Parameter · Write-Ahead Log / Settings
wal_recycle
Recycles WAL files by renaming them.
- Default:
on - Type: bool
- Applies: Session (superuser)
In plain words
Old WAL files that are no longer needed are normally renamed and reused as future segments, which avoids creating new 16 MB files. On copy-on-write file systems creating fresh files can be faster, so wal_recycle = off deletes them instead.
For DBAs and operators
At the end of each checkpoint (or restartpoint on a standby) segments older than the redo point, and not held back by slots, wal_keep_size or archiving, are either recycled (renamed to a future segment name) or removed. How many are kept for recycling depends on min_wal_size, max_wal_size and the recent WAL rate. The checkpoint log line reports both: … 0 WAL file(s) added, 3 removed, 12 recycled; …
With off, all such segments are unlinked and new ones are created (zero-filled unless wal_init_zero = off). Recommended together with wal_init_zero = off on ZFS/Btrfs only.
How to change it
A superuser (or a role granted SET) can change it per session with SET.