Thu Oct 5 11:17:13 PDT 2006
- Previous message: [Slony1-general] "Fetch 100 from LOG" too slow, Slony way behind
- Next message: [Slony1-general] "Fetch 100 from LOG" too slow, Slony way behind
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Thu, 2006-10-05 at 10:48 -0700, Hanks, Dan wrote: > Hi folks, > > > > I?m running Slony-1.1.5, and am seeing the dreaded ?fetch 100 from > LOG? show up in my slow query logs, and Slony is, predictably, falling > behind. The db is quite busy, but I don?t think overloaded. > > > > We have identified some of our code that induces ?IDLE in transaction? > sessions that sit and do nothing. We?re working to fix those. That's what your problem is problem. > > > Is there anything else we can be doing to help it along? sl_log_1 has > 15M rows in it currently. Are we at the point of being hopelessly > behind? > Kill all the 'idle in transaction' transactions, do a VACUUM ANALYZE on pg_listener, and probably a VACUUM FULL as well. This takes an exclusive lock on pg_listener, but that's okay, it won't block anything else. The slons are already doing this anyway. You'll probably also want to do a VACUUM ANALYZE on sl_log_1, as well as any other tables in the database that are frequently updated. The vacuums haven't been able to do any work since the start of the oldest 'idle in transaction' transaction. You still may have problems catch up, even after all of this. Brad.
- Previous message: [Slony1-general] "Fetch 100 from LOG" too slow, Slony way behind
- Next message: [Slony1-general] "Fetch 100 from LOG" too slow, Slony way behind
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Slony1-general mailing list