Process-death-row procedure became very slow
Bug #193656 reported by
Celso Providelo
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Launchpad itself |
Fix Released
|
Critical
|
Celso Providelo |
Bug Description
Lately, after we started to remove obsolete distroseries from the archive, the process-death-row performance was really affected, making it impossible to run both, publish-distro & process-death-row synchronously in less than 1 hour.
Changed in soyuz: | |
assignee: | nobody → cprov |
importance: | Undecided → Critical |
milestone: | none → 1.2.3 |
status: | New → Triaged |
Changed in soyuz: | |
status: | Triaged → In Progress |
Changed in soyuz: | |
status: | Fix Committed → Fix Released |
To post a comment you must log in.
We have done some experiences to run process-death-row (p-d-r) in parallel with publish-distro (p-d), but it only works correctly if p-d-r starts after the publishing table are already changes, i.e. when p-d start to run apt-ftparchive (a-f) which is approximately 7 minutes after it started.
This very specific condition is not working well, specially when archive admins have to run the publisher cycle manually, if the publishing tables change while p-d-r is running it ends up not able to commit the transaction and blocks the rest of the system consuming a lot of CPU.