We have had a few customer escalations where recovery was very slow and there were no log messages to indicate what WiredTiger was doing.
MongoDB opens wiredtiger with recovery_progress verbose messaging enabled so it can provide feedback to a user (or TSE) who may be waiting for the system to restart.
But apparently there can be long intervals (> 3 hours in one case) where we generate no recovery_progress messages. We should understand where WiredTiger might be spending time and add one or more additional recovery_progress messages so users will see continual progress.
- related to
-
WT-9607 Fix timing of compact progress messages
- Closed