problem

When using globalese pretranslation the responsible worker on our side remains running until globalese is ready. This can be multiple hours (today seen a worker running for 4 hours). This blocks apache resources and is error-prone since if the apache process goes away the whole import is defunct (if a long running import worker crashes this is also bad, but then there is no way to recover. If the globalese worker crashes on waiting for globalese this can be recovered.

solution

The solution would be that the checking worker is called periodically instead of running in long loop.

For periodically workers we need:

• A cron job triggered run workers call on the queue, since we have now cron 15minutes functionality (GarbageCollector) this would be easily possible
• The worker may not run immediatelly, but only if a given "timestamp" is reached. Probably with a start date in the future?

For globalse the check worker can run, check the globalese side, download the translation if finished. If not add another check worker not allowed to run not before a specific date (5min + now) is reached.