For various reasons, including SERVER-1423 (which can prevent fsyncUnlock from being issued if the connection on which the fsyncLock is done is closed), or plain user error in coding scripts, a mongod instance can end up being locked unexpectedly for an extended time. When this happens the cause of the resulting symptoms can be difficult to diagnose, since the only evidence in the logs is typically an fsyncLock command, which may be far removed (e.g. in a different log file) from the time when the error is noticed.
Since mongod instance being locked for an extended time is a potentially serious condition, and is likely unintended, it would be helpful for diagnosis if periodic warnings were issued to the log, e.g. "WARNING: instance is locked; all db writes and possibly all db reads will be blocked". Such a warning possibly as often as once per minute might be justified.
- is related to
-
DOCS-3681 warn users that fsyncUnlock needs to be issued on same connection as fsyncLock
- Closed