open-nomad/nomad/state/schema.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

963 lines
22 KiB
Go

package state
import (
"fmt"
"sync"
memdb "github.com/hashicorp/go-memdb"
"github.com/hashicorp/nomad/nomad/structs"
)
const (
TableNamespaces = "namespaces"
)
var (
schemaFactories SchemaFactories
factoriesLock sync.Mutex
)
// SchemaFactory is the factory method for returning a TableSchema
type SchemaFactory func() *memdb.TableSchema
type SchemaFactories []SchemaFactory
// RegisterSchemaFactories is used to register a table schema.
func RegisterSchemaFactories(factories ...SchemaFactory) {
factoriesLock.Lock()
defer factoriesLock.Unlock()
schemaFactories = append(schemaFactories, factories...)
}
func GetFactories() SchemaFactories {
return schemaFactories
}
func init() {
// Register all schemas
RegisterSchemaFactories([]SchemaFactory{
indexTableSchema,
nodeTableSchema,
jobTableSchema,
jobSummarySchema,
jobVersionSchema,
deploymentSchema,
periodicLaunchTableSchema,
evalTableSchema,
allocTableSchema,
vaultAccessorTableSchema,
siTokenAccessorTableSchema,
aclPolicyTableSchema,
aclTokenTableSchema,
oneTimeTokenTableSchema,
autopilotConfigTableSchema,
schedulerConfigTableSchema,
clusterMetaTableSchema,
csiVolumeTableSchema,
csiPluginTableSchema,
scalingPolicyTableSchema,
scalingEventTableSchema,
namespaceTableSchema,
}...)
}
// stateStoreSchema is used to return the schema for the state store
func stateStoreSchema() *memdb.DBSchema {
// Create the root DB schema
db := &memdb.DBSchema{
Tables: make(map[string]*memdb.TableSchema),
}
// Add each of the tables
for _, schemaFn := range GetFactories() {
schema := schemaFn()
if _, ok := db.Tables[schema.Name]; ok {
panic(fmt.Sprintf("duplicate table name: %s", schema.Name))
}
db.Tables[schema.Name] = schema
}
return db
}
// indexTableSchema is used for tracking the most recent index used for each table.
func indexTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "index",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "Key",
Lowercase: true,
},
},
},
}
}
// nodeTableSchema returns the MemDB schema for the nodes table.
// This table is used to store all the client nodes that are registered.
func nodeTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "nodes",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is used for node management
// and simple direct lookup. ID is required to be
// unique.
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "ID",
},
},
"secret_id": {
Name: "secret_id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "SecretID",
},
},
},
}
}
// jobTableSchema returns the MemDB schema for the jobs table.
// This table is used to store all the jobs that have been submitted.
func jobTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "jobs",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is used for job management
// and simple direct lookup. ID is required to be
// unique within a namespace.
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// Use a compound index so the tuple of (Namespace, ID) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "ID",
},
},
},
},
"type": {
Name: "type",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Type",
Lowercase: false,
},
},
"gc": {
Name: "gc",
AllowMissing: false,
Unique: false,
Indexer: &memdb.ConditionalIndex{
Conditional: jobIsGCable,
},
},
"periodic": {
Name: "periodic",
AllowMissing: false,
Unique: false,
Indexer: &memdb.ConditionalIndex{
Conditional: jobIsPeriodic,
},
},
},
}
}
// jobSummarySchema returns the memdb schema for the job summary table
func jobSummarySchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "job_summary",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// Use a compound index so the tuple of (Namespace, JobID) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "JobID",
},
},
},
},
},
}
}
// jobVersionSchema returns the memdb schema for the job version table which
// keeps a historical view of job versions.
func jobVersionSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "job_version",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// Use a compound index so the tuple of (Namespace, ID, Version) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "ID",
Lowercase: true,
},
&memdb.UintFieldIndex{
Field: "Version",
},
},
},
},
},
}
}
// jobIsGCable satisfies the ConditionalIndexFunc interface and creates an index
// on whether a job is eligible for garbage collection.
