February 27, 2013 at 2:25 pm
We have a high availability sql server environment with synchronous mirroring and witness which has been working mostly fine for more than one year.
But suddenly, last monday, the witness server became unresponsiveness and the errorlog showed lots of errors about insufficient memory.
Looking at the memory information dumped to the errorlog by sql server I found that most of the usable memory was taken up by sqlutilities:
>MEMORYCLERK_SQLUTILITIES (node 0)
>SinglePage Allocator986240
I do now know whether this is normal or not. This accounts for 94% of available memory because the witness instance has server memory setting configured 1GB.
To correct the incidence I restarted mssqlserver service on the witness, but 22 hours after that the same issue happened again.
I am pretty clueless about what went wrong so I restarted the witness once more time and then switched all our databases to asynchronous mode without witness. This seemed to work as the witness instance (now idle) ceased to consume memory.
The only other thing that came to my attention is that one of the main database servers (the one acting as secondary) has a network interface down. This server has two network interfaces bridged by a "Broadcom Advanced Server Program Driver" to provide a failsafe link. So even with one interface down this server seems to be working normally.
Any ideas?
March 7, 2013 at 10:11 am
The issue was related to misconfigured DNS server.
After fixing DNS configuration the the witness instance normalized, so the situation it is probably related to a bug or design flaw in the way sql server handles dns lookup timeouts/errors.
No further efforts were made in reproducing this problem.
regards,
Carlos.
March 7, 2013 at 10:14 am
...for the record: the environment is 64bit sql server 2008 10.0.5788 running on windows 2008 r2 OS.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply