package peloton.api.v1alpha.job.stateless

Get desktop application:
View/edit binary Protocol Buffers messages

Configuration of a job creation.

Used in: svc.CreateJobRequest

• uint32 batch_size = 1

  Batch size for the creation which controls how many instances may be created at the same time.

• uint32 max_instance_retries = 2

  Maximum number of times a failing instance will be retried during the creation. If the value is 0, the instance can be retried for infinite times.

• uint32 max_tolerable_instance_failures = 3

  Maximum number of instance failures before the creation is declared to be failed. If the value is 0, there is no limit for max failure instances and the creation is marked successful even if all of the instances fail.

• bool start_paused = 4

  If set to true, indicates that the creation should start in the paused state, requiring an explicit resume to roll forward.

Information of a job, such as job spec and status

Used in: svc.GetJobResponse

• optional peloton.JobID job_id = 1

  Job ID

• optional JobSpec spec = 2

  Job configuration

• optional JobStatus status = 3

  Job runtime status

Stateless job configuration.

Used in: JobInfo, svc.CreateJobRequest, svc.GetJobCacheResponse, svc.GetReplaceJobDiffRequest, svc.PatchJobRequest, svc.ReplaceJobRequest, private.jobmgr.GetJobCacheResponse

• optional peloton.Revision revision = 1

  Revision of the job config

• string name = 2

  Name of the job

• string owner = 3

  Owner of the job

• string owning_team = 4

  Owning team of the job

• repeated string ldap_groups = 5

  LDAP groups of the job

• string description = 6

  Description of the job

• repeated peloton.Label labels = 7

  List of user-defined labels for the job

• uint32 instance_count = 8

  Number of instances of the job

• optional SlaSpec sla = 9

  SLA config of the job

• optional pod.PodSpec default_spec = 10

  Default pod configuration of the job

• map<uint32, pod.PodSpec> instance_spec = 11

  Instance specific pod config which overwrites the default one

• optional peloton.ResourcePoolID respool_id = 12

  Resource Pool ID where this job belongs to

Runtime states of a Job.

Used in: JobStatus, QuerySpec

• JOB_STATE_INVALID = 0

  Invalid job state.

• JOB_STATE_INITIALIZED = 1

  The job has been initialized and persisted in DB.

• JOB_STATE_PENDING = 2

  All tasks have been created and persisted in DB, but no task is RUNNING yet.

• JOB_STATE_RUNNING = 3

  Any of the tasks in the job is in RUNNING state.

• JOB_STATE_SUCCEEDED = 4

  All tasks in the job are in SUCCEEDED state.

• JOB_STATE_FAILED = 5

  All tasks in the job are in terminated state and one or more tasks is in FAILED state.

• JOB_STATE_KILLED = 6

  All tasks in the job are in terminated state and one or more tasks in the job is killed by the user.

• JOB_STATE_KILLING = 7

  All tasks in the job have been requested to be killed by the user.

• JOB_STATE_UNINITIALIZED = 8

  The job is partially created and is not ready to be scheduled

• JOB_STATE_DELETED = 9

  The job has been deleted.

The current runtime status of a Job.

Used in: JobInfo, JobSummary, svc.GetJobCacheResponse, private.jobmgr.GetJobCacheResponse

• optional peloton.Revision revision = 1

  Revision of the current job status. Version in the revision is incremented every time job status changes. Thus, it can be used to order the different job status updates.

• JobState state = 2

  State of the job

• string creation_time = 3

  The time when the job was created. The time is represented in RFC3339 form with UTC timezone.

• map<string, uint32> pod_stats = 4

  The number of pods grouped by each pod state. The map key is the pod.PodState in string format and the map value is the number of tasks in the particular state.

• JobState desired_state = 5

  Goal state of the job.

• optional peloton.EntityVersion version = 6

  The current version of the job. It is used to implement optimistic concurrency control for all job write APIs. The current job configuration can be fetched based on the current resource version.

• optional WorkflowStatus workflow_status = 7

  Status of ongoing update/restart workflow.

• map<string, uint32> pod_configuration_version_stats = 8

  DEPRECATED, will remove in the next release The number of tasks grouped by which configuration version they are on. The map key is the job configuration version and the map value is the number of tasks using that particular job configuration version. The job configuration version in the map key can be fed as the value of the entity version in the GetJobRequest to fetch the job configuration.

• map<string, JobStatus.PodStateStats> pod_stats_by_configuration_version = 9

  The number of pods grouped by which entity version and the state they are on. The map key is the entity version (which can be used to fetch the job configuration) and the map value is PodStateStats.

PodStateStats indicates the number of tasks of a particular state in a job

Used in: JobStatus

• map<string, uint32> state_stats = 1

  The map key is pod state, the value is number of pods running in that state.

Summary of job spec and status. The summary will be returned by List or Query API calls. These calls will return a large number of jobs, so the content in the job summary has to be kept minimal.

Used in: svc.GetJobResponse, svc.ListJobsResponse, svc.QueryJobsResponse, watch.svc.WatchResponse

• optional peloton.JobID job_id = 1

  Job ID

• string name = 2

  Name of the job

• string owner = 4

  Owner of the job

• string owning_team = 5

  Owning team of the job

• repeated peloton.Label labels = 6

  List of user-defined labels for the job

• uint32 instance_count = 7

  Number of instances of the job

• optional peloton.ResourcePoolID respool_id = 8

  Resource Pool ID where this job belongs to

• optional JobStatus status = 9

  Job runtime status

• optional SlaSpec sla = 10

  Job SLA Spec

QuerySpec specifies the list of query criteria for jobs. All indexed fields should be part of this message. And all fields in this message have to be indexed too.

Used in: svc.QueryJobsRequest, svc.QueryJobsResponse

• optional query.PaginationSpec pagination = 1

  The spec of how to do pagination for the query results.

• repeated peloton.Label labels = 2

  List of labels to query the jobs. Will match all jobs if the list is empty.

• repeated string keywords = 3

  List of keywords to query the jobs. Will match all jobs if the list is empty. When set, will do a wildcard match on owner, name, labels, description.

• repeated JobState job_states = 4

  List of job states to query the jobs. Will match all jobs if the list is empty.

• optional respool.ResourcePoolPath respool = 5

  The resource pool to query the jobs. Will match jobs from all resource pools if unset.

• string owner = 6

  Query jobs by owner. This is case sensitive and will look for jobs with owner matching the exact owner string. Will match all jobs if owner is unset.

• string name = 7

  Query jobs by name. This is case sensitive and will look for jobs with name matching the name string. Will support partial name match. Will match all jobs if name is unset.

• optional peloton.TimeRange creation_time_range = 8

  Query jobs by creation time range. This will look for all jobs that were created within a specified time range. This search will operate based on job creation time.

• optional peloton.TimeRange completion_time_range = 9

  Query jobs by completion time range. This will look for all jobs that were completed within a specified time range. This search will operate based on job completion time.

Configuration of a job restart

Used in: WorkflowInfo, svc.RestartJobRequest

• uint32 batch_size = 1

  Batch size for the restart which controls how many instances may be restarted at the same time.

• repeated pod.InstanceIDRange ranges = 2

  The pods to restart, default to all.

• bool in_place = 3

  If set to true, peloton would try to place the task restarted on the host it previously run on. It is best effort, and has no guarantee of success.

SLA configuration for a stateless job

Used in: JobSpec, JobSummary

• uint32 priority = 1

  Priority of a job. Higher value takes priority over lower value when making scheduling decisions as well as preemption decisions.

• bool preemptible = 2

  Whether all the job instances are preemptible. If so, it might be scheduled elastic resources from other resource pools and subject to preemption when the demands of other resource pools increase. For stateless jobs, this field will overrule preemptible configuration in the pod spec.

• bool revocable = 3

  Whether all the job instances are revocable. If so, it might be scheduled using revocable resources and subject to preemption when there is resource contention on the host. For stateless jobs, this field will overrule revocable configuration in the pod spec.

• uint32 maximum_unavailable_instances = 4

  Maximum number of job instances which can be unavailable at a given time. When job instances are to be moved from one host to another for host maintenance and/or load redistribution, Peloton ensures number of job instances unavailable at a given time doesn't exceed the number configured here. However, the instances that become unavailable due to job update can exceed this number.

Configuration of a job update.

Used in: WorkflowInfo, svc.PatchJobRequest, svc.ReplaceJobRequest

• uint32 batch_size = 1

  Batch size for the update which controls how many instances may be updated at the same time.

• bool rollback_on_failure = 2

  If configured, the update be automatically rolled back to the previous job configuration on failure.

• uint32 max_instance_retries = 3

  Maximum number of times a failing instance will be retried during the update. If the value is 0, the instance can be retried for infinite times.

