I noticed that our hosted-engine suites  often fail recently, and
decided to have a look at , which are on 4.2, which should
hopefully be "rock solid" and basically never fail.
I looked at these, , which are all the ones that still
appear in  and marked as failed.
- All but one failed while "Waiting for agent to be ready" and timing
out after 10 minutes, as part of 008_restart_he_vm.py, which was added
a month ago  and then patched .
- The other one  failed while "Waiting for engine to migrate", also
eventually timing out after 10 minutes, as part of
010_local_mainentance.py, which was also added in .
I also had a look at the last ones that succeeded, builds 329 to 337
of . There:
- "Waiting for agent to be ready" took between 26 and 48 seconds
- "Waiting for engine to migrate" took between 69 and 82 seconds
Assuming these numbers are reasonable (which might be debatable), 10
minutes indeed sounds like a reasonable timeout, and I think we should
handle each failure specifically. Did anyone check them? Was it an
infra issue/load/etc.? A bug? Something else?
Suites should be monitored by their respectful maintainers, the infra team doesn't have the capasity nor resources
to monitor any new suite that is running in CI.
Having said that, if a certain infra issue is reported, either its Lago, OST or infra issue, we'll of course do our best to find and fix the issue.
I didn't check the logs yet, might do this later. Also didn't check
the failures in other jobs in .