Summary
In MongoDB, we call "log_flush" 10 times per second. When MongoDB is started with -v, or debug level 1 logging, we report 10 of these messages per second:

2022-02-02T05:33:23.113-05:00 D1 WTWRTLOG [JournalFlusher] WiredTiger message {"message":{"ts_sec":1643798003,"ts_usec":113184,"thread":"13924:0x7f2212039700","session_name":"WT_SESSION.log_flush","category":"WT_VERB_LOG","category_id":20,"verbose_level":"DEBUG","verbose_level_id":1,"msg":"log_flush: flags 0x4 LSN 11/5248"}}

This is much more noise in our logs than before, where previously debug level 1 (i.e. "-v") would only log infrequently.

We have several debug levels (1-5) in MongoDB, where 1 is the least logging, and 5 is the most. I would like to consider expanding the log levels in WiredTiger to match these levels and make changes to existing messages to fit into these new debug levels.

Motivation

• Does this affect any team outside of WT? This makes the MongoDB debug logs much more verbose by default.

• How likely is it that this use case or problem will occur? Any developer that uses "-v" to enable debug logging level 1 in MongoDB.

• If the problem does occur, what are the consequences and how severe are they? This is a productivity problem for MongoDB server developers. In order to enable logging at the first debug level, I have to configure the logging options to explicitly suppress these extra WT log message.

• Is this issue urgent? No

Acceptance Criteria (Definition of Done)

Starting mongod with the "-v" option does not produce "log_flush" messages every second.