Closed Bug 1873752 Opened 2 years ago Closed 2 years ago

[traceback] TimeoutExpired: remove_orphaned_files

Categories

(Tecken :: General, defect, P2)

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: willkg, Assigned: relud)

Details

Attachments

(3 files)

Sentry: https://mozilla.sentry.io/issues/4723701783/events/642158414c5b4b79a185e3e6e1f2e574/

TimeoutExpired: Command '['python', '/app/manage.py', 'remove_orphaned_files', '--skip-checks']'
timed out after 60 seconds

Looks like the remove_orphaned_files takes longer than 60 seconds semi-regularly. As of now, there are 92 events of this in the last 30 days.

Before making any changes, we should add a timing metric to learn more about how long it takes to run and how often it hits the 60 second timeout. Then we should adjust the timeout value.

Assignee: nobody → dthorn
Status: NEW → ASSIGNED

This was deployed just now in bug #1880345 in v2024.02.14.

I verified that it's emitting data and we can see it in the "Time to remove orphaned files" panel.

Flags: needinfo?(dthorn)

looking at when this occurs, it appears to have mostly stopped for a while after the january 9th deploy, then it picked up again on feb 1st, and stopped again with the feb 14th deploy. So I'm going to give this a month to start happening again before taking any further action.

Flags: needinfo?(dthorn)

I tweaked that graph so it knows the values are in milliseconds and thus gives the y-axis easier to think about labels. Does that change look right?

that looks great! tyty

there have been two timeouts in the last month, so we discussed it in the obs-team meeting and we're going to double the timeout, revisit in a month, and if it didn't work double the timeout again.

relud: Did the timeout get doubled? Another timeout error occured 2024-03-25, and it's still listing 60s as the timeout.

Edit: Interestingly, I don't see the 3 Sentry events that have occured in the last 30 days in the Grafana panel that exceeded the timeout. Everything looks to be below 55s? In the table for the graph data, the max is 53.5s.

Flags: needinfo?(dthorn)

i have not updated the timeout, thank you for bringing it up, I will get on that right away.

grafana not showing timeouts: correct! if i understand correctly the timeout forcibly kills the process, so the metric cannot be emitted.

Flags: needinfo?(dthorn)

Ah okay that makes sense thanks. Cool; hopefully we stop seeing those errors with the increase. It will be interesting to see how much time over 60s these runs take.

i archived the sentry issue "until escalating" so that we will notice that way if this issue needs to be revisited

Timeout doubling was deployed to production just now in bug #1890798.

resolving this until it reappears

Status: ASSIGNED → RESOLVED
Closed: 2 years ago
Resolution: --- → FIXED

it reappeared, doubling again

Status: RESOLVED → REOPENED
Resolution: FIXED → ---

marked the issue in sentry as "resolved in next release" now that we have release info

Status: REOPENED → RESOLVED
Closed: 2 years ago2 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: