open-consul/agent/proxycfg/state_test.go

1285 lines
45 KiB
Go
Raw Normal View History

package proxycfg
import (
"context"
"fmt"
"sync"
"testing"
"github.com/hashicorp/consul/agent/cache"
cachetype "github.com/hashicorp/consul/agent/cache-types"
"github.com/hashicorp/consul/agent/consul/discoverychain"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/sdk/testutil"
"github.com/stretchr/testify/require"
)
func TestStateChanged(t *testing.T) {
tests := []struct {
name string
ns *structs.NodeService
token string
mutate func(ns structs.NodeService, token string) (*structs.NodeService, string)
want bool
}{
{
name: "nil node service",
ns: structs.TestNodeServiceProxy(t),
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
return nil, token
},
want: true,
},
{
name: "same service",
ns: structs.TestNodeServiceProxy(t),
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
return &ns, token
}, want: false,
},
{
name: "same service, different token",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
return &ns, "bar"
},
want: true,
},
{
name: "different service ID",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.ID = "badger"
return &ns, token
},
want: true,
},
{
name: "different address",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.Address = "10.10.10.10"
return &ns, token
},
want: true,
},
{
name: "different port",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.Port = 12345
return &ns, token
},
want: true,
},
{
name: "different service kind",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.Kind = ""
return &ns, token
},
want: true,
},
{
name: "different proxy target",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.Proxy.DestinationServiceName = "badger"
return &ns, token
},
want: true,
},
{
name: "different proxy upstreams",
ns: structs.TestNodeServiceProxy(t),
token: "foo",
mutate: func(ns structs.NodeService, token string) (*structs.NodeService, string) {
ns.Proxy.Upstreams = nil
return &ns, token
},
want: true,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
require := require.New(t)
state, err := newState(tt.ns, tt.token)
require.NoError(err)
otherNS, otherToken := tt.mutate(*tt.ns, tt.token)
require.Equal(tt.want, state.Changed(otherNS, otherToken))
})
}
}
type testCacheNotifierRequest struct {
cacheType string
request cache.Request
ch chan<- cache.UpdateEvent
}
type testCacheNotifier struct {
lock sync.RWMutex
notifiers map[string]testCacheNotifierRequest
}
func newTestCacheNotifier() *testCacheNotifier {
return &testCacheNotifier{
notifiers: make(map[string]testCacheNotifierRequest),
}
}
func (cn *testCacheNotifier) Notify(ctx context.Context, t string, r cache.Request, correlationId string, ch chan<- cache.UpdateEvent) error {
cn.lock.Lock()
cn.notifiers[correlationId] = testCacheNotifierRequest{t, r, ch}
cn.lock.Unlock()
return nil
}
func (cn *testCacheNotifier) getNotifierRequest(t testing.TB, correlationId string) testCacheNotifierRequest {
cn.lock.RLock()
req, ok := cn.notifiers[correlationId]
cn.lock.RUnlock()
require.True(t, ok, "Correlation ID: %s is missing", correlationId)
return req
}
func (cn *testCacheNotifier) getChanForCorrelationId(t testing.TB, correlationId string) chan<- cache.UpdateEvent {
req := cn.getNotifierRequest(t, correlationId)
require.NotNil(t, req.ch)
return req.ch
}
func (cn *testCacheNotifier) sendNotification(t testing.TB, correlationId string, event cache.UpdateEvent) {
cn.getChanForCorrelationId(t, correlationId) <- event
}
func (cn *testCacheNotifier) verifyWatch(t testing.TB, correlationId string) (string, cache.Request) {
// t.Logf("Watches: %+v", cn.notifiers)
req := cn.getNotifierRequest(t, correlationId)
require.NotNil(t, req.ch)
return req.cacheType, req.request
}
type verifyWatchRequest func(t testing.TB, cacheType string, request cache.Request)
func genVerifyDCSpecificWatch(expectedCacheType string, expectedDatacenter string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, expectedCacheType, cacheType)
reqReal, ok := request.(*structs.DCSpecificRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
}
}
func genVerifyRootsWatch(expectedDatacenter string) verifyWatchRequest {
return genVerifyDCSpecificWatch(cachetype.ConnectCARootName, expectedDatacenter)
}
func genVerifyListServicesWatch(expectedDatacenter string) verifyWatchRequest {
return genVerifyDCSpecificWatch(cachetype.CatalogServiceListName, expectedDatacenter)
}
func verifyDatacentersWatch(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.CatalogDatacentersName, cacheType)
_, ok := request.(*structs.DatacentersRequest)
require.True(t, ok)
}
func genVerifyLeafWatchWithDNSSANs(expectedService string, expectedDatacenter string, expectedDNSSANs []string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.ConnectCALeafName, cacheType)
