The buildfarm has shown several intermittent failures recently:
http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=heron&dt=2008-04-20%2000:06:01
http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=cobra&dt=2008-04-26%2004:15:02
http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=heron&dt=2008-04-27%2000:06:01
all with the same signature:
*** ./expected/prepared_xacts.out Sun Apr 27 02:06:33 2008
--- ./results/prepared_xacts.out Sun Apr 27 02:10:19 2008
***************
*** 161,166 ****
--- 161,167 ---- SELECT * FROM pxtest3; ERROR: canceling statement due to statement timeout reset statement_timeout;
+ ERROR: canceling statement due to statement timeout -- Disconnect, we will continue testing in a different backend
\c- -- There should still be two prepared transactions
which so far as I recall is not a failure we've seen in the past
(though we still don't have buildfarm infrastructure that makes it
easy to confirm or refute that opinion).
One explanation for this is that these buildfarm machines were so
heavily overloaded that it really did take more than 1 sec to execute
"reset statement_timeout". That's not very plausible, though,
especially not considering that both machines logged session times
(via log_disconnect) that don't seem very far out of line.
Another theory is that we broke something recently, but I see no obvious
candidates in the CVS logs --- and I've spent the evening running 488
cycles of the parallel regression tests here, with no error.
Anyone have a theory?
regards, tom lane
