It's needed to disable the prefeching of the tasks by the
workers. It can lead in situations where a task is prefetched by
a worker loading a big repo. The prefetch task will remain blocked
until the big repo loading is done.
Related to T3857
Details
Details
- Reviewers
ardumont - Group Reviewers
System administrators - Maniphest Tasks
- T3857: unexpected long time between a task is received and the loading start for few tasks
- Commits
- rSPSITE768ce43e60f9: Reactivate the ack late configuration for the high priority tasks
diff origin/production/worker0.internal.staging.swh.network current/worker0.internal.staging.swh.network
*******************************************
File[/etc/softwareheritage/loader_high_priority.yml] =>
parameters =>
content =>
@@ -22,4 +22,5 @@
max_content_size: 104857600
celery:
+ task_acks_late: true
task_broker: amqp://swhconsumer:swh-deploy-worker-task_broker-password@scheduler0.internal.staging.swh.network:5672/%2f
task_queues:
*******************************************
*** End octocatalog-diff on worker0.internal.staging.swh.networkDiff Detail
Diff Detail
- Repository
- rSPSITE puppet-swh-site
- Lint
Automatic diff as part of commit; lint not applicable. - Unit
Automatic diff as part of commit; unit tests not applicable.