open-consul/agent/consul/state/intention.go

1087 lines
32 KiB
Go
Raw Normal View History

// Copyright (c) HashiCorp, Inc.
// SPDX-License-Identifier: MPL-2.0
package state
import (
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
"errors"
"fmt"
"sort"
"github.com/hashicorp/go-memdb"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/connect"
"github.com/hashicorp/consul/agent/structs"
)
const tableConnectIntentions = "connect-intentions"
// intentionsTableSchema returns a new table schema used for storing
// intentions for Connect.
func intentionsTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: tableConnectIntentions,
Indexes: map[string]*memdb.IndexSchema{
indexID: {
Name: indexID,
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "ID",
},
},
"destination": {
Name: "destination",
AllowMissing: true,
// This index is not unique since we need uniqueness across the whole
// 4-tuple.
Unique: false,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "DestinationNS",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "DestinationName",
Lowercase: true,
},
},
},
},
"source": {
Name: "source",
AllowMissing: true,
// This index is not unique since we need uniqueness across the whole
// 4-tuple.
Unique: false,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "SourceNS",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "SourceName",
Lowercase: true,
},
},
},
},
"source_destination": {
Name: "source_destination",
AllowMissing: true,
Unique: true,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "SourceNS",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "SourceName",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "DestinationNS",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "DestinationName",
Lowercase: true,
},
},
},
},
},
}
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyIntentions is used to pull all the intentions from the snapshot.
//
// Deprecated: service-intentions config entries are handled as config entries
// in the snapshot.
func (s *Snapshot) LegacyIntentions() (structs.Intentions, error) {
ixns, err := s.tx.Get(tableConnectIntentions, "id")
if err != nil {
return nil, err
}
var ret structs.Intentions
for wrapped := ixns.Next(); wrapped != nil; wrapped = ixns.Next() {
ret = append(ret, wrapped.(*structs.Intention))
}
return ret, nil
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyIntention is used when restoring from a snapshot.
//
// Deprecated: service-intentions config entries are handled as config entries
// in the snapshot.
func (s *Restore) LegacyIntention(ixn *structs.Intention) error {
// Insert the intention
if err := s.tx.Insert(tableConnectIntentions, ixn); err != nil {
return fmt.Errorf("failed restoring intention: %s", err)
}
if err := indexUpdateMaxTxn(s.tx, ixn.ModifyIndex, tableConnectIntentions); err != nil {
return fmt.Errorf("failed updating index: %s", err)
}
return nil
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// AreIntentionsInConfigEntries determines which table is the canonical store
// for intentions data.
func (s *Store) AreIntentionsInConfigEntries() (bool, error) {
2018-02-28 17:53:21 +00:00
tx := s.db.Txn(false)
defer tx.Abort()
return areIntentionsInConfigEntries(tx, nil)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
func areIntentionsInConfigEntries(tx ReadTxn, ws memdb.WatchSet) (bool, error) {
_, entry, err := systemMetadataGetTxn(tx, ws, structs.SystemMetadataIntentionFormatKey)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return false, fmt.Errorf("failed system metadatalookup: %s", err)
}
if entry == nil {
return false, nil
}
return entry.Value == structs.SystemMetadataIntentionFormatConfigValue, nil
}
// LegacyIntentions is like Intentions() but only returns legacy intentions.
// This is exposed for migration purposes.
func (s *Store) LegacyIntentions(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta) (uint64, structs.Intentions, error) {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
tx := s.db.Txn(false)
defer tx.Abort()
idx, results, _, err := legacyIntentionsListTxn(tx, ws, entMeta)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return idx, results, err
}
// Intentions returns the list of all intentions. The boolean response value is true if it came from config entries.
func (s *Store) Intentions(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta) (uint64, structs.Intentions, bool, error) {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
tx := s.db.Txn(false)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, ws)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return 0, nil, false, err
}
if !usingConfigEntries {
return legacyIntentionsListTxn(tx, ws, entMeta)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
return configIntentionsListTxn(tx, ws, entMeta)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
2018-02-28 17:53:21 +00:00
func legacyIntentionsListTxn(tx ReadTxn, ws memdb.WatchSet, entMeta *acl.EnterpriseMeta) (uint64, structs.Intentions, bool, error) {
2018-02-28 17:53:21 +00:00
// Get the index
idx := maxIndexTxn(tx, tableConnectIntentions)
if idx < 1 {
idx = 1
}
2018-02-28 17:53:21 +00:00
iter, err := intentionListTxn(tx, entMeta)
2018-02-28 17:53:21 +00:00
if err != nil {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return 0, nil, false, fmt.Errorf("failed intention lookup: %s", err)
2018-02-28 17:53:21 +00:00
}
2018-02-28 17:53:21 +00:00
ws.Add(iter.WatchCh())
var results structs.Intentions
for ixn := iter.Next(); ixn != nil; ixn = iter.Next() {
2018-06-12 11:26:12 +00:00
results = append(results, ixn.(*structs.Intention))
2018-02-28 17:53:21 +00:00
}
2018-06-08 12:10:06 +00:00
// Sort by precedence just because that's nicer and probably what most clients
// want for presentation.
