open-nomad/nomad/config.go
Seth Hoenig 3371214431 core: implement system batch scheduler
This PR implements a new "System Batch" scheduler type. Jobs can
make use of this new scheduler by setting their type to 'sysbatch'.

Like the name implies, sysbatch can be thought of as a hybrid between
system and batch jobs - it is for running short lived jobs intended to
run on every compatible node in the cluster.

As with batch jobs, sysbatch jobs can also be periodic and/or parameterized
dispatch jobs. A sysbatch job is considered complete when it has been run
on all compatible nodes until reaching a terminal state (success or failed
on retries).

Feasibility and preemption are governed the same as with system jobs. In
this PR, the update stanza is not yet supported. The update stanza is sill
limited in functionality for the underlying system scheduler, and is
not useful yet for sysbatch jobs. Further work in #4740 will improve
support for the update stanza and deployments.

Closes #2527
2021-08-03 10:30:47 -04:00

486 lines
18 KiB
Go

package nomad
import (
"fmt"
"io"
"net"
"os"
"runtime"
"time"
log "github.com/hashicorp/go-hclog"
"github.com/hashicorp/memberlist"
"github.com/hashicorp/nomad/helper/pluginutils/loader"
"github.com/hashicorp/nomad/helper/uuid"
"github.com/hashicorp/nomad/nomad/deploymentwatcher"
"github.com/hashicorp/nomad/nomad/structs"
"github.com/hashicorp/nomad/nomad/structs/config"
"github.com/hashicorp/nomad/scheduler"
"github.com/hashicorp/raft"
"github.com/hashicorp/serf/serf"
)
const (
DefaultRegion = "global"
DefaultDC = "dc1"
DefaultSerfPort = 4648
)
// These are the protocol versions that Nomad can understand
const (
ProtocolVersionMin uint8 = 1
ProtocolVersionMax = 1
)
// ProtocolVersionMap is the mapping of Nomad protocol versions
// to Serf protocol versions. We mask the Serf protocols using
// our own protocol version.
var protocolVersionMap map[uint8]uint8
func init() {
protocolVersionMap = map[uint8]uint8{
1: 4,
}
}
func DefaultRPCAddr() *net.TCPAddr {
return &net.TCPAddr{IP: net.ParseIP("127.0.0.1"), Port: 4647}
}
// Config is used to parameterize the server
type Config struct {
// Bootstrapped indicates if Server has bootstrapped or not.
// Its value must be 0 (not bootstrapped) or 1 (bootstrapped).
// All operations on Bootstrapped must be handled via `atomic.*Int32()` calls
Bootstrapped int32
// 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.
//
// The BootstrapExpect can be of any of the following values:
// 1: Server will form a single node cluster and become a leader immediately
// N, larger than 1: Server will wait until it's connected to N servers
// before attempting leadership and forming the cluster. No Raft Log operation
// will succeed until then.
// 0: Server will wait to get a Raft configuration from another node and may not
// attempt to form a cluster or establish leadership on its own.
BootstrapExpect int
// 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
// EnableDebug is used to enable debugging RPC endpoints
// in the absence of ACLs
EnableDebug bool
// EnableEventBroker is used to enable or disable state store
// event publishing
EnableEventBroker bool
// EventBufferSize is the amount of events to hold in memory.
EventBufferSize int64
// 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.InterceptLogger
// 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
// CSIPluginGCInterval is how often we dispatch a job to GC unused plugins.
CSIPluginGCInterval time.Duration
// CSIPluginGCThreshold is how "old" a plugin must be to be eligible
// for GC. This gives users some time to debug plugins.
CSIPluginGCThreshold time.Duration
// CSIVolumeClaimGCInterval is how often we dispatch a job to GC
// volume claims.
CSIVolumeClaimGCInterval time.Duration
// CSIVolumeClaimGCThreshold is how "old" a volume must be to be
// eligible for GC. This gives users some time to debug volumes.
CSIVolumeClaimGCThreshold time.Duration
// OneTimeTokenGCInterval is how often we dispatch a job to GC
// one-time tokens.
OneTimeTokenGCInterval 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
// DisableDispatchedJobSummaryMetrics allows for ignore dispatched jobs when
// publishing Job summary metrics
DisableDispatchedJobSummaryMetrics 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
// DefaultSchedulerConfig configures the initial scheduler config to be persisted in Raft.
// Once the cluster is bootstrapped, and Raft persists the config (from here or through API)
// and this value is ignored.
DefaultSchedulerConfig structs.SchedulerConfiguration `hcl:"default_scheduler_config"`
// 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
// RPCHandshakeTimeout is the deadline by which RPC handshakes must
// complete. The RPC handshake includes the first byte read as well as
// the TLS handshake and subsequent byte read if TLS is enabled.
//
// The deadline is reset after the first byte is read so when TLS is
// enabled RPC connections may take (timeout * 2) to complete.
//
// 0 means no timeout.
RPCHandshakeTimeout time.Duration
// RPCMaxConnsPerClient is the maximum number of concurrent RPC
// connections from a single IP address. nil/0 means no limit.
RPCMaxConnsPerClient int
// LicenseConfig is a tunable knob for enterprise license testing.
LicenseConfig *LicenseConfig
LicenseEnv string
LicensePath string
// SearchConfig provides knobs for Search API.
SearchConfig *structs.SearchConfig
// AgentShutdown is used to call agent.Shutdown from the context of a Server
// It is used primarily for licensing
AgentShutdown func() error
// DeploymentQueryRateLimit is in queries per second and is used by the
// DeploymentWatcher to throttle the amount of simultaneously deployments
DeploymentQueryRateLimit float64
}
// CheckVersion is used to check if the ProtocolVersion is valid
func (c *Config) CheckVersion() error {
if c.ProtocolVersion < ProtocolVersionMin {
return fmt.Errorf("Protocol version '%d' too low. Must be in range: [%d, %d]",
c.ProtocolVersion, ProtocolVersionMin, ProtocolVersionMax)
} else if c.ProtocolVersion > ProtocolVersionMax {
return fmt.Errorf("Protocol version '%d' too high. Must be in range: [%d, %d]",
c.ProtocolVersion, ProtocolVersionMin, ProtocolVersionMax)
}
return nil
}
// DefaultConfig returns the default configuration. Only used as the basis for
// merging agent or test parameters.
func DefaultConfig() *Config {
hostname, err := os.Hostname()
if err != nil {
panic(err)
}
c := &Config{
Region: DefaultRegion,
AuthoritativeRegion: DefaultRegion,
Datacenter: DefaultDC,
NodeName: hostname,
NodeID: uuid.Generate(),
ProtocolVersion: ProtocolVersionMax,
RaftConfig: raft.DefaultConfig(),
RaftTimeout: 10 * time.Second,
LogOutput: os.Stderr,
RPCAddr: DefaultRPCAddr(),
SerfConfig: serf.DefaultConfig(),
NumSchedulers: 1,
ReconcileInterval: 60 * time.Second,
EvalGCInterval: 5 * time.Minute,
EvalGCThreshold: 1 * time.Hour,
JobGCInterval: 5 * time.Minute,
JobGCThreshold: 4 * time.Hour,
NodeGCInterval: 5 * time.Minute,
NodeGCThreshold: 24 * time.Hour,
DeploymentGCInterval: 5 * time.Minute,
DeploymentGCThreshold: 1 * time.Hour,
CSIPluginGCInterval: 5 * time.Minute,
CSIPluginGCThreshold: 1 * time.Hour,
CSIVolumeClaimGCInterval: 5 * time.Minute,
CSIVolumeClaimGCThreshold: 5 * time.Minute,
OneTimeTokenGCInterval: 10 * time.Minute,
EvalNackTimeout: 60 * time.Second,
EvalDeliveryLimit: 3,
EvalNackInitialReenqueueDelay: 1 * time.Second,
EvalNackSubsequentReenqueueDelay: 20 * time.Second,
EvalFailedFollowupBaselineDelay: 1 * time.Minute,
EvalFailedFollowupDelayRange: 5 * time.Minute,
MinHeartbeatTTL: 10 * time.Second,
MaxHeartbeatsPerSecond: 50.0,
HeartbeatGrace: 10 * time.Second,
FailoverHeartbeatTTL: 300 * time.Second,
ConsulConfig: config.DefaultConsulConfig(),
VaultConfig: config.DefaultVaultConfig(),
RPCHoldTimeout: 5 * time.Second,
StatsCollectionInterval: 1 * time.Minute,
TLSConfig: &config.TLSConfig{},
ReplicationBackoff: 30 * time.Second,
SentinelGCInterval: 30 * time.Second,
LicenseConfig: &LicenseConfig{},
EnableEventBroker: true,
EventBufferSize: 100,
AutopilotConfig: &structs.AutopilotConfig{
CleanupDeadServers: true,
LastContactThreshold: 200 * time.Millisecond,
MaxTrailingLogs: 250,
ServerStabilizationTime: 10 * time.Second,
},
ServerHealthInterval: 2 * time.Second,
AutopilotInterval: 10 * time.Second,
DefaultSchedulerConfig: structs.SchedulerConfiguration{
SchedulerAlgorithm: structs.SchedulerAlgorithmBinpack,
PreemptionConfig: structs.PreemptionConfig{
SystemSchedulerEnabled: true,
SysBatchSchedulerEnabled: false,
BatchSchedulerEnabled: false,
ServiceSchedulerEnabled: false,
},
},
DeploymentQueryRateLimit: deploymentwatcher.LimitStateQueriesPerSecond,
}
// Enable all known schedulers by default
c.EnabledSchedulers = make([]string, 0, len(scheduler.BuiltinSchedulers))
for name := range scheduler.BuiltinSchedulers {
c.EnabledSchedulers = append(c.EnabledSchedulers, name)
}
c.EnabledSchedulers = append(c.EnabledSchedulers, structs.JobTypeCore)
// Default the number of schedulers to match the cores
c.NumSchedulers = runtime.NumCPU()
// Increase our reap interval to 3 days instead of 24h.
c.SerfConfig.ReconnectTimeout = 3 * 24 * time.Hour
// Serf should use the WAN timing, since we are using it
// to communicate between DC's
c.SerfConfig.MemberlistConfig = memberlist.DefaultWANConfig()
c.SerfConfig.MemberlistConfig.BindPort = DefaultSerfPort
// Disable shutdown on removal
c.RaftConfig.ShutdownOnRemove = false
// Default to Raft v2, update to v3 to enable new Raft and autopilot features.
c.RaftConfig.ProtocolVersion = 2
return c
}