open-consul/agent/consul/config.go

555 lines
20 KiB
Go
Raw Normal View History

2013-12-06 23:43:07 +00:00
package consul
import (
"fmt"
2013-12-06 23:43:07 +00:00
"io"
2013-12-31 22:00:25 +00:00
"net"
2013-12-06 23:43:07 +00:00
"os"
2014-01-09 23:44:25 +00:00
"time"
2017-12-12 00:38:52 +00:00
"github.com/hashicorp/consul/agent/consul/autopilot"
2018-04-09 04:56:46 +00:00
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/lib"
"github.com/hashicorp/consul/tlsutil"
2017-01-18 06:20:11 +00:00
"github.com/hashicorp/consul/types"
build: make tests independent of build tags When the metadata server is scanning the agents for potential servers it is parsing the version number which the agent provided when it joined. This version number has to conform to a certain format, i.e. 'n.n.n'. Without this version number properly set some tests fail with error messages that disguise the root cause. The default version number is currently set to 'unknown' in version/version.go which does not parse and triggers the tests to fail. The work around is to use a build tag 'consul' which will use the version number set in version_base.go instead which has the correct format and is set to the current release version. In addition, some parts of the code also require the version number to be of a certain value. Setting it to '0.0.0' for example makes some tests pass and others fail since they don't pass the semantic check. When using go build/install/test one has to remember to use '-tags consul' or tests will fail with non-obvious error messages. Using build tags makes the build process more complex and error prone since it prevents the use of the plain go toolchain and - at least in its current form - introduces subtle build and test issues. We should try to eliminate build tags for anything else but platform specific code. This patch removes all references to specific version numbers in the code and tests and sets the default version to '9.9.9' which is syntactically correct and passes the semantic check. This solves the issue of running go build/install/test without tags for the OSS build.
2017-08-28 10:20:21 +00:00
"github.com/hashicorp/consul/version"
"github.com/hashicorp/memberlist"
"github.com/hashicorp/raft"
"github.com/hashicorp/serf/serf"
"golang.org/x/time/rate"
2013-12-06 23:43:07 +00:00
)
const (
2013-12-20 23:33:13 +00:00
DefaultDC = "dc1"
2017-05-03 19:02:01 +00:00
DefaultRPCPort = 8300
2013-12-09 22:22:23 +00:00
DefaultLANSerfPort = 8301
DefaultWANSerfPort = 8302
// DefaultRaftMultiplier is used as a baseline Raft configuration that
// will be reliable on a very basic server. See docs/install/performance.html
// for information on how this value was obtained.
DefaultRaftMultiplier uint = 5
// MaxRaftMultiplier is a fairly arbitrary upper bound that limits the
// amount of performance detuning that's possible.
MaxRaftMultiplier uint = 10
2013-12-06 23:43:07 +00:00
)
2014-01-01 00:45:13 +00:00
var (
2017-05-03 19:02:01 +00:00
DefaultRPCAddr = &net.TCPAddr{IP: net.ParseIP("0.0.0.0"), Port: DefaultRPCPort}
2014-01-01 00:45:13 +00:00
2017-05-03 19:02:01 +00:00
// ProtocolVersionMap is the mapping of Consul protocol versions
// to Serf protocol versions. We mask the Serf protocols using
// our own protocol version.
protocolVersionMap map[uint8]uint8
)
func init() {
protocolVersionMap = map[uint8]uint8{
1: 4,
2: 4,
3: 4,
}
}
// (Enterprise-only) NetworkSegment is the address and port configuration
// for a network segment.
type NetworkSegment struct {
Name string
Bind string
Port int
Advertise string
RPCAddr *net.TCPAddr
SerfConfig *serf.Config
}
2013-12-06 23:43:07 +00:00
// Config is used to configure the server
type Config struct {
// Bootstrap mode is used to bring up the first Consul server.
// It is required so that it can elect a leader without any
// other nodes being present
Bootstrap bool
// BootstrapExpect mode is used to automatically bring up a collection of
// Consul servers. This can be used to automatically bring up a collection
// of nodes.
BootstrapExpect int
// Datacenter is the datacenter this Consul server represents.