sort.Sort(structs.IntentionPrecedenceSorter(results))
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return idx, results, false, nil
2018-02-28 17:53:21 +00:00
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
var ErrLegacyIntentionsAreDisabled = errors.New("Legacy intention modifications are disabled after the config entry migration.")
func (s *Store) IntentionMutation(idx uint64, op structs.IntentionOp, mut *structs.IntentionMutation) error {
tx := s.db.WriteTxn(idx)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, nil)
if err != nil {
return err
}
if !usingConfigEntries {
return errors.New("state: IntentionMutation() is not allowed when intentions are not stored in config entries")
}
switch op {
case structs.IntentionOpCreate:
if err := s.intentionMutationLegacyCreate(tx, idx, mut.Destination, mut.Value); err != nil {
return err
}
case structs.IntentionOpUpdate:
if err := s.intentionMutationLegacyUpdate(tx, idx, mut.ID, mut.Value); err != nil {
return err
}
case structs.IntentionOpDelete:
if mut.ID == "" {
if err := s.intentionMutationDelete(tx, idx, mut.Destination, mut.Source); err != nil {
return err
}
} else {
if err := s.intentionMutationLegacyDelete(tx, idx, mut.ID); err != nil {
return err
}
}
case structs.IntentionOpUpsert:
if err := s.intentionMutationUpsert(tx, idx, mut.Destination, mut.Source, mut.Value); err != nil {
return err
}
case structs.IntentionOpDeleteAll:
// This is an internal operation initiated by the leader and is not
// exposed for general RPC use.
return fmt.Errorf("Invalid Intention mutation operation '%s'", op)
default:
return fmt.Errorf("Invalid Intention mutation operation '%s'", op)
}
return tx.Commit()
}
func (s *Store) intentionMutationLegacyCreate(
tx WriteTxn,
idx uint64,
dest structs.ServiceName,
value *structs.SourceIntention,
) error {
_, configEntry, err := configEntryTxn(tx, nil, structs.ServiceIntentions, dest.Name, &dest.EnterpriseMeta)
if err != nil {
return fmt.Errorf("service-intentions config entry lookup failed: %v", err)
}
var upsertEntry *structs.ServiceIntentionsConfigEntry
if configEntry == nil {
upsertEntry = &structs.ServiceIntentionsConfigEntry{
Kind: structs.ServiceIntentions,
Name: dest.Name,
EnterpriseMeta: dest.EnterpriseMeta,
Sources: []*structs.SourceIntention{value},
}
} else {
prevEntry := configEntry.(*structs.ServiceIntentionsConfigEntry)
if err := checkLegacyIntentionApplyAllowed(prevEntry); err != nil {
return err
}
upsertEntry = prevEntry.Clone()
upsertEntry.Sources = append(upsertEntry.Sources, value)
}
if err := upsertEntry.LegacyNormalize(); err != nil {
return err
}
if err := upsertEntry.LegacyValidate(); err != nil {
return err
}
if err := ensureConfigEntryTxn(tx, idx, false, upsertEntry); err != nil {
return err
}
return nil
}
func (s *Store) intentionMutationLegacyUpdate(
tx WriteTxn,
idx uint64,
legacyID string,
value *structs.SourceIntention,
) error {
// This variant is just for legacy UUID-based intentions.
