Обсуждение: Warnings in Standby Logs
Hi,
I'm on 9.1.2 and occasionally get what are probably harmless warning messages in my logs, but because I have the responsibility of ensuring we haven't lost a bit of data between our master and standby, I feel I should know exactly what they mean. I occasionally see "invalid record length" and "record with zero length" in my logs, and they almost only occur during reboots (the "consistent recovery state reached" message gives me the warm-fuzzy, but then when they're followed by either of those two, it leaves me wondering). For example, I had to reboot my hot standby twice this month:
2012-03-16 10:23:28 CDT LOG: received smart shutdown request
2012-03-16 10:23:28 CDT FATAL: terminating walreceiver process due to administrator command
2012-03-16 10:27:33 CDT LOG: database system was interrupted while in recovery at log time 2012-03-16 10:14:15 CDT
2012-03-16 10:27:33 CDT HINT: If this has occurred more than once some data might be corrupted and you might need to choose an earlier recovery target.
2012-03-16 10:27:33 CDT LOG: entering standby mode
2012-03-16 10:27:34 CDT LOG: redo starts at 9/B98D6968
2012-03-16 10:27:34 CDT LOG: consistent recovery state reached at 9/B9F171B0
2012-03-16 10:27:34 CDT LOG: database system is ready to accept read only connections
2012-03-16 10:27:34 CDT LOG: invalid record length at 9/B9F85E78
2012-03-16 10:27:34 CDT LOG: streaming replication successfully connected to primary
2012-03-16 10:28:07 CDT LOG: received smart shutdown request
2012-03-16 10:28:07 CDT FATAL: terminating walreceiver process due to administrator command
2012-03-16 10:28:07 CDT LOG: shutting down
2012-03-16 10:47:00 CDT LOG: database system was interrupted while in recovery at log time 2012-03-16 10:14:15 CDT
2012-03-16 10:47:00 CDT HINT: If this has occurred more than once some data might be corrupted and you might need to choose an earlier recovery target.
2012-03-16 10:47:00 CDT LOG: entering standby mode
2012-03-16 10:47:01 CDT LOG: redo starts at 9/B98D6968
2012-03-16 10:47:02 CDT LOG: consistent recovery state reached at 9/BA2D3490
2012-03-16 10:47:02 CDT LOG: database system is ready to accept read only connections
2012-03-16 10:47:02 CDT LOG: invalid record length at 9/BA2D3490
2012-03-16 10:47:02 CDT LOG: streaming replication successfully connected to primary
Is this something I should be concerned about (and even if not, where can I read about what these mean in detail)?
Also, is there a more elegant shutdown command I can use so that when it starts up again, I don’t get these “invalid record length” type messages?
P.S. I have searched the forums and the docs several times, so I apologize if I missed the answer/tutorial on this somewhere.
Thanks,
Eric Simon
On Sat, Mar 31, 2012 at 1:20 AM, Eric Simon <esimon@theiqgroup.com> wrote: > 2012-03-16 10:27:34 CDT LOG: invalid record length at 9/B9F85E78 > 2012-03-16 10:27:34 CDT LOG: streaming replication successfully connected > to primary Just after startup, the standby tries to read and replay the WAL files which exist in the standby, and then if it finds invalid WAL record, IOW, it can no longer read valid WAL record locally, it starts replication and tries to read the WAL files from the master. So the above log message "invalid record ..." means that there is no valid WAL record in the standby, which triggers the replication. Since valid WAL records are streamed from the master, you don't need to worry about that log message. Regards, -- Fujii Masao NIPPON TELEGRAPH AND TELEPHONE CORPORATION NTT Open Source Software Center
I see. Is there a more elegant way that I could shut down my standby server so that it would be in a more clean state? For example, when I first built the standby and began replication, I didn't get either of those record length messages. -- View this message in context: http://postgresql.1045698.n5.nabble.com/Warnings-in-Standby-Logs-tp5607135p5613002.html Sent from the PostgreSQL - admin mailing list archive at Nabble.com.
Sorry for the delay. On Tue, Apr 3, 2012 at 1:44 AM, masterchief <esimon@theiqgroup.com> wrote: > I see. Is there a more elegant way that I could shut down my standby server > so that it would be in a more clean state? For example, when I first built > the standby and began replication, I didn't get either of those record > length messages. No, there is no more elegant way. Currently you cannot avoid to see those messages. Regards, -- Fujii Masao