Task Lifecycle

When Aurora reads a configuration file and finds a Job definition, it:

  1. Evaluates the Job definition.
  2. Splits the Job into its constituent Tasks.
  3. Sends those Tasks to the scheduler.
  4. The scheduler puts the Tasks into PENDING state, starting each Task’s life cycle.

Life of a task

Please note, a couple of task states described below are missing from this state diagram.

PENDING to RUNNING states

When a Task is in the PENDING state, the scheduler constantly searches for machines satisfying that Task’s resource request requirements (RAM, disk space, CPU time) while maintaining configuration constraints such as “a Task must run on machines dedicated to a particular role” or attribute limit constraints such as “at most 2 Tasks from the same Job may run on each rack”. When the scheduler finds a suitable match, it assigns the Task to a machine and puts the Task into the ASSIGNED state.

From the ASSIGNED state, the scheduler sends an RPC to the agent machine containing Task configuration, which the agent uses to spawn an executor responsible for the Task’s lifecycle. When the scheduler receives an acknowledgment that the machine has accepted the Task, the Task goes into STARTING state.

STARTING state initializes a Task sandbox. When the sandbox is fully initialized, Thermos begins to invoke Processes. Also, the agent machine sends an update to the scheduler that the Task is in RUNNING state, only after the task satisfies the liveness requirements. See Health Checking for more details for how to configure health checks.

RUNNING to terminal states

There are various ways that an active Task can transition into a terminal state. By definition, it can never leave this state. However, depending on nature of the termination and the originating Job definition (e.g. service, max_task_failures), a replacement Task might be scheduled.

Natural Termination: FINISHED, FAILED

A RUNNING Task can terminate without direct user interaction. For example, it may be a finite computation that finishes, even something as simple as echo hello world., or it could be an exceptional condition in a long-lived service. If the Task is successful (its underlying processes have succeeded with exit status 0 or finished without reaching failure limits) it moves into FINISHED state. If it finished after reaching a set of failure limits, it goes into FAILED state.

A terminated TASK which is subject to rescheduling will be temporarily THROTTLED, if it is considered to be flapping. A task is flapping, if its previous invocation was terminated after less than 5 minutes (scheduler default). The time penalty a task has to remain in the THROTTLED state, before it is eligible for rescheduling, increases with each consecutive failure.

Forceful Termination: KILLING, RESTARTING

You can terminate a Task by issuing an aurora job kill command, which moves it into KILLING state. The scheduler then sends the agent a request to terminate the Task. If the scheduler receives a successful response, it moves the Task into KILLED state and never restarts it.

If a Task is forced into the RESTARTING state via the aurora job restart command, the scheduler kills the underlying task but in parallel schedules an identical replacement for it.

In any case, the responsible executor on the agent follows an escalation sequence when killing a running task:

  1. If a HttpLifecycleConfig is not present, skip to (4).
  2. Send a POST to the graceful_shutdown_endpoint and wait graceful_shutdown_wait_secs seconds.
  3. Send a POST to the shutdown_endpoint and wait shutdown_wait_secs seconds.
  4. Send SIGTERM (kill) and wait at most finalization_wait seconds.
  5. Send SIGKILL (kill -9).

If the executor notices that all Processes in a Task have aborted during this sequence, it will not proceed with subsequent steps. Note that graceful shutdown is best-effort, and due to the many inevitable realities of distributed systems, it may not be performed.

Unexpected Termination: LOST

If a Task stays in a transient task state for too long (such as ASSIGNED or STARTING), the scheduler forces it into LOST state, creating a new Task in its place that’s sent into PENDING state.

In addition, if the Mesos core tells the scheduler that a agent has become unhealthy (or outright disappeared), the Tasks assigned to that agent go into LOST state and new Tasks are created in their place. From PENDING state, there is no guarantee a Task will be reassigned to the same machine unless job constraints explicitly force it there.

Giving Priority to Production Tasks: PREEMPTING

Sometimes a Task needs to be interrupted, such as when a non-production Task’s resources are needed by a higher priority production Task. This type of interruption is called a pre-emption. When this happens in Aurora, the non-production Task is killed and moved into the PREEMPTING state when both the following are true:

  • The task being killed is a non-production task.
  • The other task is a PENDING production task that hasn’t been scheduled due to a lack of resources.

The scheduler UI shows the non-production task was preempted in favor of the production task. At some point, tasks in PREEMPTING move to KILLED.

Note that non-production tasks consuming many resources are likely to be preempted in favor of production tasks.

Making Room for Maintenance: DRAINING

Cluster operators can set agent into maintenance mode. This will transition all Task running on this agent into DRAINING and eventually to KILLED. Drained Tasks will be restarted on other agents for which no maintenance has been announced yet.

State Reconciliation

Due to the many inevitable realities of distributed systems, there might be a mismatch of perceived and actual cluster state (e.g. a machine returns from a netsplit but the scheduler has already marked all its Tasks as LOST and rescheduled them).

Aurora regularly runs a state reconciliation process in order to detect and correct such issues (e.g. by killing the errant RUNNING tasks). By default, the proper detection of all failure scenarios and inconsistencies may take up to an hour.

To emphasize this point: there is no uniqueness guarantee for a single instance of a job in the presence of network partitions. If the Task requires that, it should be baked in at the application level using a distributed coordination service such as Zookeeper.