_, prevEntry, ixn, err := s.IntentionGet(nil, legacyID)
if err != nil {
return fmt.Errorf("Intention lookup failed: %v", err)
}
if ixn == nil || prevEntry == nil {
return fmt.Errorf("Cannot modify non-existent intention: '%s'", legacyID)
}
if err := checkLegacyIntentionApplyAllowed(prevEntry); err != nil {
return err
}
upsertEntry := prevEntry.Clone()
foundMatch := upsertEntry.UpdateSourceByLegacyID(
legacyID,
value,
)
if !foundMatch {
return fmt.Errorf("Cannot modify non-existent intention: '%s'", legacyID)
}
if err := upsertEntry.LegacyNormalize(); err != nil {
return err
}
if err := upsertEntry.LegacyValidate(); err != nil {
return err
}
if err := ensureConfigEntryTxn(tx, idx, false, upsertEntry); err != nil {
return err
}
return nil
}
func (s *Store) intentionMutationDelete(
tx WriteTxn,
idx uint64,
dest structs.ServiceName,
src structs.ServiceName,
) error {
_, configEntry, err := configEntryTxn(tx, nil, structs.ServiceIntentions, dest.Name, &dest.EnterpriseMeta)
if err != nil {
return fmt.Errorf("service-intentions config entry lookup failed: %v", err)
}
if configEntry == nil {
return nil
}
prevEntry := configEntry.(*structs.ServiceIntentionsConfigEntry)
upsertEntry := prevEntry.Clone()
deleted := upsertEntry.DeleteSourceByName(src)
if !deleted {
return nil
}
if upsertEntry == nil || len(upsertEntry.Sources) == 0 {
return deleteConfigEntryTxn(
tx,
idx,
structs.ServiceIntentions,
dest.Name,
&dest.EnterpriseMeta,
)
}
if err := upsertEntry.Normalize(); err != nil {
return err
}
if err := upsertEntry.Validate(); err != nil {
return err
}
if err := ensureConfigEntryTxn(tx, idx, false, upsertEntry); err != nil {
return err
}
return nil
}
func (s *Store) intentionMutationLegacyDelete(
tx WriteTxn,
idx uint64,
legacyID string,
) error {
_, prevEntry, ixn, err := s.IntentionGet(nil, legacyID)
if err != nil {
return fmt.Errorf("Intention lookup failed: %v", err)
}
if ixn == nil || prevEntry == nil {
return fmt.Errorf("Cannot delete non-existent intention: '%s'", legacyID)
}
if err := checkLegacyIntentionApplyAllowed(prevEntry); err != nil {
return err
}
upsertEntry := prevEntry.Clone()
deleted := upsertEntry.DeleteSourceByLegacyID(legacyID)
if !deleted {
return fmt.Errorf("Cannot delete non-existent intention: '%s'", legacyID)
}
if upsertEntry == nil || len(upsertEntry.Sources) == 0 {
return deleteConfigEntryTxn(
tx,
idx,
structs.ServiceIntentions,
prevEntry.Name,
&prevEntry.EnterpriseMeta,
)
}
if err := upsertEntry.LegacyNormalize(); err != nil {
return err
}
if err := upsertEntry.LegacyValidate(); err != nil {
return err
}
if err := ensureConfigEntryTxn(tx, idx, false, upsertEntry); err != nil {
return err
}
return nil
}
func (s *Store) intentionMutationUpsert(
tx WriteTxn,
idx uint64,
dest structs.ServiceName,
src structs.ServiceName,
value *structs.SourceIntention,
) error {
// This variant is just for config-entry based intentions.
_, configEntry, err := configEntryTxn(tx, nil, structs.ServiceIntentions, dest.Name, &dest.EnterpriseMeta)
if err != nil {
return fmt.Errorf("service-intentions config entry lookup failed: %v", err)
}
var prevEntry *structs.ServiceIntentionsConfigEntry
if configEntry != nil {
prevEntry = configEntry.(*structs.ServiceIntentionsConfigEntry)
}
var upsertEntry *structs.ServiceIntentionsConfigEntry
if prevEntry == nil {
upsertEntry = &structs.ServiceIntentionsConfigEntry{
Kind: structs.ServiceIntentions,
Name: dest.Name,
EnterpriseMeta: dest.EnterpriseMeta,
Sources: []*structs.SourceIntention{value},
}
} else {
upsertEntry = prevEntry.Clone()
upsertEntry.UpsertSourceByName(src, value)
}
if err := upsertEntry.Normalize(); err != nil {
return err
}
if err := upsertEntry.Validate(); err != nil {
return err
}
if err := ensureConfigEntryTxn(tx, idx, false, upsertEntry); err != nil {
return err
}
return nil
}
func checkLegacyIntentionApplyAllowed(prevEntry *structs.ServiceIntentionsConfigEntry) error {
if prevEntry == nil {
return nil
}
if prevEntry.LegacyIDFieldsAreAllSet() {
return nil
}
sn := prevEntry.DestinationServiceName()
return fmt.Errorf("cannot use legacy intention API to edit intentions with a destination of %q after editing them via a service-intentions config entry", sn.String())
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyIntentionSet creates or updates an intention.