reqReal, ok := request.(*cachetype.ConnectCALeafRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.Equal(t, expectedService, reqReal.Service)
require.ElementsMatch(t, expectedDNSSANs, reqReal.DNSSAN)
}
}
func genVerifyLeafWatch(expectedService string, expectedDatacenter string) verifyWatchRequest {
return genVerifyLeafWatchWithDNSSANs(expectedService, expectedDatacenter, nil)
}
2020-04-14 14:59:23 +00:00
func genVerifyResolverWatch(expectedService, expectedDatacenter, expectedKind string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.ConfigEntriesName, cacheType)
reqReal, ok := request.(*structs.ConfigEntryQuery)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.Equal(t, expectedService, reqReal.Name)
require.Equal(t, expectedKind, reqReal.Kind)
}
}
func genVerifyIntentionWatch(expectedService string, expectedDatacenter string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.IntentionMatchName, cacheType)
reqReal, ok := request.(*structs.IntentionQueryRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.NotNil(t, reqReal.Match)
require.Equal(t, structs.IntentionMatchDestination, reqReal.Match.Type)
require.Len(t, reqReal.Match.Entries, 1)
require.Equal(t, structs.IntentionDefaultNamespace, reqReal.Match.Entries[0].Namespace)
require.Equal(t, expectedService, reqReal.Match.Entries[0].Name)
}
}
func genVerifyPreparedQueryWatch(expectedName string, expectedDatacenter string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.PreparedQueryName, cacheType)
reqReal, ok := request.(*structs.PreparedQueryExecuteRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.Equal(t, expectedName, reqReal.QueryIDOrName)
require.Equal(t, true, reqReal.Connect)
}
}
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func genVerifyDiscoveryChainWatch(expected *structs.DiscoveryChainRequest) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.CompiledDiscoveryChainName, cacheType)
reqReal, ok := request.(*structs.DiscoveryChainRequest)
require.True(t, ok)
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
require.Equal(t, expected, reqReal)
}
}
func genVerifyGatewayWatch(expectedDatacenter string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.InternalServiceDumpName, cacheType)
reqReal, ok := request.(*structs.ServiceDumpRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.True(t, reqReal.UseServiceKind)
require.Equal(t, structs.ServiceKindMeshGateway, reqReal.ServiceKind)
require.Equal(t, structs.DefaultEnterpriseMeta(), &reqReal.EnterpriseMeta)
}
}
func genVerifyServiceSpecificRequest(expectedCacheType, expectedService, expectedFilter, expectedDatacenter string, connect bool) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, expectedCacheType, cacheType)
reqReal, ok := request.(*structs.ServiceSpecificRequest)
require.True(t, ok)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
require.Equal(t, expectedService, reqReal.ServiceName)
require.Equal(t, expectedFilter, reqReal.QueryOptions.Filter)
require.Equal(t, connect, reqReal.Connect)
}
}
func genVerifyServiceWatch(expectedService, expectedFilter, expectedDatacenter string, connect bool) verifyWatchRequest {
return genVerifyServiceSpecificRequest(cachetype.HealthServicesName, expectedService, expectedFilter, expectedDatacenter, connect)
}
func genVerifyGatewayServiceWatch(expectedService, expectedDatacenter string) verifyWatchRequest {
return genVerifyServiceSpecificRequest(cachetype.GatewayServicesName, expectedService, "", expectedDatacenter, false)
}
func genVerifyConfigEntryWatch(expectedKind, expectedName, expectedDatacenter string) verifyWatchRequest {
return func(t testing.TB, cacheType string, request cache.Request) {
require.Equal(t, cachetype.ConfigEntryName, cacheType)
reqReal, ok := request.(*structs.ConfigEntryQuery)
require.True(t, ok)
require.Equal(t, expectedKind, reqReal.Kind)
require.Equal(t, expectedName, reqReal.Name)
require.Equal(t, expectedDatacenter, reqReal.Datacenter)
}
}
func ingressConfigWatchEvent(tlsEnabled bool) cache.UpdateEvent {
return cache.UpdateEvent{
CorrelationID: gatewayConfigWatchID,
Result: &structs.ConfigEntryResponse{
Entry: &structs.IngressGatewayConfigEntry{
TLS: structs.GatewayTLSConfig{
Enabled: tlsEnabled,
},
},
},
Err: nil,
}
}
// This test is meant to exercise the various parts of the cache watching done by the state as
// well as its management of the ConfigSnapshot
//
// This test is expressly not calling Watch which in turn would execute the run function in a go
// routine. This allows the test to be fully synchronous and deterministic while still being able
// to validate the logic of most of the watching and state updating.