func jobIsGCable(obj interface{}) (bool, error) {
j, ok := obj.(*structs.Job)
if !ok {
return false, fmt.Errorf("Unexpected type: %v", obj)
}
// If the job is periodic or parameterized it is only garbage collectable if
// it is stopped.
periodic := j.Periodic != nil && j.Periodic.Enabled
parameterized := j.IsParameterized()
if periodic || parameterized {
return j.Stop, nil
}
// If the job isn't dead it isn't eligible
if j.Status != structs.JobStatusDead {
return false, nil
}
// Any job that is stopped is eligible for garbage collection
if j.Stop {
return true, nil
}
switch j.Type {
// Otherwise, batch and sysbatch jobs are eligible because they complete on
// their own without a user stopping them.
case structs.JobTypeBatch, structs.JobTypeSysBatch:
return true, nil
default:
// other job types may not be GC until stopped
return false, nil
}
}
// jobIsPeriodic satisfies the ConditionalIndexFunc interface and creates an index
// on whether a job is periodic.
func jobIsPeriodic(obj interface{}) (bool, error) {
j, ok := obj.(*structs.Job)
if !ok {
return false, fmt.Errorf("Unexpected type: %v", obj)
}
if j.Periodic != nil && j.Periodic.Enabled {
return true, nil
}
return false, nil
}
// deploymentSchema returns the MemDB schema tracking a job's deployments
func deploymentSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "deployment",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "ID",
},
},
"namespace": {
Name: "namespace",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Namespace",
},
},
// Job index is used to lookup deployments by job
"job": {
Name: "job",
AllowMissing: false,
Unique: false,
// Use a compound index so the tuple of (Namespace, JobID) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "JobID",
},
},
},
},
},
}
}
// periodicLaunchTableSchema returns the MemDB schema tracking the most recent
// launch time for a periodic job.
func periodicLaunchTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "periodic_launch",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is used for job management
// and simple direct lookup. ID is required to be
// unique.
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// Use a compound index so the tuple of (Namespace, JobID) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "ID",
},
},
},
},
},
}
}
// evalTableSchema returns the MemDB schema for the eval table.
// This table is used to store all the evaluations that are pending
// or recently completed.
func evalTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "evals",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is used for direct lookup.
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "ID",
},
},
"namespace": {
Name: "namespace",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Namespace",
},
},
// Job index is used to lookup allocations by job
"job": {
Name: "job",
AllowMissing: false,
Unique: false,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "JobID",
Lowercase: true,
},
&memdb.StringFieldIndex{
Field: "Status",
Lowercase: true,
},
},
},
},
},
}
}
// allocTableSchema returns the MemDB schema for the allocation table.
// This table is used to store all the task allocations between task groups
// and nodes.
func allocTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "allocs",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is a UUID
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "ID",
},
},
"namespace": {
Name: "namespace",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Namespace",
},
},
// Node index is used to lookup allocations by node
"node": {
Name: "node",
AllowMissing: true, // Missing is allow for failed allocations
Unique: false,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "NodeID",
Lowercase: true,
},
// Conditional indexer on if allocation is terminal
&memdb.ConditionalIndex{
Conditional: func(obj interface{}) (bool, error) {
// Cast to allocation
alloc, ok := obj.(*structs.Allocation)
if !ok {
return false, fmt.Errorf("wrong type, got %t should be Allocation", obj)
}
// Check if the allocation is terminal
return alloc.TerminalStatus(), nil
},
},
},
},
},
// Job index is used to lookup allocations by job
"job": {
Name: "job",
AllowMissing: false,
Unique: false,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "JobID",
},
},
},
},
// Eval index is used to lookup allocations by eval
"eval": {
Name: "eval",
AllowMissing: false,
Unique: false,
Indexer: &memdb.UUIDFieldIndex{
Field: "EvalID",
},
},
// Deployment index is used to lookup allocations by deployment
"deployment": {
Name: "deployment",
AllowMissing: true,
Unique: false,
Indexer: &memdb.UUIDFieldIndex{
Field: "DeploymentID",
},
},
},
}
}
// vaultAccessorTableSchema returns the MemDB schema for the Vault Accessor
// Table. This table tracks Vault accessors for tokens created on behalf of
// allocations required Vault tokens.
func vaultAccessorTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "vault_accessors",
Indexes: map[string]*memdb.IndexSchema{
// The primary index is the accessor id
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "Accessor",
},
},
"alloc_id": {
Name: "alloc_id",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "AllocID",
},
},
"node_id": {
Name: "node_id",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "NodeID",
},
},
},
}
}
// siTokenAccessorTableSchema returns the MemDB schema for the Service Identity
// token accessor table. This table tracks accessors for tokens created on behalf
// of allocations with Consul connect enabled tasks that need SI tokens.
func siTokenAccessorTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: siTokenAccessorTable,
Indexes: map[string]*memdb.IndexSchema{
// The primary index is the accessor id
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "AccessorID",
},
},
"alloc_id": {
Name: "alloc_id",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "AllocID",
},
},
"node_id": {
Name: "node_id",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "NodeID",
},
},
},
}
}
// aclPolicyTableSchema returns the MemDB schema for the policy table.
// This table is used to store the policies which are referenced by tokens
func aclPolicyTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "acl_policy",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "Name",
},
},
},
}
}
// aclTokenTableSchema returns the MemDB schema for the tokens table.
// This table is used to store the bearer tokens which are used to authenticate
func aclTokenTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "acl_token",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "AccessorID",
},
},
"secret": {
Name: "secret",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "SecretID",
},
},
"global": {
Name: "global",
AllowMissing: false,
Unique: false,
Indexer: &memdb.FieldSetIndex{
Field: "Global",
},
},
},
}
}
// oneTimeTokenTableSchema returns the MemDB schema for the tokens table.
// This table is used to store one-time tokens for ACL tokens
func oneTimeTokenTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "one_time_token",
Indexes: map[string]*memdb.IndexSchema{
"secret": {
Name: "secret",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "OneTimeSecretID",
},
},
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.UUIDFieldIndex{
Field: "AccessorID",
},
},
},
}
}
// singletonRecord can be used to describe tables which should contain only 1 entry.
// Example uses include storing node config or cluster metadata blobs.
var singletonRecord = &memdb.ConditionalIndex{
Conditional: func(interface{}) (bool, error) { return true, nil },
}
// schedulerConfigTableSchema returns the MemDB schema for the scheduler config table.
// This table is used to store configuration options for the scheduler
func schedulerConfigTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "scheduler_config",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: true,
Unique: true,
Indexer: singletonRecord, // we store only 1 scheduler config
},
},
}
}
// clusterMetaTableSchema returns the MemDB schema for the scheduler config table.
func clusterMetaTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "cluster_meta",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: singletonRecord, // we store only 1 cluster metadata
},
},
}
}
// CSIVolumes are identified by id globally, and searchable by driver
func csiVolumeTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "csi_volumes",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "ID",
},
},
},
},
"plugin_id": {
Name: "plugin_id",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "PluginID",
},
},
},
}
}
// CSIPlugins are identified by id globally, and searchable by driver
func csiPluginTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "csi_plugins",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "ID",
},
},
},
}
}
// StringFieldIndex is used to extract a field from an object
// using reflection and builds an index on that field.
type ScalingPolicyTargetFieldIndex struct {
Field string
// AllowMissing controls if the field should be ignored if the field is
// not provided.