• uint32 max_tolerable_instance_failures = 4

  Maximum number of instance failures before the update is declared to be failed. If the value is 0, there is no limit for max failure instances and the update is marked successful even if all of the instances fail.

• bool start_paused = 5

  If set to true, indicates that the update should start in the paused state, requiring an explicit resume to roll forward.

• bool in_place = 6

  If set to true, peloton would try to place the task updated on the host it previously run on. It is best effort, and has no guarantee of success.

• bool start_pods = 7

  Experimental: this flag is for experimental purpose and will be removed in the future release. When set to true, peloton would start pods updated, if the pods are killed previously. By default, killed pods would remain killed, and run with new version when running again.

WorkflowEvents are workflow state change events for a job or pod on workflow operations

Used in: WorkflowInfo, WorkflowInfo.instance_workflow_events, svc.GetWorkflowEventsResponse

• WorkflowType type = 1

  Workflow type.

• string timestamp = 2

  Timestamp of the event represented in RFC3339 form with UTC timezone.

• WorkflowState state = 3

  Current runtime state of the workflow.

Information about a workflow including its status and specification

Used in: svc.GetJobResponse, svc.ListJobWorkflowsResponse

• optional WorkflowStatus status = 1

  Workflow status

• optional UpdateSpec update_spec = 2

  Update specification for update workflow

• optional RestartSpec restart_spec = 3

  Restart specification provided for restart workflow

• optional peloton.OpaqueData opaque_data = 6

  Opaque data supplied by the client

• repeated WorkflowEvent events = 7

  job workflow events represents update state changes

• repeated pod.InstanceIDRange instances_added = 8

  Instances added by update workflow

• repeated pod.InstanceIDRange instances_removed = 9

  Instances removed by update workflow

• repeated pod.InstanceIDRange instances_updated = 10

  Instances updated by update workflow

• repeated WorkflowInfo.instance_workflow_events instance_events = 11

Used in: WorkflowInfo

• uint32 instance_id = 1

  The instance ID

• repeated WorkflowEvent events = 2

  Instance workflow events

Runtime state of a job workflow.

Used in: WorkflowEvent, WorkflowStatus

• WORKFLOW_STATE_INVALID = 0

  Invalid protobuf value

• WORKFLOW_STATE_INITIALIZED = 1

  The operation has been created but not started yet.

• WORKFLOW_STATE_ROLLING_FORWARD = 2

  The workflow is rolling forward

• WORKFLOW_STATE_PAUSED = 3

  The workflow has been paused

• WORKFLOW_STATE_SUCCEEDED = 4

  The workflow has completed successfully

• WORKFLOW_STATE_ABORTED = 5

  The update was aborted/cancelled

• WORKFLOW_STATE_FAILED = 6

  The workflow has failed to complete.

• WORKFLOW_STATE_ROLLING_BACKWARD = 7

  The update is rolling backward

• WORKFLOW_STATE_ROLLED_BACK = 8

  The update was rolled back due to failure

Runtime status of a job workflow.

Used in: JobStatus, WorkflowInfo

• WorkflowType type = 1

  Workflow type.

• WorkflowState state = 2

  Current runtime state of the workflow.

• uint32 num_instances_completed = 3

  Number of instances completed.

• uint32 num_instances_remaining = 4

  Number of instances remaining.

• uint32 num_instances_failed = 5

  Number of instances which failed to come up after the workflow.

• repeated uint32 instances_current = 6

  Current instances being operated on.

• optional peloton.EntityVersion version = 7

  Job version the workflow moved the job object to.

• optional peloton.EntityVersion prev_version = 8

  Previous job version of the job object.

• string creation_time = 9

  The time when the workflow was created. The time is represented in RFC3339 form with UTC timezone.

• string update_time = 10

  The time when the workflow was last updated. The time is represented in RFC3339 form with UTC timezone.

• WorkflowState prev_state = 11

  Previous runtime state of the workflow.

• string completion_time = 12

  The time when the workflow completed. The time is represented in RFC3339 form with UTC timezone.

The different types of job rolling workflows supported.

Used in: WorkflowEvent, WorkflowStatus

• WORKFLOW_TYPE_INVALID = 0

  Invalid protobuf value.

• WORKFLOW_TYPE_UPDATE = 1

  Job update workflow.

• WORKFLOW_TYPE_RESTART = 2

  Restart pods in a job.