Project

Profile

Help

Issue #7912

closed

The tasking system deadlocks when Redis looses its tasks

Added by bmbouter almost 2 years ago. Updated over 1 year ago.

Status:
CLOSED - CURRENTRELEASE
Priority:
Normal
Assignee:
Category:
-
Sprint/Milestone:
Start date:
Due date:
Estimated time:
Severity:
2. Medium
Version:
Platform Release:
OS:
Triaged:
Yes
Groomed:
No
Sprint Candidate:
No
Tags:
Sprint:
Sprint 89
Quarter:

Description

It's been observed that Redis can "loose" a task from it's queue. This is believed to be the situation in https://pulp.plan.io/issues/7907 for example.

If Redis looses the _release_resource task, but the worker stays running, no cleanup mechanism will release the resource locks the worker aquired. In this situation the tasking could backup indefinitely. With the bugfix of https://pulp.plan.io/issues/7907, restarting the worker will release the locks, but it would be better if a restart of processes was not required.

Simulating this experience

  1. Create a remote
  2. Call sync 5 times in a row (to cause them to all be dispatched to the same worker)
  3. Verify they are in the worker's queue with rq info
  4. Tell RQ to forget the tasks without telling Pulp by running rq empty <the_worker's_queue_name>. For example rq empty 47640@pulp3-source-fedora32.localhost.example.com.

Improving this situation

Have the resource manager perform a health check at the two sleep(0.25) statements. One here and the other here.

Those are the exact points where the resource manager cannot dispatch the next task because it's waiting for locks to be released, and if the failure situation is occurring, it will wait there indefinitely. So that's the correct (and efficient) place to check it.

Specifically have the resource manager run a health check that:

for worker in all_the_online_workers:
    for task in all tasks assigned to that worker in incomplete states:
        ask RQ if it has that job ID in the worker's queue still

# also do the same for the resource manager
for task in all tasks that are not yet assigned to a worker:
    ask RQ if the resource manager queue has the job, by it's rq_job_id

If all the jobs are in RQ then just continue to loop around in _queue_reserved_resource. Any of the jobs that are missing, call cancel() on them. If you can't connect to Redis or experience a fatal exception, just let it raise.


Related issues

Related to Pulp - Backport #7950: Backport 7912CLOSED - WONTFIX

Actions
Blocks Pulp - Backport #7951: Backport 7907CLOSED - WONTFIX

Actions
Actions #1

Updated by ttereshc almost 2 years ago

  • Subject changed from The tasking system deadlocks when Redis looses it's tasks to The tasking system deadlocks when Redis looses its tasks
Actions #2

Updated by fao89 almost 2 years ago

  • Triaged changed from No to Yes
  • Sprint set to Sprint 87
Actions #3

Updated by dalley almost 2 years ago

  • Status changed from NEW to ASSIGNED
  • Assignee set to dalley
Actions #4

Updated by daviddavis almost 2 years ago

Actions #5

Updated by fao89 almost 2 years ago

Actions #6

Updated by pulpbot almost 2 years ago

  • Status changed from ASSIGNED to POST
Actions #8

Updated by rchan almost 2 years ago

  • Sprint changed from Sprint 87 to Sprint 88
Actions #9

Updated by rchan over 1 year ago

  • Sprint changed from Sprint 88 to Sprint 89

Added by dalley over 1 year ago

Revision ca9a98b2

Remove separate task for releasing resources

Move it to an after-task action

re: #7912 https://pulp.plan.io/issues/7912

Added by dalley over 1 year ago

Revision ac83f33c

Avoid deadlocks when Redis is shut down and tasks are lost

closes: #7912 https://pulp.plan.io/issues/7912

Actions #10

Updated by dalley over 1 year ago

  • Status changed from POST to MODIFIED

Added by ttereshc over 1 year ago

Revision eaa61e99

Ensure that a task is created in pulp not earlier than the job in redis

Otherwise it can get cancelled when we check for missing tasks/jobs.

re #7912 https://pulp.plan.io/issues/7912

Actions #12

Updated by ttereshc over 1 year ago

  • Sprint/Milestone set to 3.10.0
Actions #13

Updated by pulpbot over 1 year ago

  • Status changed from MODIFIED to CLOSED - CURRENTRELEASE

Also available in: Atom PDF