Hello.
Thanks for the response.
There are several 'idle in transaction' on this server/app, but to a different db/schema.
The "repository" (JCR) schema has only a few 'idle', none 'in transaction' .
By "routine maintenance", do you mean autovacuum, or something else?
Autovacuum does appear to usually get 'auto-canceled' by a lock. However, even when it runs successfully, it doesn't seem to help with this ws_bundle Toast table size.
I am rather looking for a root cause here. Surely this table is not supposed to grow so much (100s of GB). It is even bigger than the data store!
I agree that this may not be an 'error' in Postgres, but somehow it is not playing well with Jackrabbit clustering.
On Mon, Jul 23, 2012 at 5:40 PM, Joshua D. Drake
<jd@commandprompt.com> wrote:
On 07/23/2012 02:13 PM, Gary Webster wrote:
Hello. I'm hoping someone has seen this before.
We are trying to use Postgres Plus v9.1.3 as the Persistence Manager in
Jackrabbit (Apache JCR) clustering
(http://wiki.apache.org/jackrabbit/Clustering).
Whenever the JCR is under load, the ws_bundle TOAST table in the
repository schema, grows out of control !
Some of my team members maintain that this problem doesn't occur with
MySQL, but I would rather stay with Postgres if possible...
Under load you can check your process list to see if you have long running transactions that are idle ( idle in transaction ). If you do, you have a code problem not a postgres problem and it is presenting itself through bloat.
Note: IDLE is fine. It is specifically IDLE IN TRANSACTION that is a problem.
Sincerely,
Joshua D. Drake
Thanks.
--
Command Prompt, Inc. - http://www.commandprompt.com/
PostgreSQL Support, Training, Professional Services and Development
High Availability, Oracle Conversion, Postgres-XC
@cmdpromptinc - 509-416-6579