May 26, 2004 at 5:19 pm
I've a clustered SQL 7.0 on a NT 4.0 Enterprise edition. Several weeks ago, the server started hanging in a more or less random fashion.
By hang, I mean all users connected would simultaneously freeze. This was/is true of those accessing via applications, via query analyzer, or enterprise manager. Some processes could access via command line. Internal jobs appeared to continue without pause. EG I have a job which issues a date-stamp record every minute. That continues to function
The hangs last from 2-8 minutes, though the last two weeks most of the hangs only last 2-3 minutes.
We've taken numerous PSSDIAG dumps. The only counter which seem to predict a hang was and increase from 0 to 40 or so of memory grants pending.
During the hang, CPU utilization drops to near zero, disk queue length increases, and batch requests plumment. Rows processed per the profiler drops from 9000/sec to 3000/sec.
It's as if work completed isn't being sent out, so no new requests are coming in.
Sometimes this occurs once a day during full activity, sometimes at low activity. There seems to be no real pattern other than it happens more frequently during high activity than during low.
When the hang completes all users are able to access the database, CPU spike high for a few minutes and then returns to normal. No data appears to be lost
MS has found nothing so far from the various dumps and logs we've sent them. They appear somewhat mystified.
May 27, 2004 at 6:29 am
Take an immediate snapshot of sp_who2 and use Performance Monitor to check Network, Memory, IO and CPU while the problem is occurring.
Sounds like somewhere the network may be fuzing out a bit.
Also check with Profiler to see if any specific query begins near that time, you might need to use a blackbox trace (see bol) to catch thou.
Viewing 2 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply