Negative Run Duration

Question

Post reply

Negative Run Duration

Brian Munier

SSCrazy

Points: 2875
More actions
November 29, 2007 at 1:26 pm

#184472

I am using SQL Server 2000 Enterprise Edition, Sp4 + Hotfix on a properly configured Windows 2003 Server R2 with 10 GB Ram on a two-node active-active cluster.
I have had my Job history report durations of:
-95443:-92:-42
Any thoughts?

Viewing 15 posts - 1 through 15 (of 20 total)

You must be logged in to reply to this topic. Login to reply

Steve Jones - SSC Editor SSC Guru Points: 736270 More actions · Answer 1

Sounds like a bug.

Is this the first time the job was run? Have you checked the raw data in the msdb tables?

george sibbald SSC Guru Points: 104210 More actions · Answer 2

I got exactly this run timing on a job that kicked off at exactly the time the clocks went back. Any differences in system time between the nodes. sounds like SQLAgent thinjs the job finished before it started, if you see what I mean.

---------------------------------------------------------------------

Brian Munier SSCrazy Points: 2875 More actions · Answer 3

Thank you both.

First the job has been running for years. Second I thought we dealt with the DST back in the spring on this server.

It is still occuring, any thoughts on how to stop it? Re-do the job or maintenance plan?

Thanks in advance, I am glad to know I am not crazy.

Brian

george sibbald SSC Guru Points: 104210 More actions · Answer 4

what does the job do? is it long running or a quick job, how is it kicked off?

---------------------------------------------------------------------

Steve Jones - SSC Editor SSC Guru Points: 736270 More actions · Answer 5

Can you query the raw data in msdb and see what's listed? sysjobhistory?

Brian Munier SSCrazy Points: 2875 More actions · Answer 6

Ok, let's see the job is a transaction log backup part of a maintenance plan.

The maintenance plan is actually created by LiteSpeed enterprise edition version 4.5.204.

I am running this setup on several different server (16) in several different clusters. I haven't noticed the problem on other servers, but I will double check.

Brian Munier SSCrazy Points: 2875 More actions · Answer 7

And to answer the other question, yes the raw data is in the msdb sysjobhistory table where the run duration is negative.

Steve Jones - SSC Editor SSC Guru Points: 736270 More actions · Answer 8

Very strange. Is it only the Litespeed jobs?

I wouldn't think that Litespeed has anything to do with it. The only thing that would make sense would be a time adjustment during the job, something like the DST change. How long does it take for the job to run?

Is this still continuing? What if you manually run the job now?

Brian Munier SSCrazy Points: 2875 More actions · Answer 9

It is primarily the LiteSpeed transaction backup job, which runs every 15 minutes, but I just found one instance in a job that was manually created to purge the old backup history. So I would agree it is not LiteSpeed.

I will have to double check time sources as we do try to keep them synched. maybe one of them is off.

Brian Munier SSCrazy Points: 2875 More actions · Answer 10

the t-log job normally takes about 3:16 to run so it is not that long.

The other job normally takes about 1 minute, but on the time it failed, a second report line reported it having a 0 duration.

Brian Munier SSCrazy Points: 2875 More actions · Answer 11

Another interesting note: the other node in the cluster is about 5 minutes slower than the node with issues. In fact the node with issues is 5 minutes fast.

george sibbald SSC Guru Points: 104210 More actions · Answer 12

Brian, sorry, nothing is jumpimg out at me. As I say I have only seen this when the job ran at the same time as the clocks going back and if memory serves me on very long running jobs which were eventually cancelled or failed.

Interesting the nodes are out, got to be worth getting them in synch. Not sure quite how this would affect it because I would have thought the job would fail if there was a failover whilst the job was running.

---------------------------------------------------------------------

Steve Jones - SSC Editor SSC Guru Points: 736270 More actions · Answer 13

I'm with George. Nothing jumps out. The clocks shouldn't cause this, though you probably want them to get in synch. Could possibly be something there, but I'd think the job would be dependent only on the active node.

If you've got a case to open with MS, this would be interesting to track down, though probably not terribly useful as anything other than trivia.

One last thing, are the jobs running ok, meaning getting the work done? If so, I'd wait for the next reboot, hope it goes away.

Brian Munier SSCrazy Points: 2875 More actions · Answer 14

Steve Jones - Editor (12/4/2007)
I'm with George. Nothing jumps out. The clocks shouldn't cause this, though you probably want them to get in synch. Could possibly be something there, but I'd think the job would be dependent only on the active node.
If you've got a case to open with MS, this would be interesting to track down, though probably not terribly useful as anything other than trivia.
One last thing, are the jobs running ok, meaning getting the work done? If so, I'd wait for the next reboot, hope it goes away.

I am not going to open a case with MS. The jobs appear to complete correctly. I will work to get the nodes synched.

Ironically enough, I first noticed the issue after I inadvertently rebooted the node. Oh well we will see.

thanks for your thoughts.