Fast Shutdown (SIGINT) results in a CHECKPOINT...
For a while now I've known that sending a SIGINT can trigger a CHECKPOINT, but I don't know if this is intentional or a bug. Logs from today:
2013-03-20_18:21:43.27642 LOG: received fast shutdown request
2013-03-20_18:21:43.27645 LOG: aborting any active transactions
2013-03-20_18:21:43.27647 FATAL: terminating connection due to administrator command
2013-03-20_18:22:14.25763 FATAL: the database system is shutting down
2013-03-20_18:22:14.26401 LOG: checkpoint starting: shutdown immediate
2013-03-20_18:25:36.98507 LOG: checkpoint complete: wrote 6403 buffers (2.3%); 0 transaction log file(s) added, 0 removed, 5 recycled; write=41.450 s, sync=156.157 s, total=202.943 s; sync files=50, longest=29.057 s, average=3.123 s
2013-03-20_18:25:37.65570 FATAL: the database system is shutting down
2013-03-20_18:25:37.71735 LOG: online backup mode canceled
And I believe this is where the db restarted:
2013-03-20_18:25:37.71739 DETAIL: "backup_label" was renamed to "backup_label.old".
2013-03-20_18:25:45.39001 LOG: loaded library "auto_explain"
2013-03-20_18:25:45.39647 LOG: loaded library "pg_stat_statements"
2013-03-20_18:26:04.67932 LOG: could not create socket for statistics collector: Protocol not supported
2013-03-20_18:26:04.67939 LOG: trying another address for the statistics collector
2013-03-20_18:26:05.01894 LOG: connection received: host=172.16.4.24 port=55414
2013-03-20_18:26:05.01903 FATAL: the database system is starting up
2013-03-20_18:26:05.41153 LOG: autovacuum launcher started
2013-03-20_18:26:05.41160 LOG: database system is ready to accept connections
Just shy of 5 minutes to do a fast shutdown/restart.
What I'm trying to determine is whether or not it's a bug for postmaster to CHECKPOINT during a fast shutdown or a documentation bug that a fast shutdown may cause a CHECKPOINT, which will delay the database being restarted.
I've run in to this before and in some scripts I run a CHECKPOINT before sending a SIGINT, which reduces the size of the CHECKPOINT, but doesn't necessarily eliminate it all of the time. In fact, I'm not sure it happens 100% of the time either, but today I was bit by this shutdown/startup delay and thought I'd inquire or submit a small doc patch. Thanks in advance. -sc
--
Sean Chittenden
sean@chittenden.org
--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
On 3/20/13 4:28 PM, Sean Chittenden wrote:
For a while now I've known that sending a SIGINT can trigger a CHECKPOINT, but I don't know if this is intentional or a bug.
It's intentional. If you don't want that, use SIGQUIT. That's how they
are different.
Of course, when using SIGQUIT, you will have to spend the time you saved
on the checkpoint for the recovery at startup. So you have to put in
the time either way.
It is not unheard of that a shutdown can take minutes. That's why the
-t option was added to pg_ctl some time ago.
--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general