2013-12-06 23:43:07 +00:00
Datacenter string
// PrimaryDatacenter is the authoritative datacenter for features like ACLs
// and Connect.
PrimaryDatacenter string
// DataDir is the directory to store our state in.
2013-12-06 23:43:07 +00:00
DataDir string
2015-11-29 04:40:05 +00:00
// DevMode is used to enable a development server mode.
DevMode bool
2017-01-18 06:20:11 +00:00
// NodeID is a unique identifier for this node across space and time.
NodeID types.NodeID
2013-12-06 23:43:07 +00:00
// Node name is the name we use to advertise. Defaults to hostname.
NodeName string
2015-05-11 22:15:36 +00:00
// Domain is the DNS domain for the records. Defaults to "consul."
Domain string
2013-12-06 23:43:07 +00:00
// RaftConfig is the configuration used for Raft in the local DC
RaftConfig *raft.Config
2017-03-21 23:36:44 +00:00
// (Enterprise-only) NonVoter is used to prevent this server from being added
// as a voting member of the Raft cluster.
NonVoter bool
2017-06-25 19:36:03 +00:00
// NotifyListen is called after the RPC listener has been configured.
// RPCAdvertise will be set to the listener address if it hasn't been
// configured at this point.
NotifyListen func()
2013-12-07 00:35:13 +00:00
// RPCAddr is the RPC address used by Consul. This should be reachable
// by the WAN and LAN
2014-01-01 00:45:13 +00:00
RPCAddr *net.TCPAddr
2013-12-07 00:35:13 +00:00
2013-12-31 22:00:25 +00:00
// RPCAdvertise is the address that is advertised to other nodes for
// the RPC endpoint. This can differ from the RPC address, if for example
// the RPCAddr is unspecified "0.0.0.0:8300", but this address must be
2017-06-25 19:36:03 +00:00
// reachable. If RPCAdvertise is nil then it will be set to the Listener
// address after the listening socket is configured.
2013-12-31 23:44:27 +00:00
RPCAdvertise *net.TCPAddr
2013-12-31 22:00:25 +00:00
// RPCSrcAddr is the source address for outgoing RPC connections.
RPCSrcAddr *net.TCPAddr
// (Enterprise-only) The network segment this agent is part of.
Segment string
// (Enterprise-only) Segments is a list of network segments for a server to
// bind on.
Segments []NetworkSegment
2013-12-07 01:18:09 +00:00
// SerfLANConfig is the configuration for the intra-dc serf
SerfLANConfig *serf.Config
2013-12-06 23:43:07 +00:00
2013-12-07 01:18:09 +00:00
// SerfWANConfig is the configuration for the cross-dc serf
SerfWANConfig *serf.Config
2013-12-06 23:43:07 +00:00
2017-03-15 19:26:54 +00:00
// SerfFloodInterval controls how often we attempt to flood local Serf
// Consul servers into the global areas (WAN and user-defined areas in
// Consul Enterprise).
SerfFloodInterval time.Duration
2014-01-09 23:44:25 +00:00
// ReconcileInterval controls how often we reconcile the strongly
// consistent store with the Serf info. This is used to handle nodes
// that are force removed, as well as intermittent unavailability during
// leader election.
ReconcileInterval time.Duration
2013-12-06 23:43:07 +00:00
// LogOutput is the location to write logs to. If this is not set,
// logs will go to stderr.
LogOutput io.Writer
// ProtocolVersion is the protocol version to speak. This must be between
// ProtocolVersionMin and ProtocolVersionMax.
ProtocolVersion uint8
2014-04-04 23:22:02 +00:00
// VerifyIncoming is used to verify the authenticity of incoming connections.
// This means that TCP requests are forbidden, only allowing for TLS. TLS connections
// must match a provided certificate authority. This can be used to force client auth.
VerifyIncoming bool
// VerifyOutgoing is used to force verification of the authenticity of outgoing connections.
2014-04-04 23:22:02 +00:00
// This means that TLS requests are used, and TCP requests are not made. TLS connections
// must match a provided certificate authority.
