-
Hi there, We used one Etherpad database for 20 web portals and split it into 20 separate databases. After I successfully imported all the separate databases, the load increased dramatically and some large portal databases (~28 GB) timed out. Some relevant postgresql settings(configured with PGTune):
As I describe in Discussion: ueberdb_insert_or_update always idle #5489 the postgres activities always show me: If I execute
Some other specs:
Does anyone have an idea how I can cope with the large databases and reduce the load on the server? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
If possible, try updating Etherpad. Various timeout issues should be handled better in recent versions and it should default to pooling now when using postgres. Does the timeout happen during startup? Do I read #5489 (comment) correctly and you only use a single connection? You might try out connection pooling with postgrespool database backend. You might also try increasing the buffer size in The errors you posted are not related to database problems, but iirc reconnecting and should be fixed in recent versions, too. |
Beta Was this translation helpful? Give feedback.
-
The creation of an index is the solution. I have no resource problems anymore. Without update and increased cache. |
Beta Was this translation helpful? Give feedback.
The creation of an index is the solution.
CREATE INDEX idx ON etherpad_XXXX.store(key);
I have no resource problems anymore. Without update and increased cache.