nomad

package
v0.10.0-rc1 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Oct 10, 2019 License: MPL-2.0 Imports: 70 Imported by: 0

Documentation

Index

Constants

View Source
const (
	// AutopilotRZTag is the Serf tag to use for the redundancy zone value
	// when passing the server metadata to Autopilot.
	AutopilotRZTag = "ap_zone"

	// AutopilotRZTag is the Serf tag to use for the custom version value
	// when passing the server metadata to Autopilot.
	AutopilotVersionTag = "ap_version"
)
View Source
const (
	DefaultRegion   = "global"
	DefaultDC       = "dc1"
	DefaultSerfPort = 4648
)
View Source
const (
	ProtocolVersionMin uint8 = 1
	ProtocolVersionMax       = 1
)

These are the protocol versions that Nomad can understand

View Source
const (
	// RegisterEnforceIndexErrPrefix is the prefix to use in errors caused by
	// enforcing the job modify index during registers.
	RegisterEnforceIndexErrPrefix = "Enforcing job modify index"

	// DispatchPayloadSizeLimit is the maximum size of the uncompressed input
	// data payload.
	DispatchPayloadSizeLimit = 16 * 1024
)
View Source
const (

	// NodeDrainEvents are the various drain messages
	NodeDrainEventDrainSet      = "Node drain strategy set"
	NodeDrainEventDrainDisabled = "Node drain disabled"
	NodeDrainEventDrainUpdated  = "Node drain stategy updated"

	// NodeEligibilityEventEligible is used when the nodes eligiblity is marked
	// eligible
	NodeEligibilityEventEligible = "Node marked as eligible for scheduling"

	// NodeEligibilityEventIneligible is used when the nodes eligiblity is marked
	// ineligible
	NodeEligibilityEventIneligible = "Node marked as ineligible for scheduling"

	// NodeHeartbeatEventReregistered is the message used when the node becomes
	// reregistered by the heartbeat.
	NodeHeartbeatEventReregistered = "Node reregistered by heartbeat"
)
View Source
const (
	// DefaultDequeueTimeout is used if no dequeue timeout is provided
	DefaultDequeueTimeout = time.Second
)
View Source
const (

	// NodeHeartbeatEventMissed is the event used when the Nodes heartbeat is
	// missed.
	NodeHeartbeatEventMissed = "Node heartbeat missed"
)
View Source
const (
	// StatusReap is used to update the status of a node if we
	// are handling a EventMemberReap
	StatusReap = serf.MemberStatus(-1)
)

Variables

View Source
var (
	// ErrNotOutstanding is returned if an evaluation is not outstanding
	ErrNotOutstanding = errors.New("evaluation is not outstanding")

	// ErrTokenMismatch is the outstanding eval has a different token
	ErrTokenMismatch = errors.New("evaluation token does not match")

	// ErrNackTimeoutReached is returned if an expired evaluation is reset
	ErrNackTimeoutReached = errors.New("evaluation nack timeout reached")
)
View Source
var (
	DefaultRPCAddr = &net.TCPAddr{IP: net.ParseIP("127.0.0.1"), Port: 4647}
)
View Source
var MinVersionPlanNormalization = version.Must(version.NewVersion("0.9.2"))

MinVersionPlanNormalization is the minimum version to support the normalization of Plan in SubmitPlan, and the denormalization raft log entry committed in ApplyPlanResultsRequest

Functions

func NewCoreScheduler

func NewCoreScheduler(srv *Server, snap *state.StateSnapshot) scheduler.Scheduler

NewCoreScheduler is used to return a new system scheduler instance

func NewFSM

func NewFSM(config *FSMConfig) (*nomadFSM, error)

NewFSMPath is used to construct a new FSM with a blank state

func NewPeriodicHeap added in v0.3.0

func NewPeriodicHeap() *periodicHeap

func NewVaultClient added in v0.5.0

func NewVaultClient(c *config.VaultConfig, logger log.Logger, purgeFn PurgeVaultAccessorFn) (*vaultClient, error)

NewVaultClient returns a Vault client from the given config. If the client couldn't be made an error is returned.

func NodeRpc added in v0.8.0

func NodeRpc(session *yamux.Session, method string, args, reply interface{}) error

NodeRpc is used to make an RPC call to a node. The method takes the Yamux session for the node and the method to be called.

func NodeStreamingRpc added in v0.8.0

func NodeStreamingRpc(session *yamux.Session, method string) (net.Conn, error)

NodeStreamingRpc is used to make a streaming RPC call to a node. The method takes the Yamux session for the node and the method to be called. It conducts the initial handshake and returns a connection to be used or an error. It is the callers responsibility to close the connection if there is no error.

func PoliciesFrom added in v0.5.0

func PoliciesFrom(s *vapi.Secret) ([]string, error)

PoliciesFrom parses the set of policies returned by a token lookup.

func ServersMeetMinimumVersion added in v0.7.0

func ServersMeetMinimumVersion(members []serf.Member, minVersion *version.Version, checkFailedServers bool) bool

ServersMeetMinimumVersion returns whether the Nomad servers are at least on the given Nomad version. The checkFailedServers parameter specifies whether version for the failed servers should be verified.

func TestJoin added in v0.8.0

func TestJoin(t testing.T, s1 *Server, other ...*Server)

Types

type ACL added in v0.7.0

type ACL struct {
	// contains filtered or unexported fields
}

ACL endpoint is used for manipulating ACL tokens and policies

func (*ACL) Bootstrap added in v0.7.0

Bootstrap is used to bootstrap the initial token

func (*ACL) DeletePolicies added in v0.7.0

func (a *ACL) DeletePolicies(args *structs.ACLPolicyDeleteRequest, reply *structs.GenericResponse) error

DeletePolicies is used to delete policies

func (*ACL) DeleteTokens added in v0.7.0

func (a *ACL) DeleteTokens(args *structs.ACLTokenDeleteRequest, reply *structs.GenericResponse) error

DeleteTokens is used to delete tokens

func (*ACL) GetPolicies added in v0.7.0

func (a *ACL) GetPolicies(args *structs.ACLPolicySetRequest, reply *structs.ACLPolicySetResponse) error

GetPolicies is used to get a set of policies

func (*ACL) GetPolicy added in v0.7.0

GetPolicy is used to get a specific policy

func (*ACL) GetToken added in v0.7.0

GetToken is used to get a specific token

func (*ACL) GetTokens added in v0.7.0

func (a *ACL) GetTokens(args *structs.ACLTokenSetRequest, reply *structs.ACLTokenSetResponse) error

GetTokens is used to get a set of token

func (*ACL) ListPolicies added in v0.7.0

func (a *ACL) ListPolicies(args *structs.ACLPolicyListRequest, reply *structs.ACLPolicyListResponse) error

ListPolicies is used to list the policies

func (*ACL) ListTokens added in v0.7.0

func (a *ACL) ListTokens(args *structs.ACLTokenListRequest, reply *structs.ACLTokenListResponse) error

ListTokens is used to list the tokens

func (*ACL) ResolveToken added in v0.7.0

func (a *ACL) ResolveToken(args *structs.ResolveACLTokenRequest, reply *structs.ResolveACLTokenResponse) error

ResolveToken is used to lookup a specific token by a secret ID. This is used for enforcing ACLs by clients.

func (*ACL) UpsertPolicies added in v0.7.0

func (a *ACL) UpsertPolicies(args *structs.ACLPolicyUpsertRequest, reply *structs.GenericResponse) error

UpsertPolicies is used to create or update a set of policies

func (*ACL) UpsertTokens added in v0.7.0

func (a *ACL) UpsertTokens(args *structs.ACLTokenUpsertRequest, reply *structs.ACLTokenUpsertResponse) error

UpsertTokens is used to create or update a set of tokens

type Alloc

type Alloc struct {
	// contains filtered or unexported fields
}

Alloc endpoint is used for manipulating allocations

func (*Alloc) GetAlloc

GetAlloc is used to lookup a particular allocation

func (*Alloc) GetAllocs added in v0.3.0

func (a *Alloc) GetAllocs(args *structs.AllocsGetRequest,
	reply *structs.AllocsGetResponse) error

GetAllocs is used to lookup a set of allocations

func (*Alloc) List

List is used to list the allocations in the system

func (*Alloc) Stop added in v0.9.2

Stop is used to stop an allocation and migrate it to another node.

func (*Alloc) UpdateDesiredTransition added in v0.8.0

func (a *Alloc) UpdateDesiredTransition(args *structs.AllocUpdateDesiredTransitionRequest, reply *structs.GenericResponse) error

UpdateDesiredTransition is used to update the desired transitions of an allocation.

type AllocGetter added in v0.9.6

type AllocGetter interface {
	AllocByID(ws memdb.WatchSet, id string) (*structs.Allocation, error)
}