2014-04-04 23:22:02 +00:00
VerifyOutgoing bool
// UseTLS is used to enable TLS for outgoing connections to other TLS-capable Consul
// servers. This doesn't imply any verification, it only enables TLS if possible.
UseTLS bool
2015-05-11 22:15:36 +00:00
// VerifyServerHostname is used to enable hostname verification of servers. This
// ensures that the certificate presented is valid for server.<datacenter>.<domain>.
// This prevents a compromised client from being restarted as a server, and then
// intercepting request traffic as well as being added as a raft peer. This should be
// enabled by default with VerifyOutgoing, but for legacy reasons we cannot break
// existing clients.
VerifyServerHostname bool
2014-04-04 23:22:02 +00:00
// CAFile is a path to a certificate authority file. This is used with VerifyIncoming
// or VerifyOutgoing to verify the TLS connection.
CAFile string
// CAPath is a path to a directory of certificate authority files. This is used with
// VerifyIncoming or VerifyOutgoing to verify the TLS connection.
CAPath string
2014-04-04 23:22:02 +00:00
// CertFile is used to provide a TLS certificate that is used for serving TLS connections.
// Must be provided to serve TLS connections.
CertFile string
// KeyFile is used to provide a TLS key that is used for serving TLS connections.
// Must be provided to serve TLS connections.
KeyFile string
// ServerName is used with the TLS certificate to ensure the name we
// provide matches the certificate
ServerName string
2017-02-01 20:52:04 +00:00
// TLSMinVersion is used to set the minimum TLS version used for TLS connections.
TLSMinVersion string
// TLSCipherSuites is used to specify the list of supported ciphersuites.
TLSCipherSuites []uint16
// TLSPreferServerCipherSuites specifies whether to prefer the server's ciphersuite
// over the client ciphersuites.
TLSPreferServerCipherSuites bool
// RejoinAfterLeave controls our interaction with Serf.
// When set to false (default), a leave causes a Consul to not rejoin
// the cluster until an explicit join is received. If this is set to
// true, we ignore the leave, and rejoin the cluster on start.
RejoinAfterLeave bool
2014-06-06 22:36:40 +00:00
// Build is a string that is gossiped around, and can be used to help
// operators track which versions are actively deployed
Build string
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLEnabled is used to enable ACLs
ACLsEnabled bool
// ACLEnforceVersion8 is used to gate a set of ACL policy features that
// are opt-in prior to Consul 0.8 and opt-out in Consul 0.8 and later.
ACLEnforceVersion8 bool
2014-08-05 22:36:08 +00:00
// ACLMasterToken is used to bootstrap the ACL system. It should be specified
// on the servers in the ACLDatacenter. When the leader comes online, it ensures
// that the Master token is available. This provides the initial token.
ACLMasterToken string
2014-08-05 22:20:35 +00:00
// ACLDatacenter provides the authoritative datacenter for ACL
// tokens. If not provided, ACL verification is disabled.
ACLDatacenter string
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLTokenTTL controls the time-to-live of cached ACL tokens.
// It can be set to zero to disable caching, but this adds
// a substantial cost.
ACLTokenTTL time.Duration
// ACLPolicyTTL controls the time-to-live of cached ACL policies.
2014-08-05 22:20:35 +00:00
// It can be set to zero to disable caching, but this adds
// a substantial cost.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
ACLPolicyTTL time.Duration
// ACLRoleTTL controls the time-to-live of cached ACL roles.
// It can be set to zero to disable caching, but this adds
// a substantial cost.
ACLRoleTTL time.Duration
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLDisabledTTL is the time between checking if ACLs should be
// enabled. This
ACLDisabledTTL time.Duration
// ACLTokenReplication is used to enabled token replication.
//
// By default policy-only replication is enabled. When token
// replication is off and the primary datacenter is not
// yet upgraded to the new ACLs no replication will be performed
ACLTokenReplication bool
2014-08-05 22:20:35 +00:00
// ACLDefaultPolicy is used to control the ACL interaction when
// there is no defined policy. This can be "allow" which means
// ACLs are used to black-list, or "deny" which means ACLs are
// white-lists.
