On Tue, Apr 22, 2014 at 1:28 PM, wd <wd@wdicc.com> wrote:
I'v try to make a base backup use pg_start_backup, pg_stop_backup and rsync, but failed.
After run select pg_start_backup('label') on db server, I run rsync on backup server, and got some errors like 'file has vanished: xxxx', think this should be ok since it's documented in the manual. Then run select pg_stop_backup() to end the backup.
But when I try to use this backup to restore to a specific time on backup server, got errors bellow:
[ 2014-04-11 20:50:47.610 CST]LOG: restored log file "00000001000005FF000000F8" from archive [ 2014-04-11 20:50:56.698 CST]LOG: restored log file "00000001000005FF000000F9" from archive [ 2014-04-11 20:51:05.552 CST]LOG: restored log file "00000001000005FF000000FA" from archive [ 2014-04-11 20:51:16.171 CST]LOG: restored log file "00000001000005FF000000FB" from archive [ 2014-04-11 20:51:20.785 CST]WARNING: specified item offset is too large [ 2014-04-11 20:51:20.785 CST]CONTEXT: xlog redo insert: rel 1663/1/12633; tid 1/15 [ 2014-04-11 20:51:20.785 CST]PANIC: btree_insert_redo: failed to add item [ 2014-04-11 20:51:20.785 CST]CONTEXT: xlog redo insert: rel 1663/1/12633; tid 1/15 [ 2014-04-11 20:51:20.903 CST]LOG: startup process (PID 4758) was terminated by signal 6: Aborted [ 2014-04-11 20:51:20.903 CST]LOG: terminating any other active server processes
We do this base backup everyday, and the error occur sometimes, not always happen. I think it's must something wrong in our operations, anyone know why this happen? May be we should not use rsync or there is something impotent we missed ?
What is the version used? Recovery is not able to handle properly the redo for an XLOG record related to btree.