AllocGetter is an interface for retrieving allocations by ID. It is satisfied by *state.StateStore and *state.StateSnapshot.

type AutopilotDelegate added in v0.8.0

type AutopilotDelegate struct {
	// contains filtered or unexported fields
}

AutopilotDelegate is a Nomad delegate for autopilot operations.

func (*AutopilotDelegate) AutopilotConfig added in v0.8.0

func (d *AutopilotDelegate) AutopilotConfig() *autopilot.Config

func (*AutopilotDelegate) FetchStats added in v0.8.0

func (d *AutopilotDelegate) FetchStats(ctx context.Context, servers []serf.Member) map[string]*autopilot.ServerStats

func (*AutopilotDelegate) IsServer added in v0.8.0

func (*AutopilotDelegate) NotifyHealth added in v0.8.0

func (d *AutopilotDelegate) NotifyHealth(health autopilot.OperatorHealthReply)

NotifyHealth heartbeats a metric for monitoring if we're the leader.

func (*AutopilotDelegate) PromoteNonVoters added in v0.8.0

func (d *AutopilotDelegate) PromoteNonVoters(conf *autopilot.Config, health autopilot.OperatorHealthReply) ([]raft.Server, error)

func (*AutopilotDelegate) Raft added in v0.8.0

func (d *AutopilotDelegate) Raft() *raft.Raft

func (*AutopilotDelegate) Serf added in v0.8.0

func (d *AutopilotDelegate) Serf() *serf.Serf

type BlockedEvals added in v0.3.0

type BlockedEvals struct {
	// contains filtered or unexported fields
}

BlockedEvals is used to track evaluations that shouldn't be queued until a certain class of nodes becomes available. An evaluation is put into the blocked state when it is run through the scheduler and produced failed allocations. It is unblocked when the capacity of a node that could run the failed allocation becomes available.

func NewBlockedEvals added in v0.3.0

func NewBlockedEvals(evalBroker *EvalBroker, logger log.Logger) *BlockedEvals

NewBlockedEvals creates a new blocked eval tracker that will enqueue unblocked evals into the passed broker.

func (*BlockedEvals) Block added in v0.3.0

func (b *BlockedEvals) Block(eval *structs.Evaluation)

Block tracks the passed evaluation and enqueues it into the eval broker when a suitable node calls unblock.

func (*BlockedEvals) EmitStats added in v0.3.0

func (b *BlockedEvals) EmitStats(period time.Duration, stopCh <-chan struct{})

EmitStats is used to export metrics about the blocked eval tracker while enabled

func (*BlockedEvals) Enabled added in v0.3.0

func (b *BlockedEvals) Enabled() bool

Enabled is used to check if the broker is enabled.

func (*BlockedEvals) Flush added in v0.3.0

func (b *BlockedEvals) Flush()

Flush is used to clear the state of blocked evaluations.

func (*BlockedEvals) GetDuplicates added in v0.3.0

func (b *BlockedEvals) GetDuplicates(timeout time.Duration) []*structs.Evaluation

GetDuplicates returns all the duplicate evaluations and blocks until the passed timeout.

func (*BlockedEvals) Reblock added in v0.4.0

func (b *BlockedEvals) Reblock(eval *structs.Evaluation, token string)

Reblock tracks the passed evaluation and enqueues it into the eval broker when a suitable node calls unblock. Reblock should be used over Block when the blocking is occurring by an outstanding evaluation. The token is the evaluation's token.

func (*BlockedEvals) SetEnabled added in v0.3.0

func (b *BlockedEvals) SetEnabled(enabled bool)

SetEnabled is used to control if the blocked eval tracker is enabled. The tracker should only be enabled on the active leader.

func (*BlockedEvals) SetTimetable added in v0.7.0

func (b *BlockedEvals) SetTimetable(timetable *TimeTable)

func (*BlockedEvals) Stats added in v0.3.0

func (b *BlockedEvals) Stats() *BlockedStats

Stats is used to query the state of the blocked eval tracker.

func (*BlockedEvals) Unblock added in v0.3.0

func (b *BlockedEvals) Unblock(computedClass string, index uint64)

Unblock causes any evaluation that could potentially make progress on a capacity change on the passed computed node class to be enqueued into the eval broker.

func (*BlockedEvals) UnblockClassAndQuota added in v0.7.0

func (b *BlockedEvals) UnblockClassAndQuota(class, quota string, index uint64)

UnblockClassAndQuota causes any evaluation that could potentially make progress on a capacity change on the passed computed node class or quota to be enqueued into the eval broker.

func (*BlockedEvals) UnblockFailed added in v0.4.0

func (b *BlockedEvals) UnblockFailed()

UnblockFailed unblocks all blocked evaluation that were due to scheduler failure.

func (*BlockedEvals) UnblockNode added in v0.9.4

func (b *BlockedEvals) UnblockNode(nodeID string, index uint64)

UnblockNode finds any blocked evalution that's node specific (system jobs) and enqueues it on the eval broker

func (*BlockedEvals) UnblockQuota added in v0.7.0

func (b *BlockedEvals) UnblockQuota(quota string, index uint64)

UnblockQuota causes any evaluation that could potentially make progress on a capacity change on the passed quota to be enqueued into the eval broker.

func (*BlockedEvals) Untrack added in v0.5.3

func (b *BlockedEvals) Untrack(jobID, namespace string)

Untrack causes any blocked evaluation for the passed job to be no longer tracked. Untrack is called when there is a successful evaluation for the job and a blocked evaluation is no longer needed.

type BlockedStats added in v0.3.0

type BlockedStats struct {
	// TotalEscaped is the total number of blocked evaluations that have escaped
	// computed node classes.
	TotalEscaped int

	// TotalBlocked is the total number of blocked evaluations.
	TotalBlocked int

	// TotalQuotaLimit is the total number of blocked evaluations that are due
	// to the quota limit being reached.
	TotalQuotaLimit int
}

BlockedStats returns all the stats about the blocked eval tracker.

type BrokerStats

type BrokerStats struct {
	TotalReady   int
	TotalUnacked int
	TotalBlocked int
	TotalWaiting int
	ByScheduler  map[string]*SchedulerStats
}

BrokerStats returns all the stats about the broker

type ClientAllocations added in v0.8.0

type ClientAllocations struct {
	// contains filtered or unexported fields
}

ClientAllocations is used to forward RPC requests to the targed Nomad client's Allocation endpoint.

func (*ClientAllocations) GarbageCollect added in v0.8.0

GarbageCollect is used to garbage collect an allocation on a client.

func (*ClientAllocations) GarbageCollectAll added in v0.8.0

func (a *ClientAllocations) GarbageCollectAll(args *structs.NodeSpecificRequest, reply *structs.GenericResponse) error

GarbageCollectAll is used to garbage collect all allocations on a client.

func (*ClientAllocations) Restart added in v0.9.2

Restart is used to trigger a restart of an allocation or a subtask on a client.

func (*ClientAllocations) Signal added in v0.9.2

Signal is used to send a signal to an allocation on a client.

func (*ClientAllocations) Stats added in v0.8.0

Stats is used to collect allocation statistics

type ClientStats added in v0.8.0

type ClientStats struct {
	// contains filtered or unexported fields
}

ClientStats is used to forward RPC requests to the targed Nomad client's ClientStats endpoint.

func (*ClientStats) Stats added in v0.8.0

type Config

type Config struct {
	// Bootstrap mode is used to bring up the first Nomad 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 Nomad servers. This can be used to automatically
	// bring up a collection of nodes.  All operations on BootstrapExpect
	// must be handled via `atomic.*Int32()` calls.
	BootstrapExpect int32

	// DataDir is the directory to store our state in
	DataDir string

	// DevMode is used for development purposes only and limits the
	// use of persistence or state.
	DevMode bool

	// DevDisableBootstrap is used to disable bootstrap mode while
	// in DevMode. This is largely used for testing.
	DevDisableBootstrap bool

	// LogOutput is the location to write logs to. If this is not set,
	// logs will go to stderr.
	LogOutput io.Writer

	// Logger is the logger used by the server.
	Logger log.Logger

	// ProtocolVersion is the protocol version to speak. This must be between
	// ProtocolVersionMin and ProtocolVersionMax.
	ProtocolVersion uint8

	// RPCAddr is the RPC address used by Nomad. This should be reachable
	// by the other servers and clients
	RPCAddr *net.TCPAddr

	// ClientRPCAdvertise is the address that is advertised to client nodes for
	// the RPC endpoint. This can differ from the RPC address, if for example
	// the RPCAddr is unspecified "0.0.0.0:4646", but this address must be
	// reachable
	ClientRPCAdvertise *net.TCPAddr