ACLDefaultPolicy string
// ACLDownPolicy controls the behavior of ACLs if the ACLDatacenter
// cannot be contacted. It can be either "deny" to deny all requests,
// "extend-cache" or "async-cache" which ignores the ACLCacheInterval and
// uses cached policies.
// If a policy is not in the cache, it acts like deny.
2014-08-05 22:20:35 +00:00
// "allow" can be used to allow all requests. This is not recommended.
ACLDownPolicy string
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLReplicationRate is the max number of replication rounds that can
// be run per second. Note that either 1 or 2 RPCs are used during each replication
// round
ACLReplicationRate int
2016-08-03 05:04:11 +00:00
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLReplicationBurst is how many replication RPCs can be bursted after a
// period of idleness
ACLReplicationBurst int
2016-08-03 05:04:11 +00:00
// ACLReplicationApplyLimit is the max number of replication-related
// apply operations that we allow during a one second period. This is
// used to limit the amount of Raft bandwidth used for replication.
ACLReplicationApplyLimit int
// ACLEnableKeyListPolicy is used to gate enforcement of the new "list" policy that
// protects listing keys by prefix. This behavior is opt-in
// by default in Consul 1.0 and later.
ACLEnableKeyListPolicy bool
// TombstoneTTL is used to control how long KV tombstones are retained.
2014-12-01 05:05:18 +00:00
// This provides a window of time where the X-Consul-Index is monotonic.
// Outside this window, the index may not be monotonic. This is a result
// of a few trade offs:
// 1) The index is defined by the data view and not globally. This is a
// performance optimization that prevents any write from incrementing the
// index for all data views.
// 2) Tombstones are not kept indefinitely, since otherwise storage required
// is also monotonic. This prevents deletes from reducing the disk space
// used.
// In theory, neither of these are intrinsic limitations, however for the
2015-09-15 12:22:08 +00:00
// purposes of building a practical system, they are reasonable trade offs.
2014-12-01 05:05:18 +00:00
//
// It is also possible to set this to an incredibly long time, thereby
// simulating infinite retention. This is not recommended however.
//
TombstoneTTL time.Duration
// TombstoneTTLGranularity is used to control how granular the timers are
// for the Tombstone GC. This is used to batch the GC of many keys together
// to reduce overhead. It is unlikely a user would ever need to tune this.
TombstoneTTLGranularity time.Duration
2014-12-01 05:05:18 +00:00
// Minimum Session TTL
SessionTTLMin time.Duration
// maxTokenExpirationDuration is the maximum difference allowed between
// ACLToken CreateTime and ExpirationTime values if ExpirationTime is set
// on a token.
ACLTokenMaxExpirationTTL time.Duration
// ACLTokenMinExpirationTTL is the minimum difference allowed between
// ACLToken CreateTime and ExpirationTime values if ExpirationTime is set
// on a token.
ACLTokenMinExpirationTTL time.Duration
// ServerUp callback can be used to trigger a notification that
// a Consul server is now up and known about.
ServerUp func()
// UserEventHandler callback can be used to handle incoming
// user events. This function should not block.
UserEventHandler func(serf.UserEvent)
2015-05-08 08:31:34 +00:00
// ConfigReplicationRate is the max number of replication rounds that can
// be run per second. Note that either 1 or 2 RPCs are used during each replication
// round
ConfigReplicationRate int
// ConfigReplicationBurst is how many replication rounds can be bursted after a
// period of idleness
ConfigReplicationBurst int
// ConfigReplicationApply limit is the max number of replication-related
// apply operations that we allow during a one second period. This is
// used to limit the amount of Raft bandwidth used for replication.
ConfigReplicationApplyLimit int
// CoordinateUpdatePeriod controls how long a server batches coordinate
// updates before applying them in a Raft transaction. A larger period
// leads to fewer Raft transactions, but also the stored coordinates
// being more stale.
2015-05-14 01:22:34 +00:00
CoordinateUpdatePeriod time.Duration
// CoordinateUpdateBatchSize controls the maximum number of updates a
// server batches before applying them in a Raft transaction.
CoordinateUpdateBatchSize int
// CoordinateUpdateMaxBatches controls the maximum number of batches we
// are willing to apply in one period. After this limit we will issue a
// warning and discard the remaining updates.
