'ERROR: Not going to purge WAL because LSN is invalid' when there is backup with ERROR state

[alexign]

Got some strange behavior when by error make DELTA backup before any FULL in stream remote mode

root@backupserver:~/backup-tools# pg_probackup-10 show -B /backup/postgresql/backup --instance database
==============================================================================================================
 Instance  Version  ID  Recovery Time  Mode  WAL  Current/Parent TLI  Time  Data  Start LSN  Stop LSN  Status
==============================================================================================================
root@backupserver:~/backup-tools#

backupserver:~/backup-tools# pg_probackup-10 \
>         backup \
>         -B \
>         /backup/postgresql/backup \
>         --stream \
>         -j \
>         1 \
>         --retention-redundancy=14 \
>         --delete-expired \
>         --delete-wal \
>         --instance \
>         database \
>         --remote-host \
>         databaseserver \
>         --remote-user \
>         postgres \
>         -U \
>         probackup \
>         -d \
>         pg_probackup \
>         --log-level-file=verbose \
>         --log-filename=database-20190726T182724.log \
>         --log-directory=/backup/postgresql/backup/log/database \
>         --log-rotation-age=0 \
>         --ssh-options='-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -o ServerAliveInterval=60 -o ServerAliveCountMax=5' \
>         --compress \
>         --compress-level \
>         1 \
>         --backup-mode=DELTA
INFO: Backup start, pg_probackup version: 2.1.3, backup ID: PV9J42, backup mode: delta, instance: database, stream: true, remote true
ERROR: Valid backup on current timeline is not found. Create new FULL backup before an incremental one.
WARNING: Backup PV9J42 is running, setting its status to ERROR

backupserver:~/backup-tools# pg_probackup-10 show -B /backup/postgresql/backup --instance database
=======================================================================================================================
 Instance   Version  ID      Recovery Time  Mode   WAL     Current/Parent TLI  Time  Data  Start LSN  Stop LSN  Status
=======================================================================================================================
 database  10       PV9J42  ----           DELTA  STREAM  1 / 0                 0s    0B        0/0       0/0  ERROR

backupserver:~/backup-tools# pg_probackup-10 \
>         backup \
>         -B \
>         /backup/postgresql/backup \
>         --stream \
>         -j \
>         1 \
>         --retention-redundancy=14 \
>         --delete-expired \
>         --delete-wal \
>         --instance \
>         database \
>         --remote-host \
>         databaseserver \
>         --remote-user \
>         postgres \
>         -U \
>         probackup \
>         -d \
>         pg_probackup \
>         --log-level-file=verbose \
>         --log-filename=database-20190726T182724.log \
>         --log-directory=/backup/postgresql/backup/log/database \
>         --log-rotation-age=0 \
>         --ssh-options='-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -o ServerAliveInterval=60 -o ServerAliveCountMax=5' \
>         --compress \
>         --compress-level \
>         1 \
>         --backup-mode=FULL
INFO: Backup start, pg_probackup version: 2.1.3, backup ID: PV9J61, backup mode: full, instance: database, stream: true, remote true
INFO: Start transfering data files
INFO: Data files are transfered
INFO: wait for pg_stop_backup()
INFO: pg_stop backup() successfully executed
INFO: Validating backup PV9J61
INFO: Backup PV9J61 data files are valid
INFO: Backup PV9J61 completed
INFO: Evaluate backups by retention
INFO: Backup PV9J61, mode: FULL, status: OK. Redundancy: 1/14, Time Window: 0.00d/0d. Active
INFO: Backup PV9J42, mode: DELTA, status: ERROR. Redundancy: 2/14, Time Window: 0.00d/0d. Active
ERROR: Not going to purge WAL because LSN is invalid
backupserver:~/backup-tools# echo $?
1
root@backupserver:~/backup-tools# pg_probackup-10 show -B /backup/postgresql/backup --instance database
====================================================================================================================================
 Instance   Version  ID      Recovery Time           Mode   WAL     Current/Parent TLI  Time   Data   Start LSN    Stop LSN  Status
====================================================================================================================================
 database  10       PV9J61  2019-07-26 18:54:46+00  FULL   STREAM  1 / 0                50s  296MB  4/1B000028  4/1B00BD50  OK
 database  10       PV9J42  ----                    DELTA  STREAM  1 / 0                 0s     0B         0/0         0/0  ERROR

so FULL backup end with return code <> 0

[gsmolk]

Definitely a bug, for WAL purge purposes we should have selected oldest backup with valid START LSN, not just oldest backup.

Thanks for the feedback!

[gsmolk]

Fixed, TODO: improve test coverage with this case.

[gsmolk]

Test added, merged to master