docs: one process per live file, a distinct history_name per service

Signed-off-by: disqualifier <dev@disqualifier.me>
This commit is contained in:
dsql committed 2026-10-16 04:40:13 -04:00
1 parent 5b27df65e7
commit 1a45b9778c
1 file changed
+6
+6
View File
@@ -85,6 +85,12 @@ setup_logging(name="latest", rotate="daily")
- Retention (tier counts / `backup_count`) is unchanged — it's just keyed to the project
stem now.
**One process per live file, one `history_name` per service.** Rotation and retention
assume a single writer: two processes writing the same live file both try to roll it, and
two services sharing a `log_dir` with the same history stem prune each other's history
(each counts the other's rolls against its own `backup_count`). Give each process its own
live file, and give services that share a cwd or `log_dir` distinct `history_name`s.
## Tiered retention (`keep_uncompressed` / `keep_compressed`)
The default is a flat `backup_count`: every rolled file is gzipped on roll and the oldest