Maximizing kernel dump file size
When Citrix servers that are streamed with Citrix Provisioning server crash the files that were added to the C: disk are gone. Citrix support needs these dump files to do a proper debugging of your issues. With a streamed Citrix XenApp virtual machine a D: disk is deployed to handle files that have to be kept during reboots or crashes.
Working on a Citrix case I was looking for those dump files on the D: disk, for that was were we configured the system to store them. I didn’t find any there so I went to check why that was. Here’s what I found.
First I checked the system settings.
We only want a kernel dump and not a full dump for a full dump would require to much space. A kernel dump file is much smaller and only contains the memory that took part in the crash, it could be around 1/3 of a normal dump file.
The Citrix XenApp 7.6 server have a lot of memory and a complete dump would really be to much to handle.
We’re using Cache to RAM of course so 16GB is usable the rest is for the cache.
We set the pagefile to a fixed value of 4096 for we don’t want the system to play with it for obvious reasons.
We are using a dedicated dump file as you can read in this Citrix article.
No if you don’t maximize the dump file it could grow to big for the disk so we want to maximize it.
There is a registry key for this that you need to add.
If you look under HKLMSystemCurrentControlSetControlCrashControl you see a number of settings. In the screenshot above a couple more have been added to make sure the dedicated dump file is used and the size is maximized.
The registry setting DumpFileSize controls the dump file size, the value is in MB’s so right now my server isn’t doing anything for with 2MB it can’t make a dump file. This is happening from time to time so that will be a check to do when finishing a golden image. The reason why the dump wasn’t created is obvious now so let’s change it.
To set the value to 4GB you enter 4096, after you deploy the golden image to all servers they will use this setting and can create a dump again.
Now we just wait and see until the next crash.. which will hopefully never happen 🙂
Hope this helps you guys, don’t know if I’m the only one seeing this, that the Dumpfilesize is reverted to 2 from time to time.