//
// The general strategy here is to
//
// 1. Initialize a state with a call to newState + setting some of the extra stuff like the CacheNotifier
// We will not be using the CacheNotifier to send notifications but calling handleUpdate ourselves
// 2. Iterate through a list of verification stages performing validation and updates for each.
// a. Ensure that the required watches are in place and validate they are correct
// b. Process a bunch of UpdateEvents by calling handleUpdate
// c. Validate that the ConfigSnapshot has been updated appropriately
func TestState_WatchesAndUpdates(t *testing.T) {
t.Parallel()
indexedRoots, issuedCert := TestCerts(t)
rootWatchEvent := func() cache.UpdateEvent {
return cache.UpdateEvent{
CorrelationID: rootsWatchID,
Result: indexedRoots,
Err: nil,
}
}
type verificationStage struct {
requiredWatches map[string]verifyWatchRequest
events []cache.UpdateEvent
verifySnapshot func(t testing.TB, snap *ConfigSnapshot)
}
type testCase struct {
// the state to operate on. the logger, source, cache,
// ctx and cancel fields will be filled in by the test
ns structs.NodeService
sourceDC string
stages []verificationStage
}
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
newConnectProxyCase := func(meshGatewayProxyConfigValue structs.MeshGatewayMode) testCase {
ns := structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
ID: "web-sidecar-proxy",
Service: "web-sidecar-proxy",
Address: "10.0.1.1",
Port: 443,
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "web",
Upstreams: structs.Upstreams{
structs.Upstream{
DestinationType: structs.UpstreamDestTypePreparedQuery,
DestinationName: "query",
LocalBindPort: 10001,
},
structs.Upstream{
DestinationType: structs.UpstreamDestTypeService,
DestinationName: "api",
LocalBindPort: 10002,
},
structs.Upstream{
DestinationType: structs.UpstreamDestTypeService,
DestinationName: "api-failover-remote",
Datacenter: "dc2",
LocalBindPort: 10003,
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeRemote,
},
},
structs.Upstream{
DestinationType: structs.UpstreamDestTypeService,
DestinationName: "api-failover-local",
Datacenter: "dc2",
LocalBindPort: 10004,
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeLocal,
},
},
structs.Upstream{
DestinationType: structs.UpstreamDestTypeService,
DestinationName: "api-failover-direct",
Datacenter: "dc2",
LocalBindPort: 10005,
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeNone,
},
},
structs.Upstream{
DestinationType: structs.UpstreamDestTypeService,
DestinationName: "api-dc2",
LocalBindPort: 10006,
},
},
},
}
if meshGatewayProxyConfigValue != structs.MeshGatewayModeDefault {
ns.Proxy.MeshGateway.Mode = meshGatewayProxyConfigValue
}
stage0 := verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
leafWatchID: genVerifyLeafWatch("web", "dc1"),
intentionsWatchID: genVerifyIntentionWatch("web", "dc1"),
"upstream:prepared_query:query": genVerifyPreparedQueryWatch("query", "dc1"),
"discovery-chain:api": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api",
EvaluateInDatacenter: "dc1",
EvaluateInNamespace: "default",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Datacenter: "dc1",
OverrideMeshGateway: structs.MeshGatewayConfig{
Mode: meshGatewayProxyConfigValue,
},
}),
"discovery-chain:api-failover-remote?dc=dc2": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api-failover-remote",
EvaluateInDatacenter: "dc2",
EvaluateInNamespace: "default",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Datacenter: "dc1",
OverrideMeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeRemote,
},
}),
"discovery-chain:api-failover-local?dc=dc2": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api-failover-local",
EvaluateInDatacenter: "dc2",
EvaluateInNamespace: "default",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Datacenter: "dc1",
OverrideMeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeLocal,
},
}),
"discovery-chain:api-failover-direct?dc=dc2": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api-failover-direct",
EvaluateInDatacenter: "dc2",
EvaluateInNamespace: "default",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Datacenter: "dc1",
OverrideMeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeNone,
},
}),
"discovery-chain:api-dc2": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api-dc2",
EvaluateInDatacenter: "dc1",
EvaluateInNamespace: "default",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Datacenter: "dc1",
OverrideMeshGateway: structs.MeshGatewayConfig{
Mode: meshGatewayProxyConfigValue,
