Link to home
Start Free TrialLog in
Avatar of Cyprexx IT
Cyprexx IT

asked on

Waiting for idle_blocker which we cannot find

We have been seeing long wait times at random intervals throughout the day.  When we dig into these waits, it always seems they are being caused by an idle blocker (as we have discovered through DPA as well as some extended events.)  The problem is, once we realize this has happened, it's already too late.  The SPID that was causing the idle blocking is closed and we cannot determine what was running.  We have some extended events set up that give us a bit of information about the SPID (such as the call stack, duration, wait_resource, and wait_type) but none of this tells me exactly what stored procedure, query, or other process was running.  

Does anyone have a recommended way to store all information about past SPIDs, or any other advice on how to deal with this going forward?  At this point, I'm not worried about necessarily finding out what's happened in the past (though that is welcomed as well,) but how to get this information stored going forward so I can refer back to it.
SOLUTION
Avatar of dbaSQL
dbaSQL
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of Cyprexx IT
Cyprexx IT

ASKER

I have actually read through this article, and it does have some really good info.  I have turned on the Blocked Process Report, but the article itself is only useful if you know the blocking is going to occur, which I never do.  It specifically tells you to make sure you turn off the blocked process report at the end, so the scripts in here won't really help me since I have to know when it's happening to utilize this.
ASKER CERTIFIED SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Excellent, I will set up the Extended Event and leave it on.  As long as the overhead is minimal (which I agree with you, it certainly should be, or no worse that the other extended events I have running currently,) it shouldn't be a problem.  I was just concerned about the fact the first article was explicit on the fact it should be turned off in the end.  

As far as our situation, it seems to occur once every 10-14 days.  If it was daily, or hourly, I would have had no issue with leaving it on, but I wasn't sure about leaving it on for several weeks.  Once I get some information on what's causing this, I will be sure to turn it off again.  Thank you!
Excellent responses.  Thank you!
Completely my pleasure.  In a similar situation recently, I have left blocking process session on for weeks at a time, and had no related problems.  Just be sure to check status and output intermittently, and be sure nothing is getting away from you.