Hi there,

I am using Slony 1.2.15 on gentoo with postgres 8.3.4
Context is a java web application running on tomcat.

I encountered some weird behaviour when doing multiple select statements with a plain jdbc connection triggered by a script on the master db.

It seems that the select statements are slowing down the slony replication in a major way, so that slony commits wont be processed
until the select statement finishes.
The connection pool gets exhausted by queries that do not return and stay at the COMMIT statement. Selecting pg_stat_activity shows up to 42 "hanging" COMMIT statements!

In pg_logs I can see lots of waiting queries:

duration: 5557.956 ms  statement: select "_km_cluster".forwardConfirm(2, 3,'1814465', '2009-06-26 15:55:02.034346');
15:55:09   km_prod LOG:  duration: 5219.536 ms  statement: notify "_km_cluster_Event"; insert into "_km_cluster".sl_event     
	(ev_origin, ev_seqno, ev_timestamp,      ev_minxid, ev_maxxid, ev_xip, ev_type     ) values ('2','1814466', '2009-06-26 15:55:03.98378', '75848912', '75848912', '', 'SYNC'); 
	insert into "_km_cluster".sl_confirm ^I(con_origin,con_received, con_seqno, con_timestamp)    values (2, 1, '1814466', now()); commit transaction;
15:55:10 comm_prod LOG:  duration: 17750.271 ms  statement: listen "_comm_cluster_Event";
15:55:10 comm_prod LOG:  duration: 17694.024 ms  statement: listen "_comm_cluster_Event";
15:55:11 comm_prod LOG:  duration: 16337.644 ms  statement: notify "_comm_cluster_Event"; insert
15:55:11 km_prod LOG:  duration: 13147.354 ms  execute S_2: COMMIT into "_comm_cluster".sl_event     (ev_origin, ev_seqno, ev_timestamp,      ev_minxid, ev_maxxid, ev_xip, ev_type     ) values
	('3', '1213250', '2009-06-26 15:54:53.709398', '50637491', '50637491', '', 'SYNC'); insert into "_comm_cluster".sl_confirm
	^I(con_origin, con_received, con_seqno, con_timestamp)    values (3, 1, '1213250', now()); commit transaction;
15:55:12   comm_prod LOG:  duration: 24362.984 ms  statement: notify "_comm_cluster_Event"; insert into "_comm_cluster".sl_event     (ev_origin, ev_seqno, ev_timestamp,      ev_minxid, ev_maxxid, ev_xip, ev_type     ) values
	('2', '1815023', '2009-06-26 15:54:47.1518', '75848890', '75848890', '', 'SYNC'); insert into "_comm_cluster".sl_confirm ^I(con_origin, con_received, con_seqno, con_timestamp)    values (2, 1, '1815023', now()); commit transaction;
15:55:12   comm_prod LOG:  duration: 5728.775 ms  statement: select "_comm_cluster".createEvent('_comm_cluster', 'SYNC', NULL);
15:55:13 km_prod LOG:  duration: 6187.793 ms  execute S_2: COMMIT
15:55:13 km_prod LOG:  duration: 12125.475 ms  execute S_2: COMMIT
15:55:15 km_prod LOG:  duration: 7538.928 ms  execute S_2: COMMIT
15:55:15 km_prod LOG:  duration: 7532.898 ms  execute S_2: COMMIT
15:55:15 km_prod LOG:  duration: 7531.024 ms  execute S_2: COMMIT
15:55:15 km_prod LOG:  duration: 7763.779 ms  execute S_2: COMMIT
... lots of waiting commits


Any idea or hint what may causes the problem?
I am confused because "Select statements" do not bother Slony , don't they?

Best regards
Burak
-- 
GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01