rosetta-pofile-stats script takes more than a week to complete

Hi Henning.

I think this relates to jobs that build data that is effectively cached until the next run.

AFAIK the statistics are updated weekly. But that job id quite resource consuming and may fail. As long as they have not been updated, the old statistics are displayed.
Whenever new translations are accepted, though, the statistics for that particular POFile are recalculated on the spot. If that is not happening, it's a bug.

Is this really a regression? That is, do we know it was reliable before?

The review counters were reliable in the past.

After adding the shared translation feature for the same PO Template there were some delays ... but not longer than 1 week.

The translations mentioned in this bug are out of sync for more than 2 weeks.

At least for the Romanian team, this bug is very important since the reviewers are turned down by the false positives and are no longer checking the translations status.

Cheers,

FWIW, the script seems to take roughly 10 days to run now. We should:
 1. investigate fixing pofile-stats-daily and using it exclusively
 2. fix the problems with the stats not being updated directly (even though that is attempted) and drop the regular runs of the script (or simply make them log bugs when they find discrepancies)

Solution to 2 might be to have an async job updating the stats that would take care of all the sharing pofiles as well.

As part of fastdowntime database deployments, we are expecting short database outages at least once per week. This means the script will no longer complete.

According to the database reports, the script is chewing up 72% of a core. It is active almost all the time rather than sitting idle blocked by locks etc.

If its doing incremental locks its not completing isn't a fastdowntime issue per se - it could be interrupted for several different reasons.

This would be "solved" by removing the need for this script in the first place: bug 877195.