AllowMissing bool
}
// FromObject is used to extract an index value from an
// object or to indicate that the index value is missing.
func (s *ScalingPolicyTargetFieldIndex) FromObject(obj interface{}) (bool, []byte, error) {
policy, ok := obj.(*structs.ScalingPolicy)
if !ok {
return false, nil, fmt.Errorf("object %#v is not a ScalingPolicy", obj)
}
if policy.Target == nil {
return false, nil, nil
}
val, ok := policy.Target[s.Field]
if !ok && !s.AllowMissing {
return false, nil, nil
}
// Add the null character as a terminator
val += "\x00"
return true, []byte(val), nil
}
// FromArgs is used to build an exact index lookup based on arguments
func (s *ScalingPolicyTargetFieldIndex) FromArgs(args ...interface{}) ([]byte, error) {
if len(args) != 1 {
return nil, fmt.Errorf("must provide only a single argument")
}
arg, ok := args[0].(string)
if !ok {
return nil, fmt.Errorf("argument must be a string: %#v", args[0])
}
// Add the null character as a terminator
arg += "\x00"
return []byte(arg), nil
}
// PrefixFromArgs returns a prefix that should be used for scanning based on the arguments
func (s *ScalingPolicyTargetFieldIndex) PrefixFromArgs(args ...interface{}) ([]byte, error) {
val, err := s.FromArgs(args...)
if err != nil {
return nil, err
}
// Strip the null terminator, the rest is a prefix
n := len(val)
if n > 0 {
return val[:n-1], nil
}
return val, nil
}
// scalingPolicyTableSchema returns the MemDB schema for the policy table.
func scalingPolicyTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "scaling_policy",
Indexes: map[string]*memdb.IndexSchema{
// Primary index is used for simple direct lookup.
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// UUID is uniquely identifying
Indexer: &memdb.StringFieldIndex{
Field: "ID",
},
},
// Target index is used for listing by namespace or job, or looking up a specific target.
// A given task group can have only a single scaling policies, so this is guaranteed to be unique.
"target": {
Name: "target",
Unique: false,
// Use a compound index so the tuple of (Namespace, Job, Group, Task) is
// used when looking for a policy
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&ScalingPolicyTargetFieldIndex{
Field: "Namespace",
AllowMissing: true,
},
&ScalingPolicyTargetFieldIndex{
Field: "Job",
AllowMissing: true,
},
&ScalingPolicyTargetFieldIndex{
Field: "Group",
AllowMissing: true,
},
&ScalingPolicyTargetFieldIndex{
Field: "Task",
AllowMissing: true,
},
},
},
},
// Type index is used for listing by policy type
"type": {
Name: "type",
AllowMissing: false,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Type",
},
},
// Used to filter by enabled
"enabled": {
Name: "enabled",
AllowMissing: false,
Unique: false,
Indexer: &memdb.FieldSetIndex{
Field: "Enabled",
},
},
},
}
}
// scalingEventTableSchema returns the memdb schema for job scaling events
func scalingEventTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: "scaling_event",
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
// Use a compound index so the tuple of (Namespace, JobID) is
// uniquely identifying
Indexer: &memdb.CompoundIndex{
Indexes: []memdb.Indexer{
&memdb.StringFieldIndex{
Field: "Namespace",
},
&memdb.StringFieldIndex{
Field: "JobID",
},
},
},
},
// TODO: need to figure out whether we want to index these or the jobs or ...
// "error": {
// Name: "error",
// AllowMissing: false,
// Unique: false,
// Indexer: &memdb.FieldSetIndex{
// Field: "Error",
// },
// },
},
}
}
// namespaceTableSchema returns the MemDB schema for the namespace table.
func namespaceTableSchema() *memdb.TableSchema {
return &memdb.TableSchema{
Name: TableNamespaces,
Indexes: map[string]*memdb.IndexSchema{
"id": {
Name: "id",
AllowMissing: false,
Unique: true,
Indexer: &memdb.StringFieldIndex{
Field: "Name",
},
},
"quota": {
Name: "quota",
AllowMissing: true,
Unique: false,
Indexer: &memdb.StringFieldIndex{
Field: "Quota",
},
},
},
}
}