Bug #15906

Frequent memory issues on translate.lizard

Added by intrigeri 2018-09-03 07:35:41 . Updated 2019-06-27 17:16:37 .

Status:
Resolved
Priority:
Normal
Assignee:
Category:
Infrastructure
Target version:
Start date:
2018-09-03
Due date:
% Done:

40%

Feature Branch:
Type of work:
Sysadmin
Blueprint:

Starter:
Affected tool:
Deliverable for:
309

Description

Since months icinga2 tells us that translate.lizard regularly goes to “memory is CRITICAL” or “memory is WARNING” state. Please check if that’s OK and if yes, tune the memory limits; if not, give the VM more RAM. Goal: avoid false positives in our monitoring so we don’t get used to ignoring alerts. Thanks in advance!


Subtasks


Related issues

Has duplicate Tails - Bug #15928: translate.wizard needs more memory Duplicate 2018-09-10

History

#1 Updated by intrigeri 2018-09-03 07:35:49

  • Tracker changed from Feature to Bug

#2 Updated by intrigeri 2018-09-10 10:13:53

  • has duplicate Bug #15928: translate.wizard needs more memory added

#3 Updated by intrigeri 2018-09-10 10:14:31

See Bug #15928 for info about practical consequences.

#4 Updated by bertagaz 2018-09-11 16:13:47

  • Assignee changed from groente to bertagaz

Being on sysadmin duty I had a look (also because I was annoyed too).

Two cronjobs for weblate were not correctly set:

# Puppet Name: weblate updatechecks
* 6 * * * cd /usr/local/share/weblate; ./manage.py updatechecks --all
# Puppet Name: weblate cleanup
* 4 * * * cd /usr/local/share/weblate; ./manage.py cleanuptrans

Indeed, with such lines, this scripts were run every minutes at that hours. That could explain why every morning the system was lacking memory and OOM killer came into play. I’ve fixed the related lines in the puppet manifest and deployed it. We’ll see in the next days if it fixes the issue.

#5 Updated by intrigeri 2018-09-11 16:37:49

> Indeed, with such lines, this scripts were run every minutes at that hours.

Wow, good catch!

#6 Updated by bertagaz 2018-09-13 15:25:37

  • % Done changed from 0 to 40

Two days later, no other notifications about memory issues on translate.li. I’ll lower its memory later to get it at a reasonable amount.

#7 Updated by bertagaz 2018-09-21 12:43:30

  • Assignee changed from bertagaz to groente
  • QA Check set to Info Needed

bertagaz wrote:
> Two days later, no other notifications about memory issues on translate.li. I’ll lower its memory later to get it at a reasonable amount.

So I’ve lowered it to 1248576KiB, but then we had notifications about memory issues. I’ve raised it to 1748576KB, and then no new notifications poped up.

One thing I noticed though, is that each time puppet is run in this VM, it runs pip install for every configured python packages. When a weblate cronjob (which can be quite expensive memory wise) is ran at the same time, the memory thus gets sparse.

I wonder if we shouldn’t rather use puppet’s vcsrepo module for this python packages, and have a puppet exec ‘pip install’ snippet subscribed to them. This way, puppet would run the pip install command only when the python packages’ git are upgraded. What do you think?

#8 Updated by groente 2018-09-27 13:44:22

  • Status changed from Confirmed to Resolved

hey, first of all, thanks a lot for fixing the cronjob fuckup! :)

> So I’ve lowered it to 1248576KiB, but then we had notifications about memory issues. I’ve raised it to 1748576KB, and then no new notifications poped up.

great, i think with that we can close this ticket.

> One thing I noticed though, is that each time puppet is run in this VM, it runs pip install for every configured python packages. When a weblate cronjob (which can be quite expensive memory wise) is ran at the same time, the memory thus gets sparse.
>
> I wonder if we shouldn’t rather use puppet’s vcsrepo module for this python packages, and have a puppet exec ‘pip install’ snippet subscribed to them. This way, puppet would run the pip install command only when the python packages’ git are upgraded. What do you think?

This whole pip dance will change again with the next weblate upgrade. I’ll pass your suggestion on, but for now I don’t think the extra overhead really hurts, so let’s just leave it as is.

#9 Updated by intrigeri 2018-10-01 13:29:37

  • Target version changed from Tails_3.10.1 to Tails_3.9.1

#10 Updated by Anonymous 2019-02-07 15:35:55

#11 Updated by intrigeri 2019-06-27 17:16:37

  • Assignee deleted (groente)