},
}),
},
events: []cache.UpdateEvent{
rootWatchEvent(),
cache.UpdateEvent{
CorrelationID: leafWatchID,
Result: issuedCert,
Err: nil,
},
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
cache.UpdateEvent{
CorrelationID: "discovery-chain:api",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api", "default", "dc1", "trustdomain.consul", "dc1",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = meshGatewayProxyConfigValue
}),
},
Err: nil,
},
cache.UpdateEvent{
CorrelationID: "discovery-chain:api-failover-remote?dc=dc2",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api-failover-remote", "default", "dc2", "trustdomain.consul", "dc1",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = structs.MeshGatewayModeRemote
}),
},
Err: nil,
},
cache.UpdateEvent{
CorrelationID: "discovery-chain:api-failover-local?dc=dc2",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api-failover-local", "default", "dc2", "trustdomain.consul", "dc1",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = structs.MeshGatewayModeLocal
}),
},
Err: nil,
},
cache.UpdateEvent{
CorrelationID: "discovery-chain:api-failover-direct?dc=dc2",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api-failover-direct", "default", "dc2", "trustdomain.consul", "dc1",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = structs.MeshGatewayModeNone
}),
},
Err: nil,
},
cache.UpdateEvent{
CorrelationID: "discovery-chain:api-dc2",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api-dc2", "default", "dc1", "trustdomain.consul", "dc1",
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = meshGatewayProxyConfigValue
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "api-dc2",
Redirect: &structs.ServiceResolverRedirect{
Service: "api",
Datacenter: "dc2",
},
},
),
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid())
require.True(t, snap.MeshGateway.IsEmpty())
require.Equal(t, indexedRoots, snap.Roots)
require.Equal(t, issuedCert, snap.ConnectProxy.Leaf)
require.Len(t, snap.ConnectProxy.DiscoveryChain, 5, "%+v", snap.ConnectProxy.DiscoveryChain)
require.Len(t, snap.ConnectProxy.WatchedUpstreams, 5, "%+v", snap.ConnectProxy.WatchedUpstreams)
require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 5, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
require.Len(t, snap.ConnectProxy.WatchedGateways, 5, "%+v", snap.ConnectProxy.WatchedGateways)
require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 5, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
require.Len(t, snap.ConnectProxy.WatchedServiceChecks, 0, "%+v", snap.ConnectProxy.WatchedServiceChecks)
require.Len(t, snap.ConnectProxy.PreparedQueryEndpoints, 0, "%+v", snap.ConnectProxy.PreparedQueryEndpoints)
},
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
}
stage1 := verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"upstream-target:api.default.dc1:api": genVerifyServiceWatch("api", "", "dc1", true),
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
"upstream-target:api-failover-remote.default.dc2:api-failover-remote?dc=dc2": genVerifyServiceWatch("api-failover-remote", "", "dc2", true),
"upstream-target:api-failover-local.default.dc2:api-failover-local?dc=dc2": genVerifyServiceWatch("api-failover-local", "", "dc2", true),
"upstream-target:api-failover-direct.default.dc2:api-failover-direct?dc=dc2": genVerifyServiceWatch("api-failover-direct", "", "dc2", true),
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
"mesh-gateway:dc2:api-failover-remote?dc=dc2": genVerifyGatewayWatch("dc2"),
"mesh-gateway:dc1:api-failover-local?dc=dc2": genVerifyGatewayWatch("dc1"),
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid())
require.True(t, snap.MeshGateway.IsEmpty())
require.Equal(t, indexedRoots, snap.Roots)
require.Equal(t, issuedCert, snap.ConnectProxy.Leaf)
require.Len(t, snap.ConnectProxy.DiscoveryChain, 5, "%+v", snap.ConnectProxy.DiscoveryChain)
require.Len(t, snap.ConnectProxy.WatchedUpstreams, 5, "%+v", snap.ConnectProxy.WatchedUpstreams)
require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 5, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
require.Len(t, snap.ConnectProxy.WatchedGateways, 5, "%+v", snap.ConnectProxy.WatchedGateways)
require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 5, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
require.Len(t, snap.ConnectProxy.WatchedServiceChecks, 0, "%+v", snap.ConnectProxy.WatchedServiceChecks)
require.Len(t, snap.ConnectProxy.PreparedQueryEndpoints, 0, "%+v", snap.ConnectProxy.PreparedQueryEndpoints)
},