	// ServerRPCAdvertise is the address that is advertised to other servers for
	// the RPC endpoint. This can differ from the RPC address, if for example
	// the RPCAddr is unspecified "0.0.0.0:4646", but this address must be
	// reachable
	ServerRPCAdvertise *net.TCPAddr

	// RaftConfig is the configuration used for Raft in the local DC
	RaftConfig *raft.Config

	// RaftTimeout is applied to any network traffic for raft. Defaults to 10s.
	RaftTimeout time.Duration

	// (Enterprise-only) NonVoter is used to prevent this server from being added
	// as a voting member of the Raft cluster.
	NonVoter bool

	// (Enterprise-only) RedundancyZone is the redundancy zone to use for this server.
	RedundancyZone string

	// (Enterprise-only) UpgradeVersion is the custom upgrade version to use when
	// performing upgrade migrations.
	UpgradeVersion string

	// SerfConfig is the configuration for the serf cluster
	SerfConfig *serf.Config

	// Node name is the name we use to advertise. Defaults to hostname.
	NodeName string

	// NodeID is the uuid of this server.
	NodeID string

	// Region is the region this Nomad server belongs to.
	Region string

	// AuthoritativeRegion is the region which is treated as the authoritative source
	// for ACLs and Policies. This provides a single source of truth to resolve conflicts.
	AuthoritativeRegion string

	// Datacenter is the datacenter this Nomad server belongs to.
	Datacenter string

	// Build is a string that is gossiped around, and can be used to help
	// operators track which versions are actively deployed
	Build string

	// NumSchedulers is the number of scheduler thread that are run.
	// This can be as many as one per core, or zero to disable this server
	// from doing any scheduling work.
	NumSchedulers int

	// EnabledSchedulers controls the set of sub-schedulers that are
	// enabled for this server to handle. This will restrict the evaluations
	// that the workers dequeue for processing.
	EnabledSchedulers []string

	// 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

	// EvalGCInterval is how often we dispatch a job to GC evaluations
	EvalGCInterval time.Duration

	// EvalGCThreshold is how "old" an evaluation must be to be eligible
	// for GC. This gives users some time to debug a failed evaluation.
	EvalGCThreshold time.Duration

	// JobGCInterval is how often we dispatch a job to GC jobs that are
	// available for garbage collection.
	JobGCInterval time.Duration

	// JobGCThreshold is how old a job must be before it eligible for GC. This gives
	// the user time to inspect the job.
	JobGCThreshold time.Duration

	// NodeGCInterval is how often we dispatch a job to GC failed nodes.
	NodeGCInterval time.Duration

	// NodeGCThreshold is how "old" a node must be to be eligible
	// for GC. This gives users some time to view and debug a failed nodes.
	NodeGCThreshold time.Duration

	// DeploymentGCInterval is how often we dispatch a job to GC terminal
	// deployments.
	DeploymentGCInterval time.Duration

	// DeploymentGCThreshold is how "old" a deployment must be to be eligible
	// for GC. This gives users some time to view terminal deployments.
	DeploymentGCThreshold time.Duration

	// EvalNackTimeout controls how long we allow a sub-scheduler to
	// work on an evaluation before we consider it failed and Nack it.
	// This allows that evaluation to be handed to another sub-scheduler
	// to work on. Defaults to 60 seconds. This should be long enough that
	// no evaluation hits it unless the sub-scheduler has failed.
	EvalNackTimeout time.Duration

	// EvalDeliveryLimit is the limit of attempts we make to deliver and
	// process an evaluation. This is used so that an eval that will never
	// complete eventually fails out of the system.
	EvalDeliveryLimit int

	// EvalNackInitialReenqueueDelay is the delay applied before reenqueuing a
	// Nacked evaluation for the first time. This value should be small as the
	// initial Nack can be due to a down machine and the eval should be retried
	// quickly for liveliness.
	EvalNackInitialReenqueueDelay time.Duration

	// EvalNackSubsequentReenqueueDelay is the delay applied before reenqueuing
	// an evaluation that has been Nacked more than once. This delay is
	// compounding after the first Nack. This value should be significantly
	// longer than the initial delay as the purpose it severs is to apply
	// back-pressure as evaluations are being Nacked either due to scheduler
	// failures or because they are hitting their Nack timeout, both of which
	// are signs of high server resource usage.
	EvalNackSubsequentReenqueueDelay time.Duration

	// EvalFailedFollowupBaselineDelay is the minimum time waited before
	// retrying a failed evaluation.
	EvalFailedFollowupBaselineDelay time.Duration

	// EvalFailedFollowupDelayRange defines the range of additional time from
	// the baseline in which to wait before retrying a failed evaluation. The
	// additional delay is selected from this range randomly.
	EvalFailedFollowupDelayRange time.Duration

	// MinHeartbeatTTL is the minimum time between heartbeats.
	// This is used as a floor to prevent excessive updates.
	MinHeartbeatTTL time.Duration

	// MaxHeartbeatsPerSecond is the maximum target rate of heartbeats
	// being processed per second. This allows the TTL to be increased
	// to meet the target rate.
	MaxHeartbeatsPerSecond float64

	// HeartbeatGrace is the additional time given as a grace period
	// beyond the TTL to account for network and processing delays
	// as well as clock skew.
	HeartbeatGrace time.Duration

	// FailoverHeartbeatTTL is the TTL applied to heartbeats after
	// a new leader is elected, since we no longer know the status
	// of all the heartbeats.
	FailoverHeartbeatTTL time.Duration

	// ConsulConfig is this Agent's Consul configuration
	ConsulConfig *config.ConsulConfig

	// VaultConfig is this Agent's Vault configuration
	VaultConfig *config.VaultConfig

	// 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

	// TLSConfig holds various TLS related configurations
	TLSConfig *config.TLSConfig

	// ACLEnabled controls if ACL enforcement and management is enabled.
	ACLEnabled bool

	// ReplicationBackoff is how much we backoff when replication errors.
	// This is a tunable knob for testing primarily.
	ReplicationBackoff time.Duration

	// ReplicationToken is the ACL Token Secret ID used to fetch from
	// the Authoritative Region.
	ReplicationToken string

	// SentinelGCInterval is the interval that we GC unused policies.
	SentinelGCInterval time.Duration

	// SentinelConfig is this Agent's Sentinel configuration
	SentinelConfig *config.SentinelConfig

	// StatsCollectionInterval is the interval at which the Nomad server
	// publishes metrics which are periodic in nature like updating gauges
	StatsCollectionInterval time.Duration

	// DisableTaggedMetrics determines whether metrics will be displayed via a
	// key/value/tag format, or simply a key/value format
	DisableTaggedMetrics bool

	// DisableDispatchedJobSummaryMetrics allows for ignore dispatched jobs when
	// publishing Job summary metrics
	DisableDispatchedJobSummaryMetrics bool

	// BackwardsCompatibleMetrics determines whether to show methods of
	// displaying metrics for older versions, or to only show the new format
	BackwardsCompatibleMetrics bool

	// AutopilotConfig is used to apply the initial autopilot config when
	// bootstrapping.
	AutopilotConfig *structs.AutopilotConfig

	// 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

	// PluginLoader is used to load plugins.
	PluginLoader loader.PluginCatalog

	// PluginSingletonLoader is a plugin loader that will returns singleton
	// instances of the plugins.
	PluginSingletonLoader loader.PluginCatalog
}

Config is used to parameterize the server

func DefaultConfig

func DefaultConfig() *Config

DefaultConfig returns the default configuration

func (*Config) CheckVersion

func (c *Config) CheckVersion() error

CheckVersion is used to check if the ProtocolVersion is valid

type CoreScheduler

type CoreScheduler struct {
	// contains filtered or unexported fields
}

CoreScheduler is a special "scheduler" that is registered as "_core". It is used to run various administrative work across the cluster.

func (*CoreScheduler) Process

func (c *CoreScheduler) Process(eval *structs.Evaluation) error

Process is used to implement the scheduler.Scheduler interface

type Deployment added in v0.6.0

type Deployment struct {
	// contains filtered or unexported fields
}

Deployment endpoint is used for manipulating deployments

func (*Deployment) Allocations added in v0.6.0

Allocations returns the list of allocations that are a part of the deployment

func (*Deployment) Fail added in v0.6.0

Fail is used to force fail a deployment

func (*Deployment) GetDeployment added in v0.6.0

GetDeployment is used to request information about a specific deployment

func (*Deployment) List added in v0.6.0

List returns the list of deployments in the system

func (*Deployment) Pause added in v0.6.0

Pause is used to pause a deployment

func (*Deployment) Promote added in v0.6.0

Promote is used to promote canaries in a deployment

func (*Deployment) Reap added in v0.6.0

Reap is used to cleanup terminal deployments

func (*Deployment) SetAllocHealth added in v0.6.0

SetAllocHealth is used to set the health of allocations that are part of the deployment.

