Issues with ".blg" file getting corrupted

May 17, 2010 at 3:40 PM

First of all, let me start by saying thanks so much for the PAL tool! It is a great resource and much appreciated.

I have been spending a lot of time recently troubleshooting errors with ".blg" log files I have gathered using perfmon. This is an intermittent issue where occasionally after letting perfmon log for a desired amount of time, I will stop the counter with a "logman stop countername", copy the file to my local admin machine to run it with PAL. I am running into the "data source is invalid.......sample contains less than two samples" error. I run perfmon from a .cmd batch file that references a counter.txt file. It seems like I have a 50/50 chance of getting this error. So much to the point that I now make it a habit to start PERFMON on my admin box and open the .blg that is being collected on the machine under to test just to verify that the data is good.

After researching the web, I found a KB that states you will get this error if your file is too small, it didnt run long enough, OR, the file is corrupted. The first two are out which leaves me with file corruption. After MUCH troubleshooting and tracking when and where I get this error, it seems I have it narrowed down to the "wmiapsrv" service (for the meantime ayway). Even if I check to make sure it is running before I start my perfmon script, I have had this error pop up. However, if I give it a fresh stop/start before running the perfmon job, I get consistently good data, at least so far... for the mean time I am still trying to see if this service is my culprit......

 

Just curious if anyone out there has encountered something similar?

Thanks for any help,

Shawn

 

Jun 3, 2010 at 6:01 PM

Nobody else seen this issue?

Jun 10, 2010 at 3:57 PM

I have recently started using perfmon and encountered corrupted blg files as well, though I receive slightly different error message: "... is either invalid or cannot be found. Unable to read counter information...". I am capturing large size samples (75 counters, 6-10 hour run) but by breaking them into relatively smaller log files (20MB each). What is surprising is that only few intermittent log files are corrupted, even though they are all 20MB in size. I would appreciate if you could share how you traced the problem to wmiapsrv process (which, I guess you know is critical to permoance minotring). I am unable to open the corrupt log files using perfmon or PAL.

Jun 10, 2010 at 5:32 PM

In a world where reboots seem to fix (almost) everything, I noticed that after getting corrupted log files, I rebooted the server and reran the perfmon setup again and it worked and I had good .blg data. Obviously rebooting each time is not feasible so I made it a point to restart the wmiapsrv service before the perfmon setup and this seems to have helped so far. By the way, I am using a CMD file that sets up perfmon counters, so in the script I have it stop and start the service before created the counters. I am not 100% convinced that bouncing this service is the fix but it seems to have helped so far. I will make sure to update this post if get more corrupted files (even after bouncing the service) so hopefully together we can narrow this down.

 

Coordinator
Jun 22, 2010 at 12:35 AM

Hi stjcal,

I'm not sure what is causing your logs to become corrupted. If this is an urgent issue, then please consider opening up a support case with Microsoft Customer Support Services (CSS).