We recently added some secondary tempdb files to reduce allocation contention, this seems to have worked; however we are now getting lots of slowdowns related to I/O (WRITELOG and SQLTRACE_FILE_BUFFER waits specifically).
Having had a look around the issues seem to coincide with all four of the tempdb files being written to at the same high rate. I'd expected the reads and writes to the various tempdbs to be independant, so my questions are:
- Is this a limitation with activity monitor being unable to discern the difference between the files or are they all actually being written to?
- Does this indicate the disk I/O rate is actually the limiting factor, or is there something else going on?
- Does whatever this issue is occur in newer versions of sql server?
Estimated execution plan of the query I believe to be causing the issue
Actual execution plan including the query I think is the main one causing experiencing the issue
tempdb
and can you add those queries with their execution plans to your post please? What are the specs for the disk(s) yourtempdb
, database data file, and database log file live on?** Restricted Text **
. I'll anonymise an execution plan and add it.tempdb
if that's the issue.sp_WhoIsActive
will give you better details on this. Also, you can use SentryOne Plan Explorer to automatically anonymize your execution plans.