type EnterpriseEndpoints added in v0.7.0

type EnterpriseEndpoints struct{}

EnterpriseEndpoints holds the set of enterprise only endpoints to register

func NewEnterpriseEndpoints added in v0.7.0

func NewEnterpriseEndpoints(s *Server) *EnterpriseEndpoints

NewEnterpriseEndpoints returns a stub of the enterprise endpoints since there are none in oss

func (*EnterpriseEndpoints) Register added in v0.7.0

func (e *EnterpriseEndpoints) Register(s *rpc.Server)

Register is a no-op in oss.

type EnterpriseState added in v0.7.0

type EnterpriseState struct{}

type Eval

type Eval struct {
	// contains filtered or unexported fields
}

Eval endpoint is used for eval interactions

func (*Eval) Ack

func (e *Eval) Ack(args *structs.EvalAckRequest,
	reply *structs.GenericResponse) error

Ack is used to acknowledge completion of a dequeued evaluation

func (*Eval) Allocations

func (e *Eval) Allocations(args *structs.EvalSpecificRequest,
	reply *structs.EvalAllocationsResponse) error

Allocations is used to list the allocations for an evaluation

func (*Eval) Create

func (e *Eval) Create(args *structs.EvalUpdateRequest,
	reply *structs.GenericResponse) error

Create is used to make a new evaluation

func (*Eval) Dequeue

func (e *Eval) Dequeue(args *structs.EvalDequeueRequest,
	reply *structs.EvalDequeueResponse) error

Dequeue is used to dequeue a pending evaluation

func (*Eval) GetEval

func (e *Eval) GetEval(args *structs.EvalSpecificRequest,
	reply *structs.SingleEvalResponse) error

GetEval is used to request information about a specific evaluation

func (*Eval) List

func (e *Eval) List(args *structs.EvalListRequest,
	reply *structs.EvalListResponse) error

List is used to get a list of the evaluations in the system

func (*Eval) Nack

func (e *Eval) Nack(args *structs.EvalAckRequest,
	reply *structs.GenericResponse) error

NAck is used to negative acknowledge completion of a dequeued evaluation

func (*Eval) Reap

func (e *Eval) Reap(args *structs.EvalDeleteRequest,
	reply *structs.GenericResponse) error

Reap is used to cleanup dead evaluations and allocations

func (*Eval) Reblock added in v0.4.0

func (e *Eval) Reblock(args *structs.EvalUpdateRequest, reply *structs.GenericResponse) error

Reblock is used to reinsert an existing blocked evaluation into the blocked evaluation tracker.

func (*Eval) Update

func (e *Eval) Update(args *structs.EvalUpdateRequest,
	reply *structs.GenericResponse) error

Update is used to perform an update of an Eval if it is outstanding.

type EvalBroker

type EvalBroker struct {
	// contains filtered or unexported fields
}

EvalBroker is used to manage brokering of evaluations. When an evaluation is created, due to a change in a job specification or a node, we put it into the broker. The broker sorts by evaluations by priority and scheduler type. This allows us to dequeue the highest priority work first, while also allowing sub-schedulers to only dequeue work they know how to handle. The broker is designed to be entirely in-memory and is managed by the leader node.

The broker must provide at-least-once delivery semantics. It relies on explicit Ack/Nack messages to handle this. If a delivery is not Ack'd in a sufficient time span, it will be assumed Nack'd.

func NewEvalBroker

func NewEvalBroker(timeout, initialNackDelay, subsequentNackDelay time.Duration, deliveryLimit int) (*EvalBroker, error)

NewEvalBroker creates a new evaluation broker. This is parameterized with the timeout used for messages that are not acknowledged before we assume a Nack and attempt to redeliver as well as the deliveryLimit which prevents a failing eval from being endlessly delivered. The initialNackDelay is the delay before making a Nacked evaluation available again for the first Nack and subsequentNackDelay is the compounding delay after the first Nack.

func (*EvalBroker) Ack

func (b *EvalBroker) Ack(evalID, token string) error

Ack is used to positively acknowledge handling an evaluation

func (*EvalBroker) Dequeue

func (b *EvalBroker) Dequeue(schedulers []string, timeout time.Duration) (*structs.Evaluation, string, error)

Dequeue is used to perform a blocking dequeue

func (*EvalBroker) EmitStats

func (b *EvalBroker) EmitStats(period time.Duration, stopCh <-chan struct{})

EmitStats is used to export metrics about the broker while enabled

func (*EvalBroker) Enabled

func (b *EvalBroker) Enabled() bool

Enabled is used to check if the broker is enabled.

func (*EvalBroker) Enqueue

func (b *EvalBroker) Enqueue(eval *structs.Evaluation)

Enqueue is used to enqueue a new evaluation

func (*EvalBroker) EnqueueAll added in v0.3.0

func (b *EvalBroker) EnqueueAll(evals map[*structs.Evaluation]string)

EnqueueAll is used to enqueue many evaluations. The map allows evaluations that are being re-enqueued to include their token.

When requeuing an evaluation that potentially may be already enqueued. The evaluation is handled in one of the following ways: * Evaluation not outstanding: Process as a normal Enqueue * Evaluation outstanding: Do not allow the evaluation to be dequeued til:

  • Ack received: Unblock the evaluation allowing it to be dequeued
  • Nack received: Drop the evaluation as it was created as a result of a scheduler run that was Nack'd

func (*EvalBroker) Nack

func (b *EvalBroker) Nack(evalID, token string) error

Nack is used to negatively acknowledge handling an evaluation

func (*EvalBroker) Outstanding

func (b *EvalBroker) Outstanding(evalID string) (string, bool)

Outstanding checks if an EvalID has been delivered but not acknowledged and returns the associated token for the evaluation.

func (*EvalBroker) OutstandingReset added in v0.2.0

func (b *EvalBroker) OutstandingReset(evalID, token string) error

OutstandingReset resets the Nack timer for the EvalID if the token matches and the eval is outstanding

func (*EvalBroker) PauseNackTimeout added in v0.3.1

func (b *EvalBroker) PauseNackTimeout(evalID, token string) error

PauseNackTimeout is used to pause the Nack timeout for an eval that is making progress but is in a potentially unbounded operation such as the plan queue.

func (*EvalBroker) ResumeNackTimeout added in v0.3.1

func (b *EvalBroker) ResumeNackTimeout(evalID, token string) error

ResumeNackTimeout is used to resume the Nack timeout for an eval that was paused. It should be resumed after leaving an unbounded operation.

func (*EvalBroker) SetEnabled

func (b *EvalBroker) SetEnabled(enabled bool)

SetEnabled is used to control if the broker is enabled. The broker should only be enabled on the active leader.

func (*EvalBroker) Stats

func (b *EvalBroker) Stats() *BrokerStats

Stats is used to query the state of the broker

type EvaluatePool added in v0.3.0

type EvaluatePool struct {
	// contains filtered or unexported fields
}

EvaluatePool is used to have a pool of workers that are evaluating if a plan is valid. It can be used to parallelize the evaluation of a plan.

func NewEvaluatePool added in v0.3.0

func NewEvaluatePool(workers, bufSize int) *EvaluatePool

NewEvaluatePool returns a pool of the given size.

func (*EvaluatePool) RequestCh added in v0.3.0

func (p *EvaluatePool) RequestCh() chan<- evaluateRequest

RequestCh is used to push requests

func (*EvaluatePool) ResultCh added in v0.3.0

func (p *EvaluatePool) ResultCh() <-chan evaluateResult

ResultCh is used to read the results as they are ready

func (*EvaluatePool) SetSize added in v0.3.0

func (p *EvaluatePool) SetSize(size int)

SetSize is used to resize the worker pool

func (*EvaluatePool) Shutdown added in v0.3.0

func (p *EvaluatePool) Shutdown()

Shutdown is used to shutdown the pool

func (*EvaluatePool) Size added in v0.3.0

func (p *EvaluatePool) Size() int

Size returns the current size

type FSMConfig added in v0.7.0

type FSMConfig struct {
	// EvalBroker is the evaluation broker evaluations should be added to
	EvalBroker *EvalBroker

	// Periodic is the periodic job dispatcher that periodic jobs should be
	// added/removed from
	Periodic *PeriodicDispatch

	// BlockedEvals is the blocked eval tracker that blocked evaluations should
	// be added to.
	Blocked *BlockedEvals

	// Logger is the logger used by the FSM
	Logger log.Logger

	// Region is the region of the server embedding the FSM
	Region string
}

FSMConfig is used to configure the FSM

type FileSystem added in v0.8.0

type FileSystem struct {
	// contains filtered or unexported fields
}

FileSystem endpoint is used for accessing the logs and filesystem of allocations from a Node.

func (*FileSystem) List added in v0.8.0

List is used to list the contents of an allocation's directory.