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
if meshGatewayProxyConfigValue == structs.MeshGatewayModeLocal {
stage1.requiredWatches["mesh-gateway:dc1:api-dc2"] = genVerifyGatewayWatch("dc1")
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
}
return testCase{
ns: ns,
sourceDC: "dc1",
stages: []verificationStage{stage0, stage1},
}
}
// Used in terminating-gateway cases to account for differences in OSS/ent implementations of ServiceID.String()
db := structs.NewServiceID("db", nil)
dbStr := db.String()
cases := map[string]testCase{
"initial-gateway": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindMeshGateway,
ID: "mesh-gateway",
Service: "mesh-gateway",
Address: "10.0.1.1",
Port: 443,
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
serviceListWatchID: genVerifyListServicesWatch("dc1"),
datacentersWatchID: verifyDatacentersWatch,
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without root is not valid")
require.True(t, snap.ConnectProxy.IsEmpty())
},
},
verificationStage{
events: []cache.UpdateEvent{
rootWatchEvent(),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without services is valid")
require.True(t, snap.ConnectProxy.IsEmpty())
require.Equal(t, indexedRoots, snap.Roots)
require.Empty(t, snap.MeshGateway.WatchedServices)
require.False(t, snap.MeshGateway.WatchedServicesSet)
require.Empty(t, snap.MeshGateway.WatchedDatacenters)
require.Empty(t, snap.MeshGateway.ServiceGroups)
require.Empty(t, snap.MeshGateway.ServiceResolvers)
require.Empty(t, snap.MeshGateway.GatewayGroups)
},
},
verificationStage{
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: serviceListWatchID,
Result: &structs.IndexedServiceList{
Services: make(structs.ServiceList, 0),
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway with empty service list is valid")
require.True(t, snap.ConnectProxy.IsEmpty())
require.Equal(t, indexedRoots, snap.Roots)
require.Empty(t, snap.MeshGateway.WatchedServices)
require.True(t, snap.MeshGateway.WatchedServicesSet)
require.Empty(t, snap.MeshGateway.WatchedDatacenters)
require.Empty(t, snap.MeshGateway.ServiceGroups)
require.Empty(t, snap.MeshGateway.ServiceResolvers)
require.Empty(t, snap.MeshGateway.GatewayGroups)
},
},
},
},
"mesh-gateway-do-not-cancel-service-watches": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindMeshGateway,
ID: "mesh-gateway",
Service: "mesh-gateway",
Address: "10.0.1.1",
Port: 443,
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
serviceListWatchID: genVerifyListServicesWatch("dc1"),
datacentersWatchID: verifyDatacentersWatch,
},
events: []cache.UpdateEvent{
rootWatchEvent(),
cache.UpdateEvent{
CorrelationID: serviceListWatchID,
Result: &structs.IndexedServiceList{
Services: structs.ServiceList{
{Name: "web"},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway with service list is valid")
require.Len(t, snap.MeshGateway.WatchedServices, 1)
require.True(t, snap.MeshGateway.WatchedServicesSet)
},
},
verificationStage{
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: serviceListWatchID,
Result: &structs.IndexedServiceList{
Services: structs.ServiceList{
{Name: "web"},
{Name: "api"},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway with service list is valid")
require.Len(t, snap.MeshGateway.WatchedServices, 2)
require.True(t, snap.MeshGateway.WatchedServicesSet)
},
},
},
},
"ingress-gateway": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindIngressGateway,
ID: "ingress-gateway",
Service: "ingress-gateway",
Address: "10.0.1.1",
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
gatewayConfigWatchID: genVerifyConfigEntryWatch(structs.IngressGateway, "ingress-gateway", "dc1"),
gatewayServicesWatchID: genVerifyGatewayServiceWatch("ingress-gateway", "dc1"),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without root is not valid")
require.True(t, snap.IngressGateway.IsEmpty())
},
},
verificationStage{
events: []cache.UpdateEvent{
rootWatchEvent(),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without config entry is not valid")
require.Equal(t, indexedRoots, snap.Roots)
},
},
verificationStage{
events: []cache.UpdateEvent{
ingressConfigWatchEvent(false),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without hosts set is not valid")
require.True(t, snap.IngressGateway.TLSSet)
require.False(t, snap.IngressGateway.TLSEnabled)
},
},
verificationStage{
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{
Services: structs.GatewayServices{
{
Gateway: structs.NewServiceID("ingress-gateway", nil),
Service: structs.NewServiceID("api", nil),