//
// Deprecated: Edit service-intentions config entries directly.
func (s *Store) LegacyIntentionSet(idx uint64, ixn *structs.Intention) error {
tx := s.db.WriteTxn(idx)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, nil)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return err
}
if usingConfigEntries {
return ErrLegacyIntentionsAreDisabled
}
if err := legacyIntentionSetTxn(tx, idx, ixn); err != nil {
return err
}
2020-06-02 20:34:56 +00:00
return tx.Commit()
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// legacyIntentionSetTxn is the inner method used to insert an intention with
// the proper indexes into the state store.
func legacyIntentionSetTxn(tx WriteTxn, idx uint64, ixn *structs.Intention) error {
// ID is required
if ixn.ID == "" {
return ErrMissingIntentionID
}
2018-06-12 11:26:12 +00:00
// Ensure Precedence is populated correctly on "write"
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
//nolint:staticcheck
2018-06-12 11:26:12 +00:00
ixn.UpdatePrecedence()
// Check for an existing intention
existing, err := tx.First(tableConnectIntentions, "id", ixn.ID)
if err != nil {
return fmt.Errorf("failed intention lookup: %s", err)
}
if existing != nil {
oldIxn := existing.(*structs.Intention)
ixn.CreateIndex = oldIxn.CreateIndex
ixn.CreatedAt = oldIxn.CreatedAt
} else {
ixn.CreateIndex = idx
}
ixn.ModifyIndex = idx
// Check for duplicates on the 4-tuple.
duplicate, err := tx.First(tableConnectIntentions, "source_destination",
ixn.SourceNS, ixn.SourceName, ixn.DestinationNS, ixn.DestinationName)
if err != nil {
return fmt.Errorf("failed intention lookup: %s", err)
}
if duplicate != nil {
dupIxn := duplicate.(*structs.Intention)
// Same ID is OK - this is an update
if dupIxn.ID != ixn.ID {
return fmt.Errorf("duplicate intention found: %s", dupIxn.String())
}
}
// We always force meta to be non-nil so that we its an empty map.
// This makes it easy for API responses to not nil-check this everywhere.
if ixn.Meta == nil {
ixn.Meta = make(map[string]string)
}
// Insert
if err := tx.Insert(tableConnectIntentions, ixn); err != nil {
return err
}
2021-03-08 18:02:41 +00:00
if err := tx.Insert(tableIndex, &IndexEntry{tableConnectIntentions, idx}); err != nil {
return fmt.Errorf("failed updating index: %s", err)
}
return nil
}
// IntentionGet returns the given intention by ID.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
func (s *Store) IntentionGet(ws memdb.WatchSet, id string) (uint64, *structs.ServiceIntentionsConfigEntry, *structs.Intention, error) {
tx := s.db.Txn(false)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, ws)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return 0, nil, nil, err
}
if !usingConfigEntries {
idx, ixn, err := legacyIntentionGetTxn(tx, ws, id)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return idx, nil, ixn, err
}
return configIntentionGetTxn(tx, ws, id)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
func legacyIntentionGetTxn(tx ReadTxn, ws memdb.WatchSet, id string) (uint64, *structs.Intention, error) {
// Get the table index.
idx := maxIndexTxn(tx, tableConnectIntentions)
if idx < 1 {
idx = 1
}
// Look up by its ID.
watchCh, intention, err := tx.FirstWatch(tableConnectIntentions, "id", id)
if err != nil {
return 0, nil, fmt.Errorf("failed intention lookup: %s", err)
}
ws.Add(watchCh)
// Convert the interface{} if it is non-nil
var result *structs.Intention
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
if intention != nil {
result = intention.(*structs.Intention)
}
return idx, result, nil
}
// IntentionGetExact returns the given intention by it's full unique name.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
func (s *Store) IntentionGetExact(ws memdb.WatchSet, args *structs.IntentionQueryExact) (uint64, *structs.ServiceIntentionsConfigEntry, *structs.Intention, error) {
tx := s.db.Txn(false)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, ws)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return 0, nil, nil, err
}
if !usingConfigEntries {
idx, ixn, err := s.legacyIntentionGetExactTxn(tx, ws, args)
return idx, nil, ixn, err
}
return s.configIntentionGetExactTxn(tx, ws, args)
}
func (s *Store) legacyIntentionGetExactTxn(tx ReadTxn, ws memdb.WatchSet, args *structs.IntentionQueryExact) (uint64, *structs.Intention, error) {
if err := args.Validate(); err != nil {
return 0, nil, err
}
// Get the table index.
idx := maxIndexTxn(tx, tableConnectIntentions)
if idx < 1 {
idx = 1
}
// Look up by its full name.
watchCh, intention, err := tx.FirstWatch(tableConnectIntentions, "source_destination",
args.SourceNS, args.SourceName, args.DestinationNS, args.DestinationName)
if err != nil {
return 0, nil, fmt.Errorf("failed intention lookup: %s", err)
}
ws.Add(watchCh)
// Convert the interface{} if it is non-nil
var result *structs.Intention
if intention != nil {
result = intention.(*structs.Intention)
}
return idx, result, nil
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyIntentionDelete deletes the given intention by ID.