func (*FileSystem) Stat added in v0.8.0

Stat is used to stat a file in the allocation's directory.

type Job

type Job struct {
	// contains filtered or unexported fields
}

Job endpoint is used for job interactions

func NewJobEndpoints added in v0.10.0

func NewJobEndpoints(s *Server) *Job

NewJobEndpoints creates a new job endpoint with builtin admission controllers

func (*Job) Allocations

func (j *Job) Allocations(args *structs.JobSpecificRequest,
	reply *structs.JobAllocationsResponse) error

Allocations is used to list the allocations for a job

func (*Job) BatchDeregister added in v0.8.0

BatchDeregister is used to remove a set of jobs from the cluster.

func (*Job) Deployments added in v0.6.0

func (j *Job) Deployments(args *structs.JobSpecificRequest,
	reply *structs.DeploymentListResponse) error

Deployments is used to list the deployments for a job

func (*Job) Deregister

func (j *Job) Deregister(args *structs.JobDeregisterRequest, reply *structs.JobDeregisterResponse) error

Deregister is used to remove a job the cluster.

func (*Job) Dispatch added in v0.5.3

func (j *Job) Dispatch(args *structs.JobDispatchRequest, reply *structs.JobDispatchResponse) error

Dispatch a parameterized job.

func (*Job) Evaluate

func (j *Job) Evaluate(args *structs.JobEvaluateRequest, reply *structs.JobRegisterResponse) error

Evaluate is used to force a job for re-evaluation

func (*Job) Evaluations

func (j *Job) Evaluations(args *structs.JobSpecificRequest,
	reply *structs.JobEvaluationsResponse) error

Evaluations is used to list the evaluations for a job

func (*Job) GetJob

func (j *Job) GetJob(args *structs.JobSpecificRequest,
	reply *structs.SingleJobResponse) error

GetJob is used to request information about a specific job

func (*Job) GetJobVersions added in v0.6.0

func (j *Job) GetJobVersions(args *structs.JobVersionsRequest,
	reply *structs.JobVersionsResponse) error

GetJobVersions is used to retrieve all tracked versions of a job.

func (*Job) LatestDeployment added in v0.6.0

func (j *Job) LatestDeployment(args *structs.JobSpecificRequest,
	reply *structs.SingleDeploymentResponse) error

LatestDeployment is used to retrieve the latest deployment for a job

func (*Job) List

func (j *Job) List(args *structs.JobListRequest,
	reply *structs.JobListResponse) error

List is used to list the jobs registered in the system

func (*Job) Plan added in v0.4.0

func (j *Job) Plan(args *structs.JobPlanRequest, reply *structs.JobPlanResponse) error

Plan is used to cause a dry-run evaluation of the Job and return the results with a potential diff containing annotations.

func (*Job) Register

func (j *Job) Register(args *structs.JobRegisterRequest, reply *structs.JobRegisterResponse) error

Register is used to upsert a job for scheduling

func (*Job) Revert added in v0.6.0

func (j *Job) Revert(args *structs.JobRevertRequest, reply *structs.JobRegisterResponse) error

Revert is used to revert the job to a prior version

func (*Job) Stable added in v0.6.0

Stable is used to mark the job version as stable

func (*Job) Summary added in v0.4.1

func (j *Job) Summary(args *structs.JobSummaryRequest,
	reply *structs.JobSummaryResponse) error

Summary retrieves the summary of a job

func (*Job) Validate added in v0.5.5

func (j *Job) Validate(args *structs.JobValidateRequest, reply *structs.JobValidateResponse) error

Validate validates a job

type JobEvalDispatcher added in v0.3.0

type JobEvalDispatcher interface {
	// DispatchJob takes a job a new, untracked job and creates an evaluation
	// for it and returns the eval.
	DispatchJob(job *structs.Job) (*structs.Evaluation, error)

	// RunningChildren returns whether the passed job has any running children.
	RunningChildren(job *structs.Job) (bool, error)
}

JobEvalDispatcher is an interface to submit jobs and have evaluations created for them.

type LogApplier added in v0.7.0

type LogApplier func(buf []byte, index uint64) interface{}

LogApplier is the definition of a function that can apply a Raft log

type LogAppliers added in v0.7.0

type LogAppliers map[structs.MessageType]LogApplier

LogAppliers is a mapping of the Raft MessageType to the appropriate log applier

type Node

type Node struct {
	// contains filtered or unexported fields
}

Node endpoint is used for client interactions

func (*Node) BatchDeregister added in v0.9.4

func (n *Node) BatchDeregister(args *structs.NodeBatchDeregisterRequest, reply *structs.NodeUpdateResponse) error

BatchDeregister is used to remove client nodes from the cluster.

func (*Node) Deregister

func (n *Node) Deregister(args *structs.NodeDeregisterRequest, reply *structs.NodeUpdateResponse) error

Deregister is used to remove a client from the cluster. If a client should just be made unavailable for scheduling, a status update is preferred.

func (*Node) DeriveVaultToken added in v0.5.0

func (n *Node) DeriveVaultToken(args *structs.DeriveVaultTokenRequest,
	reply *structs.DeriveVaultTokenResponse) error

DeriveVaultToken is used by the clients to request wrapped Vault tokens for tasks

func (*Node) EmitEvents added in v0.8.0

func (*Node) Evaluate

func (n *Node) Evaluate(args *structs.NodeEvaluateRequest, reply *structs.NodeUpdateResponse) error

Evaluate is used to force a re-evaluation of the node

func (*Node) GetAllocs

func (n *Node) GetAllocs(args *structs.NodeSpecificRequest,
	reply *structs.NodeAllocsResponse) error

GetAllocs is used to request allocations for a specific node

func (*Node) GetClientAllocs added in v0.3.0

func (n *Node) GetClientAllocs(args *structs.NodeSpecificRequest,
	reply *structs.NodeClientAllocsResponse) error

GetClientAllocs is used to request a lightweight list of alloc modify indexes per allocation.

func (*Node) GetNode

func (n *Node) GetNode(args *structs.NodeSpecificRequest,
	reply *structs.SingleNodeResponse) error

GetNode is used to request information about a specific node

func (*Node) List

func (n *Node) List(args *structs.NodeListRequest,
	reply *structs.NodeListResponse) error

List is used to list the available nodes

func (*Node) Register

func (n *Node) Register(args *structs.NodeRegisterRequest, reply *structs.NodeUpdateResponse) error

Register is used to upsert a client that is available for scheduling

func (*Node) UpdateAlloc

func (n *Node) UpdateAlloc(args *structs.AllocUpdateRequest, reply *structs.GenericResponse) error

UpdateAlloc is used to update the client status of an allocation

func (*Node) UpdateDrain

UpdateDrain is used to update the drain mode of a client node

func (*Node) UpdateEligibility added in v0.8.0

UpdateEligibility is used to update the scheduling eligibility of a node

func (*Node) UpdateStatus

func (n *Node) UpdateStatus(args *structs.NodeUpdateStatusRequest, reply *structs.NodeUpdateResponse) error

UpdateStatus is used to update the status of a client node

type Operator added in v0.5.5

type Operator struct {
	// contains filtered or unexported fields
}

Operator endpoint is used to perform low-level operator tasks for Nomad.

func (*Operator) AutopilotGetConfiguration added in v0.8.0

func (op *Operator) AutopilotGetConfiguration(args *structs.GenericRequest, reply *structs.AutopilotConfig) error

AutopilotGetConfiguration is used to retrieve the current Autopilot configuration.

func (*Operator) AutopilotSetConfiguration added in v0.8.0

func (op *Operator) AutopilotSetConfiguration(args *structs.AutopilotSetConfigRequest, reply *bool) error

AutopilotSetConfiguration is used to set the current Autopilot configuration.

func (*Operator) RaftGetConfiguration added in v0.5.5

func (op *Operator) RaftGetConfiguration(args *structs.GenericRequest, reply *structs.RaftConfigurationResponse) error

RaftGetConfiguration is used to retrieve the current Raft configuration.

func (*Operator) RaftRemovePeerByAddress added in v0.5.5

func (op *Operator) RaftRemovePeerByAddress(args *structs.RaftPeerByAddressRequest, reply *struct{}) error

RaftRemovePeerByAddress is used to kick a stale peer (one that it in the Raft quorum but no longer known to Serf or the catalog) by address in the form of "IP:port". The reply argument is not used, but it required to fulfill the RPC interface.

func (*Operator) RaftRemovePeerByID added in v0.8.0

func (op *Operator) RaftRemovePeerByID(args *structs.RaftPeerByIDRequest, reply *struct{}) error

RaftRemovePeerByID is used to kick a stale peer (one that is in the Raft quorum but no longer known to Serf or the catalog) by address in the form of "IP:port". The reply argument is not used, but is required to fulfill the RPC interface.