Port: 9999,
Protocol: "http",
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without leaf is not valid")
require.True(t, snap.IngressGateway.HostsSet)
require.Len(t, snap.IngressGateway.Hosts, 0)
require.Len(t, snap.IngressGateway.Upstreams, 1)
key := IngressListenerKey{Protocol: "http", Port: 9999}
require.Equal(t, snap.IngressGateway.Upstreams[key], structs.Upstreams{
{
DestinationNamespace: "default",
DestinationName: "api",
LocalBindPort: 9999,
Config: map[string]interface{}{
"protocol": "http",
},
},
})
require.Len(t, snap.IngressGateway.WatchedDiscoveryChains, 1)
require.Contains(t, snap.IngressGateway.WatchedDiscoveryChains, "api")
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
leafWatchID: genVerifyLeafWatch("ingress-gateway", "dc1"),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: leafWatchID,
Result: issuedCert,
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway with root and leaf certs is valid")
require.Equal(t, issuedCert, snap.IngressGateway.Leaf)
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"discovery-chain:api": genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
Name: "api",
EvaluateInDatacenter: "dc1",
EvaluateInNamespace: "default",
Datacenter: "dc1",
}),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: "discovery-chain:api",
Result: &structs.DiscoveryChainResponse{
Chain: discoverychain.TestCompileConfigEntries(t, "api", "default", "dc1", "trustdomain.consul", "dc1", nil),
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.Len(t, snap.IngressGateway.WatchedUpstreams, 1)
require.Len(t, snap.IngressGateway.WatchedUpstreams["api"], 1)
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"upstream-target:api.default.dc1:api": genVerifyServiceWatch("api", "", "dc1", true),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: "upstream-target:api.default.dc1:api",
Result: &structs.IndexedCheckServiceNodes{
Nodes: structs.CheckServiceNodes{
{
Node: &structs.Node{
Node: "node1",
Address: "127.0.0.1",
},
Service: &structs.NodeService{
ID: "api1",
Service: "api",
},
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.Len(t, snap.IngressGateway.WatchedUpstreamEndpoints, 1)
require.Contains(t, snap.IngressGateway.WatchedUpstreamEndpoints, "api")
require.Len(t, snap.IngressGateway.WatchedUpstreamEndpoints["api"], 1)
require.Contains(t, snap.IngressGateway.WatchedUpstreamEndpoints["api"], "api.default.dc1")
require.Equal(t, snap.IngressGateway.WatchedUpstreamEndpoints["api"]["api.default.dc1"],
structs.CheckServiceNodes{
{
Node: &structs.Node{
Node: "node1",
Address: "127.0.0.1",
},
Service: &structs.NodeService{
ID: "api1",
Service: "api",
},
},
},
)
},
},
},
},
"ingress-gateway-with-tls-update-upstreams": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindIngressGateway,
ID: "ingress-gateway",
Service: "ingress-gateway",
Address: "10.0.1.1",
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
gatewayConfigWatchID: genVerifyConfigEntryWatch(structs.IngressGateway, "ingress-gateway", "dc1"),
gatewayServicesWatchID: genVerifyGatewayServiceWatch("ingress-gateway", "dc1"),
},
events: []cache.UpdateEvent{
rootWatchEvent(),
ingressConfigWatchEvent(true),
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{
Services: structs.GatewayServices{
{
Gateway: structs.NewServiceID("ingress-gateway", nil),
Service: structs.NewServiceID("api", nil),
Hosts: []string{"test.example.com"},
Port: 9999,
},
},
},
Err: nil,
},
cache.UpdateEvent{
CorrelationID: leafWatchID,
Result: issuedCert,
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid())
require.True(t, snap.IngressGateway.TLSSet)
require.True(t, snap.IngressGateway.TLSEnabled)
require.True(t, snap.IngressGateway.HostsSet)
require.Len(t, snap.IngressGateway.Hosts, 1)
require.Len(t, snap.IngressGateway.Upstreams, 1)
require.Len(t, snap.IngressGateway.WatchedDiscoveryChains, 1)
require.Contains(t, snap.IngressGateway.WatchedDiscoveryChains, "api")
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
leafWatchID: genVerifyLeafWatchWithDNSSANs("ingress-gateway", "dc1", []string{
"test.example.com",
"*.ingress.consul.",
"*.ingress.dc1.consul.",
"*.ingress.alt.consul.",
"*.ingress.dc1.alt.consul.",
}),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid())
require.Len(t, snap.IngressGateway.Upstreams, 0)
require.Len(t, snap.IngressGateway.WatchedDiscoveryChains, 0)
require.NotContains(t, snap.IngressGateway.WatchedDiscoveryChains, "api")
},
},
},
},
"terminating-gateway-initial": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindTerminatingGateway,