//
// Deprecated: Edit service-intentions config entries directly.
func (s *Store) LegacyIntentionDelete(idx uint64, id string) error {
tx := s.db.WriteTxn(idx)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, nil)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return err
}
if usingConfigEntries {
return ErrLegacyIntentionsAreDisabled
}
if err := legacyIntentionDeleteTxn(tx, idx, id); err != nil {
return fmt.Errorf("failed intention delete: %s", err)
}
2020-06-02 20:34:56 +00:00
return tx.Commit()
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// legacyIntentionDeleteTxn is the inner method used to delete a legacy intention
// with the proper indexes into the state store.
func legacyIntentionDeleteTxn(tx WriteTxn, idx uint64, queryID string) error {
// Pull the query.
wrapped, err := tx.First(tableConnectIntentions, "id", queryID)
if err != nil {
return fmt.Errorf("failed intention lookup: %s", err)
}
if wrapped == nil {
return nil
}
// Delete the query and update the index.
if err := tx.Delete(tableConnectIntentions, wrapped); err != nil {
return fmt.Errorf("failed intention delete: %s", err)
}
2021-03-08 18:02:41 +00:00
if err := tx.Insert(tableIndex, &IndexEntry{tableConnectIntentions, idx}); err != nil {
return fmt.Errorf("failed updating index: %s", err)
}
return nil
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyIntentionDeleteAll deletes all legacy intentions. This is part of the
// config entry migration code.
func (s *Store) LegacyIntentionDeleteAll(idx uint64) error {
tx := s.db.WriteTxn(idx)
defer tx.Abort()
// Delete the table and update the index.
if _, err := tx.DeleteAll(tableConnectIntentions, "id"); err != nil {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return fmt.Errorf("failed intention delete-all: %s", err)
}
2021-03-08 18:02:41 +00:00
if err := tx.Insert(tableIndex, &IndexEntry{tableConnectIntentions, idx}); err != nil {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return fmt.Errorf("failed updating index: %s", err)
}
// Also bump the index for the config entry table so that
// secondaries can correctly know when they've replicated all of the service-intentions
// config entries that USED to exist in the old intentions table.
2021-03-08 18:02:41 +00:00
if err := tx.Insert(tableIndex, &IndexEntry{tableConfigEntries, idx}); err != nil {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return fmt.Errorf("failed updating index: %s", err)
}
// Also set a system metadata flag indicating the transition has occurred.
metadataEntry := &structs.SystemMetadataEntry{
Key: structs.SystemMetadataIntentionFormatKey,
Value: structs.SystemMetadataIntentionFormatConfigValue,
RaftIndex: structs.RaftIndex{
CreateIndex: idx,
ModifyIndex: idx,
},
}
if err := systemMetadataSetTxn(tx, idx, metadataEntry); err != nil {
return fmt.Errorf("failed updating system metadata key %q: %s", metadataEntry.Key, err)
}
return tx.Commit()
}
type IntentionDecisionOpts struct {
Target string
Namespace string
Partition string
Intentions structs.Intentions
MatchType structs.IntentionMatchType
DefaultDecision acl.EnforcementDecision
AllowPermissions bool
}
// IntentionDecision returns whether a connection should be allowed to a source or destination given a set of intentions.
//
// allowPermissions determines whether the presence of L7 permissions leads to a DENY decision.
// This should be false when evaluating a connection between a source and destination, but not the request that will be sent.
func (s *Store) IntentionDecision(opts IntentionDecisionOpts) (structs.IntentionDecisionSummary, error) {
// Figure out which source matches this request.
var ixnMatch *structs.Intention
for _, ixn := range opts.Intentions {
if _, ok := connect.AuthorizeIntentionTarget(opts.Target, opts.Namespace, opts.Partition, ixn, opts.MatchType); ok {
ixnMatch = ixn
break
}
}
resp := structs.IntentionDecisionSummary{
DefaultAllow: opts.DefaultDecision == acl.Allow,
}
if ixnMatch == nil {
// No intention found, fall back to default
resp.Allowed = resp.DefaultAllow
return resp, nil
}
// Intention found, combine action + permissions
resp.Allowed = ixnMatch.Action == structs.IntentionActionAllow
if len(ixnMatch.Permissions) > 0 {
2021-03-16 14:03:52 +00:00
// If any permissions are present, fall back to allowPermissions.
// We are not evaluating requests so we cannot know whether the L7 permission requirements will be met.
resp.Allowed = opts.AllowPermissions
resp.HasPermissions = true
}
resp.ExternalSource = ixnMatch.Meta[structs.MetaExternalSource]
// Intentions with wildcard namespaces but specific names are not allowed (*/web -> */api)
// So we don't check namespaces to see if there's an exact intention
if ixnMatch.SourceName != structs.WildcardSpecifier && ixnMatch.DestinationName != structs.WildcardSpecifier {
resp.HasExact = true
}
return resp, nil
}
// IntentionMatch returns the list of intentions that match the namespace and
// name for either a source or destination. This applies the resolution rules
// so wildcards will match any value.