CoordinateUpdateMaxBatches int
// RPCHoldTimeout is how long an RPC can be "held" before it is errored.
// This is used to paper over a loss of leadership by instead holding RPCs,
// so that the caller experiences a slow response rather than an error.
// This period is meant to be long enough for a leader election to take
// place, and a small jitter is applied to avoid a thundering herd.
RPCHoldTimeout time.Duration
// RPCRate and RPCMaxBurst control how frequently RPC calls are allowed
// to happen. In any large enough time interval, rate limiter limits the
// rate to RPCRate tokens per second, with a maximum burst size of
// RPCMaxBurst events. As a special case, if RPCRate == Inf (the infinite
// rate), RPCMaxBurst is ignored.
//
// See https://en.wikipedia.org/wiki/Token_bucket for more about token
// buckets.
RPCRate rate.Limit
RPCMaxBurst int
// LeaveDrainTime is used to wait after a server has left the LAN Serf
// pool for RPCs to drain and new requests to be sent to other servers.
LeaveDrainTime time.Duration
// AutopilotConfig is used to apply the initial autopilot config when
// bootstrapping.
2017-12-12 00:38:52 +00:00
AutopilotConfig *autopilot.Config
// ServerHealthInterval is the frequency with which the health of the
// servers in the cluster will be updated.
ServerHealthInterval time.Duration
// AutopilotInterval is the frequency with which the leader will perform
// autopilot tasks, such as promoting eligible non-voters and removing
// dead servers.
AutopilotInterval time.Duration
2018-04-09 04:56:46 +00:00
// ConnectEnabled is whether to enable Connect features such as the CA.
ConnectEnabled bool
2018-04-09 04:56:46 +00:00
// CAConfig is used to apply the initial Connect CA configuration when
// bootstrapping.
CAConfig *structs.CAConfiguration
// ConfigEntryBootstrap contains a list of ConfigEntries to ensure are created
// If entries of the same Kind/Name exist already these will not update them.
ConfigEntryBootstrap []structs.ConfigEntry
2013-12-06 23:43:07 +00:00
}
func (c *Config) ToTLSUtilConfig() tlsutil.Config {
return tlsutil.Config{
VerifyIncoming: c.VerifyIncoming,
VerifyOutgoing: c.VerifyOutgoing,
CAFile: c.CAFile,
CAPath: c.CAPath,
CertFile: c.CertFile,
KeyFile: c.KeyFile,
NodeName: c.NodeName,
ServerName: c.ServerName,
TLSMinVersion: c.TLSMinVersion,
CipherSuites: c.TLSCipherSuites,
PreferServerCipherSuites: c.TLSPreferServerCipherSuites,
}
}
2017-05-03 19:02:01 +00:00
// CheckProtocolVersion validates the protocol version.
func (c *Config) CheckProtocolVersion() error {
if c.ProtocolVersion < ProtocolVersionMin {
2017-05-03 19:02:01 +00:00
return fmt.Errorf("Protocol version '%d' too low. Must be in range: [%d, %d]", c.ProtocolVersion, ProtocolVersionMin, ProtocolVersionMax)
}
if c.ProtocolVersion > ProtocolVersionMax {
return fmt.Errorf("Protocol version '%d' too high. Must be in range: [%d, %d]", c.ProtocolVersion, ProtocolVersionMin, ProtocolVersionMax)
}
return nil
}
2017-05-03 19:02:01 +00:00
// CheckACL validates the ACL configuration.
2014-08-05 22:20:35 +00:00
func (c *Config) CheckACL() error {
switch c.ACLDefaultPolicy {
case "allow":
case "deny":
default:
return fmt.Errorf("Unsupported default ACL policy: %s", c.ACLDefaultPolicy)
}
switch c.ACLDownPolicy {
case "allow":
case "deny":
case "async-cache", "extend-cache":
2014-08-05 22:20:35 +00:00
default:
return fmt.Errorf("Unsupported down ACL policy: %s", c.ACLDownPolicy)
}
return nil
}
2017-05-03 19:02:01 +00:00
// DefaultConfig returns a sane default configuration.
