archiver : ERROR: recovery is in progress HINT: pg_walfile_name() cannot be executed during recovery.

[Krysztophe]

After a switchover, on the secondary (formerly primary), the archiver service fails with

ERROR:  recovery is in progress
HINT:  pg_walfile_name() cannot be executed during recovery.

The service was fine on the secondary before the switchover.

I'm surprised to see this only now.

Cause: pg_stat_archiver is not as empty as I see usually, it contains WAL names, eg:

select pg_is_in_recovery() , current_setting('archive_mode') as archive_mode, * from  pg_stat_archiver \gx
-[ RECORD 1 ]------+------------------------------
pg_is_in_recovery  | t
archive_mode       | on
archived_count     | 860475
last_archived_wal  | 0000000100000DA200000037
last_archived_time | 2025-09-08 14:02:02.458681+02
failed_count       | 1227
last_failed_wal    | 0000000100000D9C000000EC
last_failed_time   | 2025-09-08 11:46:34.035206+02
stats_reset        | 2025-07-16 12:18:51.395272+02

Fix : Add CASE WHEN pg_is_in_recovery() THEN null ELSE extract ... there ?

[frost242]

Hi @Krysztophe

It's fine for this fix. Would you like to do it ?

[Krysztophe]

WIP

[Krysztophe]

In fact,it sounds similar to #358 and the code has a FIXME:

# FIXME: on a slave with archive_mode=always:
        #  1) fails while parsing an .history file
pg_last_wal_receive_lsn always returns zero if the slave is fed
        #     with pure log shipping (streaming is ok

I'd like to solve this all at once, and I need to test different versions, so it may take more time than I thought

Edit : sorry, I could not work on it and I don't know when it will be possible

[frost242]

@Krysztophe I took care of issue #358 as I sometime face it with my job. I'll try to fix this one ASAP too.