Jul 15, 2008 at 3:58 PM
Edited Jul 28, 2008 at 2:05 PM
I have added some counters to PAL but at the report they appear as "Stats only" but I have configured some tresholds. How can I see the alerts of this counters?
Jul 29, 2008 at 12:38 AM
The results of the analysis should only be "Stats Only" only if you have not specified any thresholds. Please send me or post the threshold file (XML), so I can see what is going on. Did you add the threshold using the PAL Editor?
Jul 29, 2008 at 9:48 AM
Edited Jul 29, 2008 at 2:29 PM
Hello Clint and thank you for helping me.

I have this problem with all the counter of Physycal Disk and with Processor % Processor Time and I have added the threshold with the Pal Editor. The report don't show graphs for Physical Disk.

I can't paste here the XML because it is too long. I have sent it to your user.
Aug 9, 2008 at 5:23 AM
My apologies for taking so long to reply. Other work is keeping me busy. Can you post it to the Issue Tracker area, or try sending it to me again ( I tried searching my email today for your email, but can't seem to find it. Thank you.
Aug 25, 2008 at 3:33 PM
Don't worry about it, I just arrived today from my holidays.

I sent it to your e-mail.

Thank you.
Aug 30, 2008 at 1:41 AM

Hi Riesaz,

I tested your threshold file and found a few things.

First, you will stats only when there are no threshold defined for the analyses. When I looked at your Physical Disk analyses your Current Disk Queue Length is the only one that has a threshold defined. The threshold works fine when I tested it. You've got a great start, you just need to add the thresholds to the analysis.

Also, PAL is not showing a graphical chart until you define one. This means you need to click the "Edit Chart..." button, then click OK when you are done choosing a chart template. After you create the chart, be sure you click the Update Analysis button for the changes to take effect, and click File, Save, to save it to the XML file.

Regarding the thresholds you are creating, we don't use % Disk Time counters because this counter was originally based on a single physical disk, therefore we cannot draw any conclusions or recommendations from this counter. The Current Disk Queue Length counter varies widely in values because disk queue can get very large and empty out in a few milliseconds, therefore Avg. Disk Queue Length is better. We don't use Avg. Disk Queue Length because in order to get a good conclusion from it, then we would need to know the number of phyiscal spindles behind the LUN. With SAN becoming more common, it's difficult to know how many real physical disks there are, therefore we don't generally use this counter. The only disk counters that we can draw conclusions from is the Avg Disk sec/Read and Avg Disk sec/Write. These counters give us the time in seconds of how long disk I/O requests are taking. Average I/O taking longer than 15ms is considered to be slow, therefore PAL has thresholds for this.

We don't use % Idle Time because again it is based on a single physical disk versus a RAID array. We had a lot of heated debates on this counter here at Microsoft and finally went to the source code of it. The source shows that it simply checks if the disk queue has anything in it or not and calculates the percentage of time it is empty. This means if you have 1 outstanding I/O and have 100 disks behind the LUN, then it would be 0% idle, but in effect it would not be a bottleneck.

For the same reasons as above, we don't use phyiscal disk counters much because we don't really know how many disks there are and applications and services get to a disk by drive letter (logical disk), therefore Logical Disk performance is the thing to look at.