2013-12-06 23:43:07 +00:00
func DefaultConfig() *Config {
hostname, err := os.Hostname()
if err != nil {
panic(err)
}
conf := &Config{
Build: version.Version,
Datacenter: DefaultDC,
NodeName: hostname,
RPCAddr: DefaultRPCAddr,
RaftConfig: raft.DefaultConfig(),
SerfLANConfig: lib.SerfDefaultConfig(),
SerfWANConfig: lib.SerfDefaultConfig(),
SerfFloodInterval: 60 * time.Second,
ReconcileInterval: 60 * time.Second,
ProtocolVersion: ProtocolVersion2Compatible,
ACLRoleTTL: 30 * time.Second,
ACLPolicyTTL: 30 * time.Second,
ACLTokenTTL: 30 * time.Second,
ACLDefaultPolicy: "allow",
ACLDownPolicy: "extend-cache",
ACLReplicationRate: 1,
ACLReplicationBurst: 5,
ACLReplicationApplyLimit: 100, // ops / sec
ConfigReplicationRate: 1,
ConfigReplicationBurst: 5,
ConfigReplicationApplyLimit: 100, // ops / sec
TombstoneTTL: 15 * time.Minute,
TombstoneTTLGranularity: 30 * time.Second,
SessionTTLMin: 10 * time.Second,
ACLTokenMinExpirationTTL: 1 * time.Minute,
ACLTokenMaxExpirationTTL: 24 * time.Hour,
// These are tuned to provide a total throughput of 128 updates
// per second. If you update these, you should update the client-
// side SyncCoordinateRateTarget parameter accordingly.
CoordinateUpdatePeriod: 5 * time.Second,
CoordinateUpdateBatchSize: 128,
CoordinateUpdateMaxBatches: 5,
RPCRate: rate.Inf,
RPCMaxBurst: 1000,
2017-02-01 20:52:04 +00:00
TLSMinVersion: "tls10",
// TODO (slackpad) - Until #3744 is done, we need to keep these
2017-12-13 18:52:06 +00:00
// in sync with agent/config/default.go.
2017-12-12 00:38:52 +00:00
AutopilotConfig: &autopilot.Config{
CleanupDeadServers: true,
LastContactThreshold: 200 * time.Millisecond,
MaxTrailingLogs: 250,
ServerStabilizationTime: 10 * time.Second,
},
2017-12-13 18:52:06 +00:00
CAConfig: &structs.CAConfiguration{
Provider: "consul",
Config: map[string]interface{}{
"RotationPeriod": "2160h",
"LeafCertTTL": "72h",
},
},
ServerHealthInterval: 2 * time.Second,
AutopilotInterval: 10 * time.Second,
2013-12-06 23:43:07 +00:00
}
// Increase our reap interval to 3 days instead of 24h.
conf.SerfLANConfig.ReconnectTimeout = 3 * 24 * time.Hour
conf.SerfWANConfig.ReconnectTimeout = 3 * 24 * time.Hour
2013-12-07 01:18:09 +00:00
// WAN Serf should use the WAN timing, since we are using it
2013-12-06 23:43:07 +00:00
// to communicate between DC's
2013-12-07 01:18:09 +00:00
conf.SerfWANConfig.MemberlistConfig = memberlist.DefaultWANConfig()
2013-12-06 23:43:07 +00:00
// Ensure we don't have port conflicts
2013-12-27 20:51:15 +00:00
conf.SerfLANConfig.MemberlistConfig.BindPort = DefaultLANSerfPort
conf.SerfWANConfig.MemberlistConfig.BindPort = DefaultWANSerfPort
2013-12-06 23:43:07 +00:00
// Raft protocol version 3 only works with other Consul servers running
// 0.8.0 or later.
conf.RaftConfig.ProtocolVersion = 3
// Disable shutdown on removal
conf.RaftConfig.ShutdownOnRemove = false
// Check every 5 seconds to see if there are enough new entries for a snapshot, can be overridden
conf.RaftConfig.SnapshotInterval = 30 * time.Second
// Snapshots are created every 16384 entries by default, can be overridden
conf.RaftConfig.SnapshotThreshold = 16384
2013-12-06 23:43:07 +00:00
return conf
}