//
// The returned value is the list of intentions in the same order as the
// entries in args. The intentions themselves are sorted based on the
// intention precedence rules. i.e. result[0][0] is the highest precedent
// rule to match for the first entry.
func (s *Store) IntentionMatch(ws memdb.WatchSet, args *structs.IntentionQueryMatch) (uint64, []structs.Intentions, error) {
tx := s.db.Txn(false)
defer tx.Abort()
usingConfigEntries, err := areIntentionsInConfigEntries(tx, ws)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return 0, nil, err
}
if !usingConfigEntries {
return s.legacyIntentionMatchTxn(tx, ws, args)
}
return s.configIntentionMatchTxn(tx, ws, args)
}
func (s *Store) legacyIntentionMatchTxn(tx ReadTxn, ws memdb.WatchSet, args *structs.IntentionQueryMatch) (uint64, []structs.Intentions, error) {
// Get the table index.
idx := maxIndexTxn(tx, tableConnectIntentions)
if idx < 1 {
idx = 1
}
// Make all the calls and accumulate the results
results := make([]structs.Intentions, len(args.Entries))
for i, entry := range args.Entries {
ixns, err := intentionMatchOneTxn(tx, ws, entry, args.Type)
if err != nil {
return 0, nil, err
}
// Sort the results by precedence
sort.Sort(structs.IntentionPrecedenceSorter(ixns))
// Store the result
results[i] = ixns
}
return idx, results, nil
}
// IntentionMatchOne returns the list of intentions that match the namespace and
// name for a single source or destination. This applies the resolution rules
// so wildcards will match any value.
//
// The returned intentions are sorted based on the intention precedence rules.
// i.e. result[0] is the highest precedent rule to match
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
func (s *Store) IntentionMatchOne(
ws memdb.WatchSet,
entry structs.IntentionMatchEntry,
matchType structs.IntentionMatchType,
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
destinationType structs.IntentionTargetType,
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
) (uint64, structs.Intentions, error) {
tx := s.db.Txn(false)
defer tx.Abort()
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
return compatIntentionMatchOneTxn(tx, ws, entry, matchType, destinationType)
2021-03-16 14:03:52 +00:00
}
func compatIntentionMatchOneTxn(
tx ReadTxn,
ws memdb.WatchSet,
entry structs.IntentionMatchEntry,
matchType structs.IntentionMatchType,
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
destinationType structs.IntentionTargetType,
2021-03-16 14:03:52 +00:00
) (uint64, structs.Intentions, error) {
usingConfigEntries, err := areIntentionsInConfigEntries(tx, ws)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if err != nil {
return 0, nil, err
}
if !usingConfigEntries {
return legacyIntentionMatchOneTxn(tx, ws, entry, matchType)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
return configIntentionMatchOneTxn(tx, ws, entry, matchType, destinationType)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
func legacyIntentionMatchOneTxn(
tx ReadTxn,
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
ws memdb.WatchSet,
entry structs.IntentionMatchEntry,
matchType structs.IntentionMatchType,
) (uint64, structs.Intentions, error) {
// Get the table index.
idx := maxIndexTxn(tx, tableConnectIntentions)
if idx < 1 {
idx = 1
}
results, err := intentionMatchOneTxn(tx, ws, entry, matchType)
if err != nil {
return 0, nil, err
}
sort.Sort(structs.IntentionPrecedenceSorter(results))
return idx, results, nil
}
func intentionMatchOneTxn(tx ReadTxn, ws memdb.WatchSet,
entry structs.IntentionMatchEntry, matchType structs.IntentionMatchType) (structs.Intentions, error) {
// Each search entry may require multiple queries to memdb, so this
// returns the arguments for each necessary Get. Note on performance:
// this is not the most optimal set of queries since we repeat some
// many times (such as */*). We can work on improving that in the
// future, the test cases shouldn't have to change for that.
getParams, err := intentionMatchGetParams(entry)
if err != nil {
return nil, err
}
// Perform each call and accumulate the result.
var result structs.Intentions
for _, params := range getParams {
iter, err := tx.Get(tableConnectIntentions, string(matchType), params...)
if err != nil {
return nil, fmt.Errorf("failed intention lookup: %s", err)
}
ws.Add(iter.WatchCh())
for ixn := iter.Next(); ixn != nil; ixn = iter.Next() {
result = append(result, ixn.(*structs.Intention))
}
}
return result, nil
}
2021-09-15 15:37:46 +00:00
// TODO(partitions): Update for partitions
// intentionMatchGetParams returns the tx.Get parameters to find all the
// intentions for a certain entry.