func (*Operator) SchedulerGetConfiguration added in v0.9.0

func (op *Operator) SchedulerGetConfiguration(args *structs.GenericRequest, reply *structs.SchedulerConfigurationResponse) error

SchedulerGetConfiguration is used to retrieve the current Scheduler configuration.

func (*Operator) SchedulerSetConfiguration added in v0.9.0

func (op *Operator) SchedulerSetConfiguration(args *structs.SchedulerSetConfigRequest, reply *structs.SchedulerSetConfigurationResponse) error

SchedulerSetConfiguration is used to set the current Scheduler configuration.

func (*Operator) ServerHealth added in v0.8.0

func (op *Operator) ServerHealth(args *structs.GenericRequest, reply *autopilot.OperatorHealthReply) error

ServerHealth is used to get the current health of the servers.

type PendingEvaluations

type PendingEvaluations []*structs.Evaluation

PendingEvaluations is a list of waiting evaluations. We implement the container/heap interface so that this is a priority queue

func (PendingEvaluations) Len

func (p PendingEvaluations) Len() int

Len is for the sorting interface

func (PendingEvaluations) Less

func (p PendingEvaluations) Less(i, j int) bool

Less is for the sorting interface. We flip the check so that the "min" in the min-heap is the element with the highest priority

func (PendingEvaluations) Peek

Peek is used to peek at the next element that would be popped

func (*PendingEvaluations) Pop

func (p *PendingEvaluations) Pop() interface{}

Pop is used to remove an evaluation from the slice

func (*PendingEvaluations) Push

func (p *PendingEvaluations) Push(e interface{})

Push is used to add a new evaluation to the slice

func (PendingEvaluations) Swap

func (p PendingEvaluations) Swap(i, j int)

Swap is for the sorting interface

type PendingPlans

type PendingPlans []*pendingPlan

PendingPlans is a list of waiting plans. We implement the container/heap interface so that this is a priority queue

func (PendingPlans) Len

func (p PendingPlans) Len() int

Len is for the sorting interface

func (PendingPlans) Less

func (p PendingPlans) Less(i, j int) bool

Less is for the sorting interface. We flip the check so that the "min" in the min-heap is the element with the highest priority. For the same priority, we use the enqueue time of the evaluation to give a FIFO ordering.

func (PendingPlans) Peek

func (p PendingPlans) Peek() *pendingPlan

Peek is used to peek at the next element that would be popped

func (*PendingPlans) Pop

func (p *PendingPlans) Pop() interface{}

Pop is used to remove an evaluation from the slice

func (*PendingPlans) Push

func (p *PendingPlans) Push(e interface{})

Push is used to add a new evaluation to the slice

func (PendingPlans) Swap

func (p PendingPlans) Swap(i, j int)

Swap is for the sorting interface

type Periodic added in v0.3.0

type Periodic struct {
	// contains filtered or unexported fields
}

Periodic endpoint is used for periodic job interactions

func (*Periodic) Force added in v0.3.0

Force is used to force a new instance of a periodic job

type PeriodicDispatch added in v0.3.0

type PeriodicDispatch struct {
	// contains filtered or unexported fields
}

PeriodicDispatch is used to track and launch periodic jobs. It maintains the set of periodic jobs and creates derived jobs and evaluations per instantiation which is determined by the periodic spec.

func NewPeriodicDispatch added in v0.3.0

func NewPeriodicDispatch(logger log.Logger, dispatcher JobEvalDispatcher) *PeriodicDispatch

NewPeriodicDispatch returns a periodic dispatcher that is used to track and launch periodic jobs.

func (*PeriodicDispatch) Add added in v0.3.0

func (p *PeriodicDispatch) Add(job *structs.Job) error

Add begins tracking of a periodic job. If it is already tracked, it acts as an update to the jobs periodic spec. The method returns whether the job was added and any error that may have occurred.

func (*PeriodicDispatch) ForceRun added in v0.3.0

func (p *PeriodicDispatch) ForceRun(namespace, jobID string) (*structs.Evaluation, error)

ForceRun causes the periodic job to be evaluated immediately and returns the subsequent eval.

func (*PeriodicDispatch) LaunchTime added in v0.3.0

func (p *PeriodicDispatch) LaunchTime(jobID string) (time.Time, error)

LaunchTime returns the launch time of the job. This is only valid for jobs created by PeriodicDispatch and will otherwise return an error.

func (*PeriodicDispatch) Remove added in v0.3.0

func (p *PeriodicDispatch) Remove(namespace, jobID string) error

Remove stops tracking the passed job. If the job is not tracked, it is a no-op.

func (*PeriodicDispatch) SetEnabled added in v0.3.0

func (p *PeriodicDispatch) SetEnabled(enabled bool)

SetEnabled is used to control if the periodic dispatcher is enabled. It should only be enabled on the active leader. Disabling an active dispatcher will stop any launched go routine and flush the dispatcher.

func (*PeriodicDispatch) Tracked added in v0.3.0

func (p *PeriodicDispatch) Tracked() []*structs.Job

Tracked returns the set of tracked job IDs.

type Plan

type Plan struct {
	// contains filtered or unexported fields
}

Plan endpoint is used for plan interactions

func (*Plan) Submit

func (p *Plan) Submit(args *structs.PlanRequest, reply *structs.PlanResponse) error

Submit is used to submit a plan to the leader

type PlanFuture

type PlanFuture interface {
	Wait() (*structs.PlanResult, error)
}

PlanFuture is used to return a future for an enqueue

type PlanQueue

type PlanQueue struct {
	// contains filtered or unexported fields
}

PlanQueue is used to submit commit plans for task allocations to the current leader. The leader verifies that resources are not over-committed and commits to Raft. This allows sub-schedulers to be optimistically concurrent. In the case of an overcommit, the plan may be partially applied if allowed, or completely rejected (gang commit).

func NewPlanQueue

func NewPlanQueue() (*PlanQueue, error)

NewPlanQueue is used to construct and return a new plan queue

func (*PlanQueue) Dequeue

func (q *PlanQueue) Dequeue(timeout time.Duration) (*pendingPlan, error)

Dequeue is used to perform a blocking dequeue

func (*PlanQueue) EmitStats

func (q *PlanQueue) EmitStats(period time.Duration, stopCh <-chan struct{})

EmitStats is used to export metrics about the broker while enabled

func (*PlanQueue) Enabled

func (q *PlanQueue) Enabled() bool

Enabled is used to check if the queue is enabled.

func (*PlanQueue) Enqueue

func (q *PlanQueue) Enqueue(plan *structs.Plan) (PlanFuture, error)

Enqueue is used to enqueue a plan

func (*PlanQueue) Flush

func (q *PlanQueue) Flush()

Flush is used to reset the state of the plan queue

func (*PlanQueue) SetEnabled

func (q *PlanQueue) SetEnabled(enabled bool)

SetEnabled is used to control if the queue is enabled. The queue should only be enabled on the active leader.

func (*PlanQueue) Stats

func (q *PlanQueue) Stats() *QueueStats

Stats is used to query the state of the queue

type PurgeVaultAccessorFn added in v0.5.0

type PurgeVaultAccessorFn func(accessors []*structs.VaultAccessor) error

PurgeVaultAccessor is called to remove VaultAccessors from the system. If the function returns an error, the token will still be tracked and revocation will retry till there is a success

type QueueStats

type QueueStats struct {
	Depth int
}

QueueStats returns all the stats about the plan queue

type RPCContext added in v0.8.0

type RPCContext struct {
	// Conn exposes the raw connection.
	Conn net.Conn

	// Session exposes the multiplexed connection session.
	Session *yamux.Session

	// TLS marks whether the RPC is over a TLS based connection
	TLS bool

	// VerifiedChains is is the Verified certificates presented by the incoming
	// connection.
	VerifiedChains [][]*x509.Certificate

	// NodeID marks the NodeID that initiated the connection.
	NodeID string
}

RPCContext provides metadata about the RPC connection.

type RaftLayer

type RaftLayer struct {
	// contains filtered or unexported fields
}

RaftLayer implements the raft.StreamLayer interface, so that we can use a single RPC layer for Raft and Nomad

func NewRaftLayer

func NewRaftLayer(addr net.Addr, tlsWrap tlsutil.Wrapper) *RaftLayer

NewRaftLayer is used to initialize a new RaftLayer which can be used as a StreamLayer for Raft. If a tlsConfig is provided, then the connection will use TLS.

func (*RaftLayer) Accept

func (l *RaftLayer) Accept() (net.Conn, error)

Accept is used to return connection which are dialed to be used with the Raft layer

func (*RaftLayer) Addr

func (l *RaftLayer) Addr() net.Addr

Addr is used to return the address of the listener

func (*RaftLayer) Close

func (l *RaftLayer) Close() error

Close is used to stop listening for Raft connections

func (*RaftLayer) Dial

func (l *RaftLayer) Dial(address raft.ServerAddress, timeout time.Duration) (net.Conn, error)

