Re: [GENERAL] Autovacuum help..

2006-08-07 Thread Sundar Narayanaswamy
I haven't had issues about the max_fsm_pages. The problem I had was caused because I was not committing (or rolling back) after JDBC select calls. I didn't think that was necessary because the select calls didnt modify database. But, once I modified my selects to rollback after the data is read fro

Re: [GENERAL] Autovacuum help..

2006-08-06 Thread Francisco Reyes
Matthew T. O'Connor writes: In many instances the default thresholds for autovacuum are too conservative. You can try making it more aggressive and see what happens, you can do this generally, or on a table specific basis. I think something is very wrong with the estimates I am getting back.

Re: [GENERAL] Autovacuum help..

2006-08-06 Thread Matthew T. O'Connor
Francisco Reyes wrote: Using 8.1.4 Have autovacuum running and it shows on the logs as running. Ever couple of days in the last week when I tried to run vacuum analyze postgresql complained that I needed to increase my fsm_pages. I am going to start scheduling vacuum analyze in crontab.. but i

Re: [GENERAL] Autovacuum help..

2006-08-06 Thread Francisco Reyes
Martijn van Oosterhout writes: My question is, is autovacuum expected to work in situations like this where I have a long lived connection to the database ? After I shutdown my application, I am having a simmilar situation, but I am just starting to debug it. Using 8.1.4 Have autovacuum runni

Re: [GENERAL] Autovacuum help..

2006-08-02 Thread Martijn van Oosterhout
On Wed, Aug 02, 2006 at 07:36:09AM -0700, Sundar Narayanaswamy wrote: > select * from ; > read rows from result set > (autovacuum cannot remove dead rows) > > (autovacuum cannot remove dead rows) > . > . > The last select operation is the one of concern. I was just raising the point > that s

Re: [GENERAL] Autovacuum help..

2006-08-02 Thread Sundar Narayanaswamy
> > > > Thanks again. I am wondering as to why the state changes to "Transaction in > > idle" when a query is executed. It'll be nice if that happens only when > > a real change is made (transaction starts) to the database and not when > > a select query occurs. > > This makes no sense. A select

Re: [GENERAL] Autovacuum help..

2006-08-01 Thread Martijn van Oosterhout
On Tue, Aug 01, 2006 at 11:48:04AM -0700, Sundar Narayanaswamy wrote: > Now, in this situation, if some databases stay in "Idle in transaction", > would the dead rows be unremovable from other databases (that are in "idle" > state) as well ? In other words, should ALL of the databases/connections >

Re: [GENERAL] Autovacuum help..

2006-08-01 Thread Sundar Narayanaswamy
> Well, you need to work out why they are unremovable. For example, do > you have any really long lived open transactions. These are generally a > bad idea, for all sorts of reasons. If you don't commit occasionally, > none of your changes will be visible to anyone else. I have multiple databases

Re: [GENERAL] Autovacuum help..

2006-08-01 Thread John Purser
On Tue, 1 Aug 2006 08:02:59 -0700 (PDT) Sundar Narayanaswamy <[EMAIL PROTECTED]> wrote: > Hi, > > I need your help/suggestions with a problem I am facing related to > autovacuum. > > I am using PostgreSQL 8.1.2 through a JDBC connection. The connection > is long lived (established when the appli

Re: [GENERAL] Autovacuum help..

2006-08-01 Thread Martijn van Oosterhout
On Tue, Aug 01, 2006 at 08:02:59AM -0700, Sundar Narayanaswamy wrote: > I need your help/suggestions with a problem I am facing related to autovacuum. > > I am using PostgreSQL 8.1.2 through a JDBC connection. The connection is > long lived (established when the application starts up and is close

[GENERAL] Autovacuum help..

2006-08-01 Thread Sundar Narayanaswamy
Hi, I need your help/suggestions with a problem I am facing related to autovacuum. I am using PostgreSQL 8.1.2 through a JDBC connection. The connection is long lived (established when the application starts up and is closed only when the application is shutdown). I have enabled the autovacuu