ID: "terminating-gateway",
Service: "terminating-gateway",
Address: "10.0.1.1",
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
gatewayServicesWatchID: genVerifyServiceSpecificRequest(gatewayServicesWatchID,
"terminating-gateway", "", "dc1", false),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.False(t, snap.Valid(), "gateway without root is not valid")
require.True(t, snap.ConnectProxy.IsEmpty())
2020-04-20 19:42:33 +00:00
require.True(t, snap.MeshGateway.IsEmpty())
require.True(t, snap.IngressGateway.IsEmpty())
},
},
verificationStage{
events: []cache.UpdateEvent{
rootWatchEvent(),
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway without services is valid")
require.True(t, snap.ConnectProxy.IsEmpty())
2020-04-20 19:42:33 +00:00
require.True(t, snap.MeshGateway.IsEmpty())
require.True(t, snap.IngressGateway.IsEmpty())
2020-04-24 00:16:09 +00:00
require.True(t, snap.TerminatingGateway.IsEmpty())
require.Equal(t, indexedRoots, snap.Roots)
},
},
},
},
"terminating-gateway-handle-update": testCase{
ns: structs.NodeService{
Kind: structs.ServiceKindTerminatingGateway,
ID: "terminating-gateway",
Service: "terminating-gateway",
Address: "10.0.1.1",
},
sourceDC: "dc1",
stages: []verificationStage{
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
rootsWatchID: genVerifyRootsWatch("dc1"),
gatewayServicesWatchID: genVerifyServiceSpecificRequest(gatewayServicesWatchID,
"terminating-gateway", "", "dc1", false),
},
events: []cache.UpdateEvent{
rootWatchEvent(),
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{
Services: structs.GatewayServices{
{
Service: structs.NewServiceID("db", nil),
Gateway: structs.NewServiceID("terminating-gateway", nil),
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.True(t, snap.Valid(), "gateway with service list is valid")
require.Len(t, snap.TerminatingGateway.WatchedServices, 1)
},
},
verificationStage{
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{
Services: structs.GatewayServices{
{
Service: structs.NewServiceID("db", nil),
Gateway: structs.NewServiceID("terminating-gateway", nil),
},
{
Service: structs.NewServiceID("billing", nil),
Gateway: structs.NewServiceID("terminating-gateway", nil),
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
db := structs.NewServiceID("db", nil)
billing := structs.NewServiceID("billing", nil)
require.True(t, snap.Valid(), "gateway with service list is valid")
require.Len(t, snap.TerminatingGateway.WatchedServices, 2)
require.Contains(t, snap.TerminatingGateway.WatchedServices, db)
require.Contains(t, snap.TerminatingGateway.WatchedServices, billing)
require.Len(t, snap.TerminatingGateway.WatchedIntentions, 2)
require.Contains(t, snap.TerminatingGateway.WatchedIntentions, db)
require.Contains(t, snap.TerminatingGateway.WatchedIntentions, billing)
require.Len(t, snap.TerminatingGateway.WatchedLeaves, 2)
require.Contains(t, snap.TerminatingGateway.WatchedLeaves, db)
require.Contains(t, snap.TerminatingGateway.WatchedLeaves, billing)
2020-04-14 14:59:23 +00:00
require.Len(t, snap.TerminatingGateway.WatchedResolvers, 2)
require.Contains(t, snap.TerminatingGateway.WatchedResolvers, db)
require.Contains(t, snap.TerminatingGateway.WatchedResolvers, billing)
require.Len(t, snap.TerminatingGateway.GatewayServices, 2)
require.Contains(t, snap.TerminatingGateway.GatewayServices, db)
require.Contains(t, snap.TerminatingGateway.GatewayServices, billing)
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"external-service:" + dbStr: genVerifyServiceWatch("db", "", "dc1", false),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: "external-service:" + dbStr,
Result: &structs.IndexedCheckServiceNodes{
Nodes: structs.CheckServiceNodes{
{
Node: &structs.Node{
Node: "node1",
Address: "127.0.0.1",
},
Service: &structs.NodeService{
ID: "db",
Service: "db",
},
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.Len(t, snap.TerminatingGateway.ServiceGroups, 1)
require.Equal(t, snap.TerminatingGateway.ServiceGroups[structs.NewServiceID("db", nil)],
structs.CheckServiceNodes{
{
Node: &structs.Node{
Node: "node1",
Address: "127.0.0.1",
},
Service: &structs.NodeService{
ID: "db",
Service: "db",
},
},
},
)
},
},
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"service-leaf:" + dbStr: genVerifyLeafWatch("db", "dc1"),
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: "service-leaf:" + dbStr,
Result: issuedCert,
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
require.Equal(t, snap.TerminatingGateway.ServiceLeaves[structs.NewServiceID("db", nil)], issuedCert)
},
},
2020-04-14 14:59:23 +00:00
verificationStage{
requiredWatches: map[string]verifyWatchRequest{
"service-resolver:" + dbStr: genVerifyResolverWatch("db", "dc1", structs.ServiceResolver),
2020-04-14 14:59:23 +00:00
},
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: "service-resolver:" + dbStr,
2020-04-14 14:59:23 +00:00
Result: &structs.IndexedConfigEntries{
Kind: structs.ServiceResolver,
Entries: []structs.ConfigEntry{
&structs.ServiceResolverConfigEntry{
Name: "db",
Kind: structs.ServiceResolver,
Redirect: &structs.ServiceResolverRedirect{
Service: "db",
Datacenter: "dc2",
},
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
want := &structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
Redirect: &structs.ServiceResolverRedirect{
Service: "db",
Datacenter: "dc2",
},
}
require.Equal(t, want, snap.TerminatingGateway.ServiceResolvers[structs.NewServiceID("db", nil)])
},
},
verificationStage{
events: []cache.UpdateEvent{
cache.UpdateEvent{
CorrelationID: gatewayServicesWatchID,
Result: &structs.IndexedGatewayServices{
Services: structs.GatewayServices{
{
Service: structs.NewServiceID("billing", nil),
Gateway: structs.NewServiceID("terminating-gateway", nil),
},
},
},
Err: nil,
},
},
verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
billing := structs.NewServiceID("billing", nil)
require.True(t, snap.Valid(), "gateway with service list is valid")
2020-04-20 19:33:19 +00:00
// All the watches should have been cancelled for db
require.Len(t, snap.TerminatingGateway.WatchedServices, 1)
require.Contains(t, snap.TerminatingGateway.WatchedServices, billing)
require.Len(t, snap.TerminatingGateway.WatchedIntentions, 1)
require.Contains(t, snap.TerminatingGateway.WatchedIntentions, billing)
require.Len(t, snap.TerminatingGateway.WatchedLeaves, 1)
require.Contains(t, snap.TerminatingGateway.WatchedLeaves, billing)
2020-04-14 14:59:23 +00:00
require.Len(t, snap.TerminatingGateway.WatchedResolvers, 1)
require.Contains(t, snap.TerminatingGateway.WatchedResolvers, billing)
require.Len(t, snap.TerminatingGateway.GatewayServices, 1)
require.Contains(t, snap.TerminatingGateway.GatewayServices, billing)
// There was no update event for billing's leaf/endpoints, so length is 0
require.Len(t, snap.TerminatingGateway.ServiceGroups, 0)
require.Len(t, snap.TerminatingGateway.ServiceLeaves, 0)
2020-04-14 14:59:23 +00:00
require.Len(t, snap.TerminatingGateway.ServiceResolvers, 0)
},
},
},
},
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
"connect-proxy": newConnectProxyCase(structs.MeshGatewayModeDefault),
"connect-proxy-mesh-gateway-local": newConnectProxyCase(structs.MeshGatewayModeLocal),
}
for name, tc := range cases {
t.Run(name, func(t *testing.T) {
state, err := newState(&tc.ns, "")
// verify building the initial state worked
require.NoError(t, err)
require.NotNil(t, state)
// setup the test logger to use the t.Log
state.logger = testutil.Logger(t)
// setup a new testing cache notifier
cn := newTestCacheNotifier()
state.cache = cn
// setup the local datacenter information
state.source = &structs.QuerySource{
Datacenter: tc.sourceDC,
}
state.dnsConfig = DNSConfig{
Domain: "consul.",
AltDomain: "alt.consul.",
}
// setup the ctx as initWatches expects this to be there
state.ctx, state.cancel = context.WithCancel(context.Background())
// ensure the initial watch setup did not error
require.NoError(t, state.initWatches())
// get the initial configuration snapshot
snap := state.initialConfigSnapshot()
//--------------------------------------------------------------------
//
// All the nested subtests here are to make failures easier to
// correlate back with the test table
//
//--------------------------------------------------------------------
for idx, stage := range tc.stages {
require.True(t, t.Run(fmt.Sprintf("stage-%d", idx), func(t *testing.T) {
for correlationId, verifier := range stage.requiredWatches {
require.True(t, t.Run(correlationId, func(t *testing.T) {
// verify that the watch was initiated
cacheType, request := cn.verifyWatch(t, correlationId)
// run the verifier if any
if verifier != nil {
verifier(t, cacheType, request)
}
}))
}
// the state is not currently executing the run method in a goroutine
// therefore we just tell it about the updates
for eveIdx, event := range stage.events {
require.True(t, t.Run(fmt.Sprintf("update-%d", eveIdx), func(t *testing.T) {
require.NoError(t, state.handleUpdate(event, &snap))
}))
}
// verify the snapshot
if stage.verifySnapshot != nil {
stage.verifySnapshot(t, &snap)
}
}))
}
})
}
}