Обсуждение: pg_xlog error
Hi, I have problems in my database. I think it is corrupted. Folow my log when I tried to start it standalone. I have some questions: 1. I saw that the error is in base/30518/449778670_vm file. Can I rebuild this file or somethink like this? 2. In the last line of log, we can see "DEBUG: shmem_exit(1): 8 callbacks to make". Where can I find these transactions to do callback and/or how can I do these callbacks? If anyone can help me, I will be really glad. DEBUG: invoking IpcMemoryCreate(size=39149568) DEBUG: removing file "pg_notify/0000" DEBUG: InitPostgres DEBUG: my backend id is 1 LOG: database system shutdown was interrupted; last known up at 2011-07-11 11:20:55 BRT DEBUG: checkpoint record is at 601/C760AA08 DEBUG: redo record is at 601/C760AA08; shutdown TRUE DEBUG: next transaction ID: 0/2960582638; next OID: 476667689 DEBUG: next MultiXactId: 234588; next MultiXactOffset: 522535 DEBUG: oldest unfrozen transaction ID: 914493351, in database 18415351 DEBUG: transaction ID wrap limit is 3061976998, limited by database with OID 18415351 LOG: database system was not properly shut down; automatic recovery in progress LOG: consistent recovery state reached at 601/C760AA60 LOG: redo starts at 601/C760AA60 LOG: record with zero length at 601/CA6A0CE8 LOG: redo done at 601/CA6A0CA0 LOG: last completed transaction was at log time 2011-07-10 15:34:02.372812-03 LOG: checkpoint starting: end-of-recovery immediate FATAL: xlog flush request 61B/245DD1F0 is not satisfied --- flushed only to 601/CA6A0CE8 CONTEXT: writing block 0 of relation base/30518/449778670_vm DEBUG: shmem_exit(1): 8 callbacks to make -- View this message in context: http://postgresql.1045698.n5.nabble.com/pg-xlog-error-tp4575216p4575216.html Sent from the PostgreSQL - hackers mailing list archive at Nabble.com.
On 11.07.2011 17:33, jcamera wrote: > Hi, > > I have problems in my database. I think it is corrupted. Folow my log > when I tried to start it standalone. > > I have some questions: > > 1. I saw that the error is in base/30518/449778670_vm file. Can I rebuild > this file or somethink like this? *_vm files contain the visibility maps of each relation. Visibility maps don't contain any user data, it's just bookkeeping information for vacuum to skip parts of tables that don't need vacuuming. They can be safely removed, the next vacuum will just take somewhat longer than otherwise. But clearly that error is an indication of corruption, and the visibility map probably was not the only thing that got corrupted. It looks like you're missing some WAL files, or they got corrupted. It's likely that you need to restore from a backup. And you should try to figure out what caused the corruption in the first place. Broken hardware, perhaps, or fsync=off and a power outage. > 2. In the last line of log, we can see "DEBUG: shmem_exit(1): 8 callbacks > to make". Where can I find these transactions to do callback and/or how can > I do these callbacks? That's just an internal debugging line you get when the startup process ends. You can ignore it. -- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com