action #134837
closed
QA (public) - coordination #121720: [saga][epic] Migration to QE setup in PRG2+NUE3 while ensuring availability
QA (public) - coordination #123800: [epic] Provide SUSE QE Tools services running in PRG2 aka. Prg CoLo
SLE test repo not updated on OSD, cron service was not running since 2023-08-29, fetchneedles not called size:M
Added by okurz over 1 year ago.
Updated over 1 year ago.
Category:
Regressions/Crashes
Description
Observation¶
From https://suse.slack.com/archives/C02CANHLANP/p1693394887125729 Paolo Stivanin made us aware that the test repository checkout for os-autoinst-distri-opensuse on OSD was not up-to-date. I found that the repository checkout had commits from 2023-08-29 but not the expected ones for 2023-08-30. journalctl -e -u cron showed as last entries
Aug 29 08:05:01 openqa CRON[11204]: (root) CMD (touch /var/lib/openqa/factory/repo/cvd/*)
Aug 29 08:05:02 openqa CRON[11199]: (root) CMDEND (touch /var/lib/openqa/factory/repo/cvd/*)
Aug 29 09:05:01 openqa CRON[19801]: (root) CMD (touch /var/lib/openqa/factory/repo/cvd/*)
Aug 29 10:05:01 openqa CRON[24320]: (root) CMD (touch /var/lib/openqa/factory/repo/cvd/*)
Aug 29 10:05:02 openqa CRON[24317]: (root) CMDEND (touch /var/lib/openqa/factory/repo/cvd/*)
and the service
# systemctl status cron
. cron.service - Command Scheduler
Loaded: loaded (/usr/lib/systemd/system/cron.service; enabled; vendor preset: enabled)
Active: inactive (dead)
I called fetchneedles manually and started the cron service and it was fine again. But we did not receive any alert and no other notice that the service was enabled but not running. Can we find a way to be alerted about systemd services enabled but not running?
Acceptance Criteria¶
- AC1: We are alerted about enabled services not running (or active?)
Suggestions¶
- Research about how to find enabled systemd services not running and how to monitor for that
- We already have a check triggered by telegraf for failed services as well as masked services. Look into that and possible extend or be inspired by those solutions
- Ensure to have monitoring panels and alerts
Workaround¶
Restart the service manually
- Copied from action #132146: Support migration of osd VM to PRG2 - 2023-08-29 size:M added
I called fetchneedles manually and started the cron service and it was fine again. But we did not receive any alert and no other notice that the service was enabled but not running. Can we find a way to be alerted about systemd services enabled but not running?
Hrm. I was thinking of #134519 at first where the issue is that we're not seeing alerts for individual cron jobs. If cron was not running that should have triggered a systemd services alert, though?
telegraf-webui.conf mentions cron, however there is no monitoring/grafana/webui.services.json to sync with and no other instance of cron in salt-states-openqa (or pillars)? Was there some change that broke this inadvertendly?
- Status changed from New to In Progress
- Assignee set to livdywan
I'm taking a look. Maybe we can make this an if-or by adjusting our script that checks the service state
- Related to action #134519: We were not notified that backup.qa.suse.de did not create backups size:M added
livdywan wrote in #note-2:
I called fetchneedles manually and started the cron service and it was fine again. But we did not receive any alert and no other notice that the service was enabled but not running. Can we find a way to be alerted about systemd services enabled but not running?
Hrm. I was thinking of #134519 at first where the issue is that we're not seeing alerts for individual cron jobs
Well, it might avoid the issue by using systemd timers
If cron was not running that should have triggered a systemd services alert, though?
That would be good but so far we are only looking for failed alerts. I researched yesterday shortly with mkittler+nicksinger how to find enabled systemd services that should be running but aren't but the research was so far inconclusive.
telegraf-webui.conf mentions cron, however there is no monitoring/grafana/webui.services.json to sync with and no other instance of cron in salt-states-openqa (or pillars)?
https://gitlab.suse.de/openqa/salt-states-openqa/-/blob/master/etc/master/cron.d/SLES.CRON
Was there some change that broke this inadvertendly?
I suspect it was a mistake by us during OSD migration that triggered this and is unlikely to happen again.
- Due date set to 2023-09-14
Setting due date based on mean cycle time of SUSE QE Tools
- Subject changed from SLE test repo not updated on OSD, cron service was not running since 2023-08-29, fetchneedles not called to SLE test repo not updated on OSD, cron service was not running since 2023-08-29, fetchneedles not called size:M
- Description updated (diff)
- Due date deleted (
2023-09-14)
- Status changed from In Progress to Resolved
I also brought this up in jitsi. Ideas considered were having regularly run pipelines with salt state apply similar to what we're doing with package upgrades since we might not apply and e.g. enable services for several days. We probably don't want to interfere with temporary manual changes, though - so for now I'm hesitant to go this route. A bit of online research suggests it's common to simply restart services via cron or vice versa, but we may not want to increase interdependency here.
For now I'd probably say we can consider if we want the above simple MR, but otherwise leave this as-is.
Also available in: Atom
PDF