package tws.proto.job

Get desktop application:
View/edit binary Protocol Buffers messages

compute resource for a worker cpu as a double, can be fractional RAM as mega bytes Disk as giga bytes. Disk is volatile disk. instances: number of instances that will be created from this compute resource workers_per_pod: in kubernetes, on each compute resource (on each pod) this many workers will be started number of workers running on this resource type is: instances * workers_per_pod index: a unique index is assigned to each resource starting from 0

Used in: Job, jobmaster.WorkerInfo

• double cpu = 1

• int32 ram_mega_bytes = 2

• double disk_giga_bytes = 3

• int32 instances = 4

• int32 workers_per_pod = 5

• oneof scalableRequired

  • bool scalable = 6

• oneof indexRequired

  • int32 index = 7

defines the job parameters. This will define the required resources for the job, what type of a job this is (a basic container job, fully featured dataflow job etc)

Used in: htgjob.SubGraph, Job

• repeated Config.KeyValue kvs = 1

• map<string, bytes> config_byte_map = 2

Used in: Config

• string key = 1

• string value = 2

  Only one of the below are set

• bytes serialized_value = 3

this is a Twister job.

• string job_name = 1

• string job_id = 2

  automatically generated job-id

• string worker_class_name = 3

  worker class to run

• string driver_class_name = 4

  driver class to run

• int32 number_of_workers = 5

• repeated ComputeResource compute_resource = 6

  the requested resources

• optional Config config = 7

  the configurations specified

• optional JobFormat job_format = 8

  how the job is layed out

Used in: Job

• JobFormatType type = 1

• string job_file = 2

Used in: JobFormat

• JAR = 0

• JAVA_ZIP = 1

• PYTHON = 2

• PYTHON_ZIP = 3

a worker can be any one of these states

• STARTING = 0

• STARTED = 1

• COMPLETED = 2

• FAILED = 3

• KILLED = 4

• string job_name = 1

• bool failure = 2

  If the worker terminated with an error

• string worker_message = 3

• string stack_trace = 4

  If the worker failed with an exception the full stack trace is set here