Did that really fucken unjam it
Timeline
Post
Remote status
Context
3
@pwm lol what happned? I was wondering why you were posting from dsmc
@bajax I spent a week (because work in between) moving from my old cluster to the new host for it, and after I loaded up the database dump, the objects table needed reindexing or something, because you literally couldn't query it within a 15 second timeout, so federation was stuck will I made pleroma prune it and I think that triggered a reindex or something idk exactly.
The old triple master cluster was on worse hardware and I made bad architectural decisions by using a networked disk soltuion (was being too clever)
now it's still in a kuberenetes thing but better hardware choices.
The old triple master cluster was on worse hardware and I made bad architectural decisions by using a networked disk soltuion (was being too clever)
now it's still in a kuberenetes thing but better hardware choices.
@pwm @bajax
>moving from my old cluster to the new host for it, and after I loaded up the database dump, the objects table needed reindexing or something
Next time after a restore:
>moving from my old cluster to the new host for it, and after I loaded up the database dump, the objects table needed reindexing or something
Next time after a restore:
vacuumdb --all --analyze-in-stages --verboseWithout statistics the query planner makes completely unoptimized decisions. What fixed it was actually autovacuum (vacuum and analyze) getting triggered on the table after enough changes were done to the table by the prune.
Replies
2
@bajax @pwm I blame 
for never mentioning it in his dump/restore gitlab snippet that is widely used: https://gitlab.com/-/snippets/2228488
It's actually mentioned in the Pleroma docs, but the currently live docs have a different problem where the restoration takes an eternity (it's already fixed for a long time, but not deployed yet...)
for never mentioning it in his dump/restore gitlab snippet that is widely used: https://gitlab.com/-/snippets/2228488It's actually mentioned in the Pleroma docs, but the currently live docs have a different problem where the restoration takes an eternity (it's already fixed for a long time, but not deployed yet...)