Each time there is an update or insert in the storage back-end, we fire off background-update jobs for revision policy enforcement and secondary-indices updates. However, these asynchronous jobs are executed in parallel with the normal request execution pipeline, which means they are left unmonitored and unprotected. We need to make sure they are bullet-proof and that their execution cannot get interrupted by unhandled (and possibly synchronously-thrown) exceptions as that might leave the process hanging.
Description
Description
Details
Details
Subject | Repo | Branch | Lines +/- | |
---|---|---|---|---|
Update restbase to 03ed384 | mediawiki/services/restbase/deploy | master | +187 -124 |
Related Objects
Related Objects
Event Timeline
Comment Actions
Since it is friday and the background update jobs keep leaving RESTBase processes in a comatose state, I am going to redeploy the last-deployed version of RESTBase, and only rebuild the module dependencies, which will pick up the new version.
Comment Actions
This has been deployed @ 12:43 UTC, the system seems stable and functional. So far so good - no unhandled exception errors in local syslogs.
Comment Actions
Change 223297 had a related patch set uploaded (by Mobrovac):
Update restbase to 03ed384