Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-20 Thread Dave Owens
produced a patch that demotes some heavy-hitting queries down to Read Committed, and we will see if this makes an impact on the number of SIReadLocks. Is it interesting that only 101557 out of 7 million SIReadLocks have a pid associated with them? -Dave Owens -- Sent via pgsql-performance mailing

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-19 Thread Dave Owens
On Tue, Aug 19, 2014 at 11:01 AM, Kevin Grittner wrote: > If restart is an option, that sounds like a great idea. If you > could capture the data into tables where we can summarize to > analyze it in a meaningful way, that would be ideal. Something > like: > > CREATE TABLE activity_snap_1 AS SEL

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-19 Thread Dave Owens
I wonder if it would be helpful to restart the database, then begin gathering information pg_locks while it can still respond to queries. I speculate that this is possible because the amount of memory needed to query pg_locks continues to grow (around 1900MB now). Dave Owens -- Sent via pgsql

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-19 Thread Dave Owens
Hi Andres, On Tue, Aug 19, 2014 at 10:17 AM, Andres Freund wrote: >> max_connections = 450 ...we have found that we run out of shared >> memory when max_pred_locks_per_transaction is less than 30k. > > What was the precise error message when that happened? 2014-07-31 15:00:25 PDT 53dabbea.29c7ER

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-19 Thread Dave Owens
On Tue, Aug 19, 2014 at 9:40 AM, Kevin Grittner wrote: > Hmm, that's not outrageous. How about long-running transactions? > Please check pg_stat_activity and pg_prepared_xacts for xact_start > or prepared (respectively) values older than a few minutes. Since > predicate locks may need to be kept

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-19 Thread Dave Owens
Relic reports about 30k calls per minute, from our main webapp). That database alone consists of 575 tables and 732 indexes. Dave Owens -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql

Re: [PERFORM] query against pg_locks leads to large memory alloc

2014-08-18 Thread Dave Owens
On Mon, Aug 18, 2014 at 2:21 PM, Matheus de Oliveira wrote: > Do you really need such large values? What is your max_connections value? max_connections = 450 ...we have found that we run out of shared memory when max_pred_locks_per_transaction is less than 30k. On Mon, Aug 18, 2014 at 2:29 PM, M

[PERFORM] query against pg_locks leads to large memory alloc

2014-08-18 Thread Dave Owens
Huge Pages are disabled :-) Thanks in advance for your time and expertise. Dave Owens -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance

Re: [PERFORM] CPU spikes and transactions

2014-05-13 Thread Dave Owens
the report differently to be useful. We will begin reducing shared_buffers incrementally over the coming days. Dave Owens 541-359-2602 TU Facebook<https://app.getsignals.com/link?url=https%3A%2F%2Fwww.facebook.com%2Fteamunify&ukey=agxzfnNpZ25hbHNjcnhyGAsSC1VzZXJQcm9maWxlGICAgOCP-IMLDA&k=17