func intentionMatchGetParams(entry structs.IntentionMatchEntry) ([][]interface{}, error) {
// We always query for "*/*" so include that. If the namespace is a
// wildcard, then we're actually done.
result := make([][]interface{}, 0, 3)
result = append(result, []interface{}{structs.WildcardSpecifier, structs.WildcardSpecifier})
if entry.Namespace == structs.WildcardSpecifier {
return result, nil
}
// Search for NS/* intentions. If we have a wildcard name, then we're done.
result = append(result, []interface{}{entry.Namespace, structs.WildcardSpecifier})
if entry.Name == structs.WildcardSpecifier {
return result, nil
}
// Search for the exact NS/N value.
result = append(result, []interface{}{entry.Namespace, entry.Name})
return result, nil
}
type ServiceWithDecision struct {
Name structs.ServiceName
Decision structs.IntentionDecisionSummary
}
// IntentionTopology returns the upstreams or downstreams of a service. Upstreams and downstreams are inferred from
// intentions. If intentions allow a connection from the target to some candidate service, the candidate service is considered
// an upstream of the target.
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
func (s *Store) IntentionTopology(
ws memdb.WatchSet,
target structs.ServiceName,
downstreams bool,
defaultDecision acl.EnforcementDecision,
Egress gtw/intention rpc endpoint (#13354) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * change state store API to get an IntentionTarget parameter * add intentions tests * add destination upstream endpoint * fix failing test * fix failing test and a bug with wildcard intentions * fix failing test * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * add missing test and clarify doc * fix style * gofmt intention.go * fix merge introduced issue Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:55:02 +00:00
intentionTarget structs.IntentionTargetType,
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
) (uint64, structs.ServiceList, error) {
2021-03-16 14:03:52 +00:00
tx := s.db.ReadTxn()
defer tx.Abort()
Egress gtw/intention rpc endpoint (#13354) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * change state store API to get an IntentionTarget parameter * add intentions tests * add destination upstream endpoint * fix failing test * fix failing test and a bug with wildcard intentions * fix failing test * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * add missing test and clarify doc * fix style * gofmt intention.go * fix merge introduced issue Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:55:02 +00:00
idx, services, err := s.intentionTopologyTxn(tx, ws, target, downstreams, defaultDecision, intentionTarget)
if err != nil {
requested := "upstreams"
if downstreams {
requested = "downstreams"
}
return 0, nil, fmt.Errorf("failed to fetch %s for %s: %v", requested, target.String(), err)
}
2021-04-14 16:52:05 +00:00
resp := make(structs.ServiceList, 0)
for _, svc := range services {
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
resp = append(resp, structs.ServiceName{Name: svc.Name.Name, EnterpriseMeta: svc.Name.EnterpriseMeta})
}
return idx, resp, nil
}
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
func (s *Store) intentionTopologyTxn(
tx ReadTxn, ws memdb.WatchSet,
target structs.ServiceName,
downstreams bool,
defaultDecision acl.EnforcementDecision,
Egress gtw/intention rpc endpoint (#13354) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * change state store API to get an IntentionTarget parameter * add intentions tests * add destination upstream endpoint * fix failing test * fix failing test and a bug with wildcard intentions * fix failing test * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * add missing test and clarify doc * fix style * gofmt intention.go * fix merge introduced issue Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:55:02 +00:00
intentionTarget structs.IntentionTargetType,
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
) (uint64, []ServiceWithDecision, error) {
var maxIdx uint64
// If querying the upstreams for a service, we first query intentions that apply to the target service as a source.
// That way we can check whether intentions from the source allow connections to upstream candidates.
2021-03-16 15:49:24 +00:00
// The reverse is true for downstreams.
intentionMatchType := structs.IntentionMatchSource
if downstreams {
2021-03-16 15:49:24 +00:00
intentionMatchType = structs.IntentionMatchDestination
}
entry := structs.IntentionMatchEntry{
Namespace: target.NamespaceOrDefault(),
Partition: target.PartitionOrDefault(),
Name: target.Name,
}
Egress gtw/intention rpc endpoint (#13354) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * change state store API to get an IntentionTarget parameter * add intentions tests * add destination upstream endpoint * fix failing test * fix failing test and a bug with wildcard intentions * fix failing test * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * add missing test and clarify doc * fix style * gofmt intention.go * fix merge introduced issue Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:55:02 +00:00
index, intentions, err := compatIntentionMatchOneTxn(tx, ws, entry, intentionMatchType, intentionTarget)
if err != nil {
return 0, nil, fmt.Errorf("failed to query intentions for %s", target.String())
}
if index > maxIdx {
maxIdx = index
}
// TODO(tproxy): One remaining improvement is that this includes non-Connect services (typical services without a proxy)
// Ideally those should be excluded as well, since they can't be upstreams/downstreams without a proxy.
// Maybe narrow serviceNamesOfKindTxn to services represented by proxies? (ingress, sidecar-
wildcardMeta := structs.WildcardEnterpriseMetaInPartition(structs.WildcardSpecifier)
Egress gtw/intention rpc endpoint (#13354) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * change state store API to get an IntentionTarget parameter * add intentions tests * add destination upstream endpoint * fix failing test * fix failing test and a bug with wildcard intentions * fix failing test * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * add missing test and clarify doc * fix style * gofmt intention.go * fix merge introduced issue Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:55:02 +00:00
var services []*KindServiceName
if intentionTarget == structs.IntentionTargetService {
index, services, err = serviceNamesOfKindTxn(tx, ws, structs.ServiceKindTypical, *wildcardMeta)
} else {
// destinations can only ever be upstream, since they are only allowed as intention destination.
index, services, err = serviceNamesOfKindTxn(tx, ws, structs.ServiceKindDestination, *wildcardMeta)
}
if err != nil {
return index, nil, fmt.Errorf("failed to list ingress service names: %v", err)
}
if index > maxIdx {
maxIdx = index
}
if downstreams {
// Ingress gateways can only ever be downstreams, since mesh services don't dial them.
index, ingress, err := serviceNamesOfKindTxn(tx, ws, structs.ServiceKindIngressGateway, *wildcardMeta)
if err != nil {
return index, nil, fmt.Errorf("failed to list ingress service names: %v", err)
}
if index > maxIdx {
maxIdx = index
}
services = append(services, ingress...)
}
2021-03-16 15:49:24 +00:00
// When checking authorization to upstreams, the match type for the decision is `destination` because we are deciding
// if upstream candidates are covered by intentions that have the target service as a source.
// The reverse is true for downstreams.
decisionMatchType := structs.IntentionMatchDestination
if downstreams {
decisionMatchType = structs.IntentionMatchSource
}
result := make([]ServiceWithDecision, 0, len(services))
for _, svc := range services {
candidate := svc.Service
if candidate.Name == structs.ConsulServiceName {
continue
}
opts := IntentionDecisionOpts{
Target: candidate.Name,
Namespace: candidate.NamespaceOrDefault(),
Partition: candidate.PartitionOrDefault(),
Intentions: intentions,
MatchType: decisionMatchType,
DefaultDecision: defaultDecision,
AllowPermissions: true,
}
decision, err := s.IntentionDecision(opts)
if err != nil {
src, dst := target, candidate
if downstreams {
src, dst = candidate, target
}
return 0, nil, fmt.Errorf("failed to get intention decision from (%s) to (%s): %v",
src.String(), dst.String(), err)
}
if !decision.Allowed || target.Matches(candidate) {
continue
}
result = append(result, ServiceWithDecision{
Egress gtw/connect destination intentions (#13341) * update gateway-services table with endpoints * fix failing test * remove unneeded config in test * rename "endpoint" to "destination" * more endpoint renaming to destination in tests * update isDestination based on service-defaults config entry creation * use a 3 state kind to be able to set the kind to unknown (when neither a service or a destination exist) * set unknown state to empty to avoid modifying alot of tests * fix logic to set the kind correctly on CRUD * fix failing tests * add missing tests and fix service delete * fix failing test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * fix a bug with kind and add relevant test * fix compile error * fix failing tests * add kind to clone * fix failing tests * fix failing tests in catalog endpoint * fix service dump test * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * remove duplicate tests * first draft of destinations intention in connect proxy * remove ServiceDestinationList * fix failing tests * fix agent/consul failing tests * change to filter intentions in the state store instead of adding a field. * fix failing tests * fix comment * fix comments * store service kind destination and add relevant tests * changes based on review * filter on destinations when querying source match * Apply suggestions from code review Co-authored-by: alex <8968914+acpana@users.noreply.github.com> * fix style * Apply suggestions from code review Co-authored-by: Dan Stough <dan.stough@hashicorp.com> * rename destinationType to targetType. Co-authored-by: Dan Stough <dan.stough@hashicorp.com> Co-authored-by: alex <8968914+acpana@users.noreply.github.com> Co-authored-by: github-team-consul-core <github-team-consul-core@hashicorp.com>
2022-06-07 19:03:59 +00:00
Name: structs.ServiceName{Name: candidate.Name, EnterpriseMeta: candidate.EnterpriseMeta},
Decision: decision,
})
}
return maxIdx, result, err
}