Dial is used to create a new outgoing connection

func (*RaftLayer) Handoff

func (l *RaftLayer) Handoff(ctx context.Context, c net.Conn) error

Handoff is used to hand off a connection to the RaftLayer. This allows it to be Accept()'ed

func (*RaftLayer) ReloadTLS added in v0.8.0

func (l *RaftLayer) ReloadTLS(tlsWrap tlsutil.Wrapper)

ReloadTLS swaps the TLS wrapper. This is useful when upgrading or downgrading TLS connections.

type Region added in v0.2.1

type Region struct {
	// contains filtered or unexported fields
}

Region is used to query and list the known regions

func (*Region) List added in v0.2.1

func (r *Region) List(args *structs.GenericRequest, reply *[]string) error

List is used to list all of the known regions. No leader forwarding is required for this endpoint because memberlist is used to populate the peers list we read from.

type SchedulerStats

type SchedulerStats struct {
	Ready   int
	Unacked int
}

SchedulerStats returns the stats per scheduler

type Search struct {
	// contains filtered or unexported fields
}

Search endpoint is used to look up matches for a given prefix and context

func (*Search) PrefixSearch added in v0.6.1

func (s *Search) PrefixSearch(args *structs.SearchRequest, reply *structs.SearchResponse) error

PrefixSearch is used to list matches for a given prefix, and returns matching jobs, evaluations, allocations, and/or nodes.

type Server

type Server struct {

	// EnterpriseState is used to fill in state for Pro/Ent builds
	EnterpriseState
	// contains filtered or unexported fields
}

Server is Nomad server which manages the job queues, schedulers, and notification bus for agents.

func NewServer

func NewServer(config *Config, consulCatalog consul.CatalogAPI) (*Server, error)

NewServer is used to construct a new Nomad server from the configuration, potentially returning an error

func TestACLServer added in v0.8.0

func TestACLServer(t testing.T, cb func(*Config)) (*Server, *structs.ACLToken)

func TestServer added in v0.8.0

func TestServer(t testing.T, cb func(*Config)) *Server

func (*Server) Datacenter added in v0.5.0

func (s *Server) Datacenter() string

Datacenter returns the data center of the server

func (*Server) DispatchJob added in v0.3.0

func (s *Server) DispatchJob(job *structs.Job) (*structs.Evaluation, error)

DispatchJob creates an evaluation for the passed job and commits both the evaluation and the job to the raft log. It returns the eval.

func (*Server) EmitRaftStats added in v0.9.4

func (s *Server) EmitRaftStats(period time.Duration, stopCh <-chan struct{})

EmitRaftStats is used to export metrics about raft indexes and state store snapshot index

func (*Server) Encrypted

func (s *Server) Encrypted() bool

Encrypted determines if gossip is encrypted

func (*Server) GetConfig added in v0.5.0

func (s *Server) GetConfig() *Config

GetConfig returns the config of the server for testing purposes only

func (*Server) IsLeader

func (s *Server) IsLeader() bool

IsLeader checks if this server is the cluster leader

func (*Server) IsShutdown

func (s *Server) IsShutdown() bool

IsShutdown checks if the server is shutdown

func (*Server) Join

func (s *Server) Join(addrs []string) (int, error)

Join is used to have Nomad join the gossip ring The target address should be another node listening on the Serf address

func (*Server) KeyManager

func (s *Server) KeyManager() *serf.KeyManager

KeyManager returns the Serf keyring manager

func (*Server) Leave

func (s *Server) Leave() error

Leave is used to prepare for a graceful shutdown of the server

func (*Server) LocalMember

func (s *Server) LocalMember() serf.Member

LocalMember is used to return the local node

func (*Server) Members

func (s *Server) Members() []serf.Member

Members is used to return the members of the serf cluster

func (*Server) RPC

func (s *Server) RPC(method string, args interface{}, reply interface{}) error

RPC is used to make a local RPC call

func (*Server) Region added in v0.5.0

func (s *Server) Region() string

Region returns the region of the server

func (*Server) Regions added in v0.2.1

func (s *Server) Regions() []string

Regions returns the known regions in the cluster.

func (*Server) Reload added in v0.5.5

func (s *Server) Reload(newConfig *Config) error

Reload handles a config reload specific to server-only configuration. Not all config fields can handle a reload.

func (*Server) RemoveFailedNode

func (s *Server) RemoveFailedNode(node string) error

RemoveFailedNode is used to remove a failed node from the cluster

func (*Server) ReplicationToken added in v0.7.0

func (s *Server) ReplicationToken() string

ReplicationToken returns the token used for replication. We use a method to support dynamic reloading of this value later.

func (*Server) ResolveToken added in v0.7.0

func (s *Server) ResolveToken(secretID string) (*acl.ACL, error)

ResolveToken is used to translate an ACL Token Secret ID into an ACL object, nil if ACLs are disabled, or an error.

func (*Server) RunningChildren added in v0.3.0

func (s *Server) RunningChildren(job *structs.Job) (bool, error)

RunningChildren checks whether the passed job has any running children.

func (*Server) Shutdown

func (s *Server) Shutdown() error

Shutdown is used to shutdown the server

func (*Server) State

func (s *Server) State() *state.StateStore

State returns the underlying state store. This should *not* be used to modify state directly.

func (*Server) Stats

func (s *Server) Stats() map[string]map[string]string

Stats is used to return statistics for debugging and insight for various sub-systems

func (*Server) StreamingRpcHandler added in v0.8.0

func (s *Server) StreamingRpcHandler(method string) (structs.StreamingRpcHandler, error)

StreamingRpcHandler is used to make a streaming RPC call.

type SnapshotRestorer added in v0.7.0

type SnapshotRestorer func(restore *state.StateRestore, dec *codec.Decoder) error

SnapshotRestorer is the definition of a function that can apply a Raft log

type SnapshotRestorers added in v0.7.0

type SnapshotRestorers map[SnapshotType]SnapshotRestorer

SnapshotRestorers is a mapping of the SnapshotType to the appropriate snapshot restorer.

type SnapshotType

type SnapshotType byte

SnapshotType is prefixed to a record in the FSM snapshot so that we can determine the type for restore

const (
	NodeSnapshot SnapshotType = iota
	JobSnapshot
	IndexSnapshot
	EvalSnapshot
	AllocSnapshot
	TimeTableSnapshot
	PeriodicLaunchSnapshot
	JobSummarySnapshot
	VaultAccessorSnapshot
	JobVersionSnapshot
	DeploymentSnapshot
	ACLPolicySnapshot
	ACLTokenSnapshot
	SchedulerConfigSnapshot
)

type StatsFetcher added in v0.8.0

type StatsFetcher struct {
	// contains filtered or unexported fields
}

StatsFetcher has two functions for autopilot. First, lets us fetch all the stats in parallel so we are taking a sample as close to the same time as possible, since we are comparing time-sensitive info for the health check. Second, it bounds the time so that one slow RPC can't hold up the health check loop; as a side effect of how it implements this, it also limits to a single in-flight RPC to any given server, so goroutines don't accumulate as we run the health check fairly frequently.

func NewStatsFetcher added in v0.8.0

func NewStatsFetcher(logger log.Logger, pool *pool.ConnPool, region string) *StatsFetcher

NewStatsFetcher returns a stats fetcher.

func (*StatsFetcher) Fetch added in v0.8.0

func (f *StatsFetcher) Fetch(ctx context.Context, members []serf.Member) map[string]*autopilot.ServerStats

Fetch will attempt to query all the servers in parallel.

type Status

type Status struct {
	// contains filtered or unexported fields
}

Status endpoint is used to check on server status

func (*Status) HasNodeConn added in v0.8.0

func (s *Status) HasNodeConn(args *structs.NodeSpecificRequest, reply *structs.NodeConnQueryResponse) error

HasNodeConn returns whether the server has a connection to the requested Node.

func (*Status) Leader

func (s *Status) Leader(args *structs.GenericRequest, reply *string) error

Leader is used to get the address of the leader

func (*Status) Members added in v0.5.0

func (s *Status) Members(args *structs.GenericRequest, reply *structs.ServerMembersResponse) error

Members return the list of servers in a cluster that a particular server is aware of

func (*Status) Peers

func (s *Status) Peers(args *structs.GenericRequest, reply *[]string) error

Peers is used to get all the Raft peers

func (*Status) Ping

func (s *Status) Ping(args struct{}, reply *struct{}) error

Ping is used to just check for connectivity

func (*Status) RaftStats added in v0.8.0

func (s *Status) RaftStats(args struct{}, reply *autopilot.ServerStats) error

Used by Autopilot to query the raft stats of the local server.

