LOGMGR is a wait that occurs when a database is being closed and a thread has to wait for all outstanding transaction log I/O to complete before the log can be shut down. Databases close during instance shutdown, when a database is taken offline or detached, and when the AUTO_CLOSE option closes an idle database.
Do not confuse it with LOGMGR_QUEUE, which is the log writer thread idling and is completely benign. Plain LOGMGR is about closing the log down cleanly.
Is It a Problem?
Almost never in a steady-state wait profile, because databases are not usually closing during normal operation. Where it can bite is in two situations. First, slow shutdowns or offline/detach operations that hang while in-flight log writes drain to slow storage. Second, databases with AUTO_CLOSE ON cycling open and closed repeatedly, paying this cost over and over.
If LOGMGR accumulates on a production instance, ask why databases are closing at all.
Common Causes
AUTO_CLOSE ON, common on databases restored from Express-edition or vendor-supplied backups, causing constant close/open cycles.- Taking databases offline, detaching them, or shutting the instance down while a heavy write workload still has log I/O in flight.
- Slow log-drive storage stretching out the final flush during a close.
What To Do
- Check for auto-close:
SELECT name FROM sys.databases WHERE is_auto_close_on = 1;and turn it off on anything a user or application touches:ALTER DATABASE [db] SET AUTO_CLOSE OFF;. - If shutdowns or offlines hang, look at log write latency in
sys.dm_io_virtual_file_statsbefore blaming the engine; the close is waiting on the same slow disk your commits are. - Quiesce write activity before planned detach or offline operations so there is little log left to drain.
How To See It
Rank it against everything else with Get-WaitStatistics. If LOGMGR registers meaningfully, correlate it with database state changes in the error log rather than treating it as a workload wait.
Part of the SQL Server Wait Types Library.
Related deep dive: WRITELOG Wait Type.
Leave a Reply