Tales From A Lazy Fat DBA

Hi All,

Ever encountered a situation where the backup history (TIMELINE_HISTORY) file was deleted by mistake or maybe someone removed it purposely, it was quite old and you try to restore a new backup. I remember many issues related with replication, backup tools (BARMAN & BART) that you might face if that file is removed from PG_WAL directory. Would like to discuss a problem that we encountered while taking BART backup on EDB version 10.

These ‘timeline files’ are quite important, as using the timeline history files, the pg_basebackup can follow the latest timeline present in the primary, just as it can follow new timelines appearing in an archive WAL directory. So, in short, it shows which timeline it branched off from and when. These history files are necessary to allow the system to pick the right WAL segment files when recovering from an archive that contains multiple timelines. So, its important to have this file there in WAL directory.

[enterprisedb@fatdba ~]$ bart -c /usr/edb-bart-1.1/etc/bart.cfg BACKUP -s edbserver --backup-name MAINFULLBKP_10-13-20
INFO:  DebugTarget - getVar(checkDiskSpace.bytesAvailable)
INFO:  creating full backup using pg_basebackup for server 'edbserver'
INFO:  creating backup for server 'edbserver'
INFO:  backup identifier: '1602788909136'
ERROR: backup failed for server 'edbserver' 

pg_basebackup: could not send replication command "TIMELINE_HISTORY": ERROR:  could not open file "pg_wal/00000002.history": No such file or directory 

1633701/1633701 kB (100%), 2/2 tablespaces
pg_basebackup: child process exited with error 1
pg_basebackup: removing data directory "/edbbackup/edbserver/1602788909136"
 

The file is not there under said directory.

[enterprisedb@fatdba ~]$ cd /edb/as10/as10/data/pg_wal/
[enterprisedb@fatdba pg_wal]$ ls
0000000200000005000000EA  0000000200000005000000EB.00000060.backup  0000000200000005000000ED  archive_status
0000000200000005000000EB  0000000200000005000000EC                  0000000200000005000000EE
 

In case of file missing/moved, you can always create a brand new empty file and that will be used by the respective utility and will be populated with metadata soon after. So, in order to quickly restore this issue, let’s create one.

 [enterprisedb@fatdba pg_wal]$ touch 00000002.history
[enterprisedb@fatdba pg_wal]$
[enterprisedb@fatdba pg_wal]$ ls *hist*
00000002.history 

Let’s try to take the backup once again.

[enterprisedb@fatdba pg_wal]$ bart -c /usr/edb-bart-1.1/etc/bart.cfg BACKUP -s edbserver --backup-name MAINFULLBKP_10-13-20
INFO:  DebugTarget - getVar(checkDiskSpace.bytesAvailable)
INFO:  creating full backup using pg_basebackup for server 'edbserver'
INFO:  creating backup for server 'edbserver'
INFO:  backup identifier: '1602789425665'
INFO:  backup completed successfully
INFO:
BART VERSION: 2.5.5
BACKUP DETAILS:
BACKUP STATUS: active
BACKUP IDENTIFIER: 1602789425665
BACKUP NAME: MAINFULLBKP_10-13-20
BACKUP PARENT: none
BACKUP LOCATION: /edbbackup/edbserver/1602789425665
BACKUP SIZE: 1.57 GB
BACKUP FORMAT: tar
BACKUP TIMEZONE: Europe/Berlin
XLOG METHOD: stream
BACKUP CHECKSUM(s): 0
TABLESPACE(s): 1
 Oid     Name      Location
 42250   UNKNOWN   /edb/as10/as10/data_test/pg_tblspc

START WAL LOCATION: 0000000200000005000000ED
BACKUP METHOD: streamed
BACKUP FROM: master
START TIME: 2020-10-15 21:17:05 CEST
STOP TIME: 2020-10-15 21:17:38 CEST
TOTAL DURATION: 33 sec(s)


[enterprisedb@fatdba pg_wal]$  bart -c /usr/edb-bart-1.1/etc/bart.cfg SHOW-BACKUPS
 SERVER NAME   BACKUP ID       BACKUP NAME            BACKUP PARENT   BACKUP TIME                BACKUP SIZE   WAL(s) SIZE   WAL FILES   STATUS

 edbserver     1602789425665   MAINFULLBKP_10-13-20   none            2020-10-15 21:17:38 CEST   1.57 GB       16.00 MB      1           active
 

And it worked.

Hope It Helped!
Prashant Dixit