func (*Status) Version

func (s *Status) Version(args *structs.GenericRequest, reply *structs.VersionResponse) error

Version is used to allow clients to determine the capabilities of the server

type System added in v0.3.0

type System struct {
	// contains filtered or unexported fields
}

System endpoint is used to call invoke system tasks.

func (*System) GarbageCollect added in v0.3.0

func (s *System) GarbageCollect(args *structs.GenericRequest, reply *structs.GenericResponse) error

GarbageCollect is used to trigger the system to immediately garbage collect nodes, evals and jobs.

func (*System) ReconcileJobSummaries added in v0.4.1

func (s *System) ReconcileJobSummaries(args *structs.GenericRequest, reply *structs.GenericResponse) error

ReconcileSummaries reconciles the summaries of all the jobs in the state store

type TestVaultClient added in v0.5.0

type TestVaultClient struct {
	// LookupTokenErrors maps a token to an error that will be returned by the
	// LookupToken call
	LookupTokenErrors map[string]error

	// LookupTokenSecret maps a token to the Vault secret that will be returned
	// by the LookupToken call
	LookupTokenSecret map[string]*vapi.Secret

	// CreateTokenErrors maps a token to an error that will be returned by the
	// CreateToken call
	CreateTokenErrors map[string]map[string]error

	// CreateTokenSecret maps a token to the Vault secret that will be returned
	// by the CreateToken call
	CreateTokenSecret map[string]map[string]*vapi.Secret

	RevokedTokens []*structs.VaultAccessor
}

TestVaultClient is a Vault client appropriate for use during testing. Its behavior is programmable such that endpoints can be tested under various circumstances.

func (*TestVaultClient) CreateToken added in v0.5.0

func (v *TestVaultClient) CreateToken(ctx context.Context, a *structs.Allocation, task string) (*vapi.Secret, error)

func (*TestVaultClient) EmitStats added in v0.5.5

func (v *TestVaultClient) EmitStats(period time.Duration, stopCh <-chan struct{})

func (*TestVaultClient) LookupToken added in v0.5.0

func (v *TestVaultClient) LookupToken(ctx context.Context, token string) (*vapi.Secret, error)

func (*TestVaultClient) RevokeTokens added in v0.5.0

func (v *TestVaultClient) RevokeTokens(ctx context.Context, accessors []*structs.VaultAccessor, committed bool) error

func (*TestVaultClient) Running added in v0.5.5

func (v *TestVaultClient) Running() bool

func (*TestVaultClient) SetActive added in v0.5.0

func (v *TestVaultClient) SetActive(enabled bool)

func (*TestVaultClient) SetConfig added in v0.5.0

func (v *TestVaultClient) SetConfig(config *config.VaultConfig) error

func (*TestVaultClient) SetCreateTokenError added in v0.5.0

func (v *TestVaultClient) SetCreateTokenError(allocID, task string, err error)

SetCreateTokenError sets the error that will be returned by the token creation

func (*TestVaultClient) SetCreateTokenSecret added in v0.5.0

func (v *TestVaultClient) SetCreateTokenSecret(allocID, task string, secret *vapi.Secret)

SetCreateTokenSecret sets the secret that will be returned by the token creation

func (*TestVaultClient) SetLookupTokenAllowedPolicies added in v0.5.0

func (v *TestVaultClient) SetLookupTokenAllowedPolicies(token string, policies []string)

SetLookupTokenAllowedPolicies is a helper that adds a secret that allows the given policies

func (*TestVaultClient) SetLookupTokenError added in v0.5.0

func (v *TestVaultClient) SetLookupTokenError(token string, err error)

SetLookupTokenSecret sets the error that will be returned by the token lookup

func (*TestVaultClient) SetLookupTokenSecret added in v0.5.0

func (v *TestVaultClient) SetLookupTokenSecret(token string, secret *vapi.Secret)

SetLookupTokenSecret sets the secret that will be returned by the token lookup

func (*TestVaultClient) Stats added in v0.5.5

func (v *TestVaultClient) Stats() map[string]string

func (*TestVaultClient) Stop added in v0.5.0

func (v *TestVaultClient) Stop()

type TimeTable

type TimeTable struct {
	// contains filtered or unexported fields
}

TimeTable is used to associate a Raft index with a timestamp. This is used so that we can quickly go from a timestamp to an index or visa versa.

func NewTimeTable

func NewTimeTable(granularity time.Duration, limit time.Duration) *TimeTable

NewTimeTable creates a new time table which stores entries at a given granularity for a maximum limit. The storage space required is (limit/granularity)

func (*TimeTable) Deserialize

func (t *TimeTable) Deserialize(dec *codec.Decoder) error

Deserialize is used to deserialize the time table and restore the state

func (*TimeTable) NearestIndex

func (t *TimeTable) NearestIndex(when time.Time) uint64

NearestIndex returns the nearest index older than the given time

func (*TimeTable) NearestTime

func (t *TimeTable) NearestTime(index uint64) time.Time

NearestTime returns the nearest time older than the given index

func (*TimeTable) Serialize

func (t *TimeTable) Serialize(enc *codec.Encoder) error

Serialize is used to serialize the time table

func (*TimeTable) Witness

func (t *TimeTable) Witness(index uint64, when time.Time)

Witness is used to witness a new index and time.

type TimeTableEntry

type TimeTableEntry struct {
	Index uint64
	Time  time.Time
}

TimeTableEntry is used to track a time and index

type VaultClient added in v0.5.0

type VaultClient interface {
	// SetActive activates or de-activates the Vault client. When active, token
	// creation/lookup/revocation operation are allowed.
	SetActive(active bool)

	// SetConfig updates the config used by the Vault client
	SetConfig(config *config.VaultConfig) error

	// CreateToken takes an allocation and task and returns an appropriate Vault
	// Secret
	CreateToken(ctx context.Context, a *structs.Allocation, task string) (*vapi.Secret, error)

	// LookupToken takes a token string and returns its capabilities.
	LookupToken(ctx context.Context, token string) (*vapi.Secret, error)

	// RevokeTokens takes a set of tokens accessor and revokes the tokens
	RevokeTokens(ctx context.Context, accessors []*structs.VaultAccessor, committed bool) error

	// Stop is used to stop token renewal
	Stop()

	// Running returns whether the Vault client is running
	Running() bool

	// Stats returns the Vault clients statistics
	Stats() map[string]string

	// EmitStats emits that clients statistics at the given period until stopCh
	// is called.
	EmitStats(period time.Duration, stopCh <-chan struct{})
}

VaultClient is the Servers interface for interfacing with Vault

type VaultStats added in v0.5.5

type VaultStats struct {
	// TrackedForRevoke is the count of tokens that are being tracked to be
	// revoked since they could not be immediately revoked.
	TrackedForRevoke int

	// TokenTTL is the time-to-live duration for the current token
	TokenTTL time.Duration

	// TokenExpiry is the recorded expiry time of the current token
	TokenExpiry time.Time
}

VaultStats returns all the stats about Vault tokens created and managed by Nomad.

type Worker

type Worker struct {
	// contains filtered or unexported fields
}

Worker is a single threaded scheduling worker. There may be multiple running per server (leader or follower). They are responsible for dequeuing pending evaluations, invoking schedulers, plan submission and the lifecycle around making task allocations. They bridge the business logic of the scheduler with the plumbing required to make it all work.

func NewWorker

func NewWorker(srv *Server) (*Worker, error)

NewWorker starts a new worker associated with the given server

func (*Worker) CreateEval

func (w *Worker) CreateEval(eval *structs.Evaluation) error

CreateEval is used to create a new evaluation. This allows the worker to act as the planner for the scheduler.

func (*Worker) ReblockEval added in v0.4.0

func (w *Worker) ReblockEval(eval *structs.Evaluation) error

ReblockEval is used to reinsert a blocked evaluation into the blocked eval tracker. This allows the worker to act as the planner for the scheduler.

func (*Worker) SetPause

func (w *Worker) SetPause(p bool)

SetPause is used to pause or unpause a worker

func (*Worker) SubmitPlan

func (w *Worker) SubmitPlan(plan *structs.Plan) (*structs.PlanResult, scheduler.State, error)

SubmitPlan is used to submit a plan for consideration. This allows the worker to act as the planner for the scheduler.

func (*Worker) UpdateEval

func (w *Worker) UpdateEval(eval *structs.Evaluation) error

UpdateEval is used to submit an updated evaluation. This allows the worker to act as the planner for the scheduler.

Directories

Path Synopsis
deploymentwatcher creates and tracks Deployments, which hold meta data describing the process of upgrading a running job to a new set of Allocations.
deploymentwatcher creates and tracks Deployments, which hold meta data describing the process of upgrading a running job to a new set of Allocations.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL