Thread: fsync error when restoring from archived xlogs
Hi All, Trying to do a PITR on postgres 8.1.8; I've restored the last full backup of the cluster dir and put the required WAL files into place. Yet when I attempt the recovery I see the following: Jun 18 15:44:11 postgres[29730]: [3-1] LOG: starting archive recovery Jun 18 15:44:11 postgres[29730]: [4-1] LOG: restore_command = "cp / var/lib/pgsql/backups/oldwal/%f %p" Jun 18 15:44:11 postgres[29730]: [5-1] LOG: recovery_target_time = 2007-06-16 22:00:00+09:30 Jun 18 15:44:11 postgres[29730]: [6-1] LOG: restored log file "000000010000000000000002.00DD29D8.backup" from archive Jun 18 15:44:11 postgres[29730]: [7-1] LOG: restored log file "000000010000000000000002" from archive Jun 18 15:44:11 postgres[29730]: [8-1] LOG: checkpoint record is at 0/2DD29D8 Jun 18 15:44:11 postgres[29730]: [9-1] LOG: redo record is at 0/2DD29D8; undo record is at 0/0; shutdown FALSE Jun 18 15:44:11 postgres[29730]: [10-1] LOG: next transaction ID: 4236; next OID: 32591 Jun 18 15:44:11 postgres[29730]: [11-1] LOG: next MultiXactId: 1; next MultiXactOffset: 0 Jun 18 15:44:11 postgres[29730]: [12-1] LOG: automatic recovery in progress Jun 18 15:44:11 postgres[29730]: [13-1] LOG: redo starts at 0/2DD2A1C Jun 18 15:44:13 postgres[29730]: [14-1] LOG: restored log file "000000010000000000000003" from archive ... Jun 18 15:44:39 postgres[29730]: [49-1] LOG: restored log file "000000010000000000000026" from archive Jun 18 15:44:39 postgres[29730]: [50-1] LOG: recovery stopping before commit of transaction 1809419, time 2007-06-16 22:24:47 CST Jun 18 15:44:39 postgres[29730]: [51-1] LOG: redo done at 0/264A55A0 Jun 18 15:44:39 postgres[29730]: [52-1] LOG: selected new timeline ID: 2 Jun 18 15:44:39 postgres[29730]: [53-1] LOG: archive recovery complete Jun 18 15:44:39 postgres[29730]: [54-1] LOG: could not fsync segment 0 of relation 1663/42607/44092: No such file or directory Jun 18 15:44:39 postgres[29730]: [55-1] PANIC: storage sync failed on magnetic disk: No such file or directory Jun 18 15:44:39 postgres[29726]: [2-1] LOG: startup process (PID 29730) was terminated by signal 6 Jun 18 15:44:39 postgres[29726]: [3-1] LOG: aborting startup due to startup process failure Any ideas? I'm not a subscriber so please cc: me on any responses. Regards, Tom -- Tom Lanyon Systems Administrator NetSpot Pty Ltd
On Mon, 2007-06-18 at 16:06 +0930, Tom Lanyon wrote: > Trying to do a PITR on postgres 8.1.8; I've restored the last full > backup of the cluster dir and put the required WAL files into place. This problem is fixed in 8.1.9. Seems it was inadvertently introduced in 8.1.7 (Jan 2007) and was not reported as a bug until after 8.1.8 was out, in early April. The 8.1.9 release notes make no mention of the bug fix, which isn't very helpful, I'll agree. > Jun 18 15:44:39 postgres[29730]: [54-1] LOG: could not fsync segment > 0 of relation 1663/42607/44092: No such file or directory > Jun 18 15:44:39 postgres[29730]: [55-1] PANIC: storage sync failed > on magnetic disk: No such file or directory It's a duplicate bug. Please check next time before you report. -- Simon Riggs EnterpriseDB http://www.enterprisedb.com
On 20/06/2007, at 6:48 AM, Simon Riggs wrote: > This problem is fixed in 8.1.9. Seems it was inadvertently > introduced in > 8.1.7 (Jan 2007) and was not reported as a bug until after 8.1.8 was > out, in early April. > > The 8.1.9 release notes make no mention of the bug fix, which isn't > very > helpful, I'll agree. > > It's a duplicate bug. Please check next time before you report. Simon, Thanks for the reply. Good to know that I wasn't just doing something stupid. :) To be fair, I did search for an answer before posting to the list. However, I must've been searching for the wrong terms as my search didn't turn up anything on the -bugs list! Regards, Tom