Accessing worker ID

[Schlevidon]

Hi, sorry if I missed something in the docs, but is it possible to access the worker ID within the worker-start-cmd for allocation queues (e.g. by using an environment variable)? Similarly, is there a way to retrieve the ID of the worker that was assigned to a job submission?

[Kobzol]

Hi :)
Regarding worker-start-cmd, this is a bit complicated. The start command is executed within a PBS/Slurm allocation even before a HQ worker is started, something like this:

#PBS ...
<worker-start-cmd>
hq worker start
<worker-stop-cmd>

Therefore, it's not really possible to know the worker ID here (at least with the current design), because the worker hasn't been started yet at the moment when worker-start-cmd is executed.

Regarding worker ID, workers are not assigned to jobs, but to individual tasks within a job. But even then it's not so simple! Because we have multi-node tasks, so a single task might have multiple workers, and furthermore, if a worker fails while executing a task, a different worker can be assigned to such task later. So it's not easy to say what is the worker ID of a task. That being said, if you execute hq task info <job-id> <task-id>, HQ will print the worker that is either currently executing the task, or the worker who has finished executing that task in the past.

[Schlevidon]

I see, that makes sense. Thank you very much for your explanation!