Job Information
A job, derived from JobInProgress, in general holds some basic information, including:
- jobId: BSPJobID. The id of this job.
- jobName: String. The name of this job.
- user: String. See BSPJobContext.getUser() wherein configures via conf.get("user.name")
- jobFile: Path. job.xml. See BSPJobClient.submitJobInternal()
- localJobFile: Path. A local copied jobFile.
- localJarFile: Path. A local copied jar file.
- numBSPTasks: int. The number of BSP tasks to be
- master: String. The name of master server that is responsible for this job.
- maxTaskAttempts: int. default to 2.
- MAX_TASK_EXEC: int. bound to 1.
- inputPath: Path. Dir where split files are stored. i/o.
- status: JobStatus. Mainly JobStatus.State. job status.
- progress: long. % of bsp(). job status.
- setupProgress: long. % of setup() job status.
- cleanupProgress: long. % of cleanup() job status.
- startTime: long. The time this job is started. job status.
- finishTime: long. The time this job is finished. job status.
- superstepCount: long. This job's superstep count. job satus.
- conf: Configuration. Specific setting for this job.
- Tasks: 2d Task array. See TaskInProgress and Job Task Relation section.
- failed: Boolean. If a task fails or not. default to false.
- NUM_ATTEMPTS_PER_RESTART: int. Set to 1000.
- partition: The _i_th numBSPTasks
- taskId: TaskID. Id of a task.
- attemptId: TaskAttemptID. AttemptId for a task.
- status: TaskStatus.
- Phase: SETUP, COMPUTE, BARRIER_SYNC, CLEANUP
- State: RUNNING, SUCCEEDED, FAILED, KILLED
- completed: Boolean. if a task is completed/ succeed. default to false.
- groom: String. On which groom a task runs.
- startTime: System.currentTimeMillis(). default -1.
- RawSplit: The _i_th RawSplit if RawSplit is not null
Job Status
A job status contains:
- PREP: This status is started when a job is created until all tasks are dispatched to groom servers.
- SetUp a job
- Scheding
- RUNNING: When tasks are executed in GroomServers.
- SUCCEEDED: When all tasks are in the SUCCEEDED state.
- FAILED: When one or more tasks exceed its max attempt
- KILLED: User/ Admin issues a kill command.
Task Phase
A single task phase.
- STARTING: Initial phase
- COMPUTE: Within bsp()
- BARRIER_SYNC: Reaching sync().
- CLEANUP: Finish entire supersteps execution.
Task State
This only indicate a task state.
- UNASSIGNED: Initial state of a task.
- RUNNING: Only valid when a task is either in COMPUTE or in BARRIER_SYNC phase.
- SUCCEEDED: A task completes its execution at the end of CLEANUP phase.
- FAILED: An exception or an error occurrs.
- KILLED
Job Task Relation