PostgreSQL, idle in transaction and vacuum not vacuuming
PostgreSQL, idle in transaction and vacuum not vacuuming
- Subject: PostgreSQL, idle in transaction and vacuum not vacuuming
- From: Miguel Arroz <email@hidden>
- Date: Mon, 29 Jun 2009 00:30:06 +0100
Hi!
One of our PostgreSQL databases is growing a lot. This database
contains tables with rows that are updated very often, and if we dump
the SQL and re-import it, the DB takes up much less space than it did
before (less than one third).
After some research, I found out vacuum says there are almost one
hundred thousand dead rows in a particular table that cannot be reused
yet.
I'm trying to find out why, and some pages I read state that it may
be caused my pending transactions. My FSM values seem to be good, they
are much higher than what VACUUM VERBOSE claims to need.
When I do a select * from pg_stat_activity I see that I have two
connections per app per DB, one usually "<IDLE>" and the other "<IDLE>
in transaction". The "<IDLE> in transaction" one has a xact_start from
some days ago, which may be the culprit of this. I'm not 100% sure,
because apparently the "fatty" table is not locked by any of those
transactions, but still, it's not a good sign.
I was told this was related to jdbc2info, and it's an old EOF bug.
So, some questions:
1) This must be fixed in Wonder somewhere. :) Where and how?
2) Do you guys have the same problem with PostgreSQL?
3) If this is not the cause to my problem, any idea of what it
might be?
Yours
Miguel Arroz
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden