ESXi 5.5 (2718055) vmkdump Folder and dumpfile Generated on Every Boot

[sexconker]

I have a brand new server that was initially installed with ESXi 5.1 (custom image from Dell with needed network and RAID controller drivers) and immediately updated to ESXi 5.5 via Update Manager.

The system only has a single, local datastore.

When the system boots, I get a vmkdump folder with a dump file (396 MB) in the datastore.  I can clear this file out using esxcli, but it's recreated on the next boot.  Nothing appears to be wrong with the server, and I'm not getting any PSoDs.

I've read that there are changes to the way coredumps are handled in 5.5 and the fact that I first installed (and partitioned) with 5.1 means that the existing partitioning scheme doesn't leave enough room for 5.5's coredumps, so thy have to go to a datastore.  But what I'm not clear on is whether or not this file is supposed to be created on every boot or not.  I was under the impression that they were only created when a PSoD occurs, but I presume ESXi 5.5 could also create it on boot in order to reserve space for it.

Is the presence of a vmkdump folder on my local (only) datastore in ESXi 5.5 (upgraded from 5.1) normal?

Is the generation of a 396 MB dumpfile on boot normal?

If this is normal, is there any (relatively easy) way to resize the partitions such that the coredump can live in its own partition off the datastore?  How much space is required?  Losing the 110 MB (from the unused partition that's too small in 5.5) isn't a big deal, but I'd like to keep the datastore clean.

This is happening on several servers, 2 new ones as described above, and an older one that is in production that went from 5.0 to 5.5.  This leads me to believe everything is normal, but 'd like confirmation.

Thanks

[Spyingjack]

I am sure you might be getting PSOD screen, Can you please post screenshot? Mostly this is due to hardware failures (Memory Threshold) or CPU failures.

its generates machine code. " 0X0000000000000"

[sexconker]

There is no PSOD. I physically watch the entire cold boot process.

Additionally, only one file is ever created. If I leave it on the server and reboot, a second file is not created and the timestamp of the first file is not updated.

[cykVM · Accepted solution]

Hi,

if you read through the 5.5 documentation about coredump partitions/to file: vSphere 5.5 Documentation Center

Set Up a File as Core Dump Location

If the size of your available core dump partition size is insufficient, you can configure ESXi to generate core dump as a file.

Typically, the core dump partition for new installations of ESXi 5.5 and later is 2.5GB. For upgrades from previous releases to ESXi 5.5, the core dump partition is limited to 100MB for ESXi crashes. For many large systems, this limit is not enough space and the core dump log files are truncated. For upgrades with limits on 100MB partitions, during the boot process the system might create a core dump file on a VMFS datastore. If it does not create a core dump file, you can manually create the file.

So this is expected behaviour.

cykVM

[sexconker]

Thanks, that what I had thought.

[cykVM]

For your other question about resizing the coredump partition I presume that does not work. At least accoring to this KB: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=201236...

You may just add another disk with a larger diagnostic partition.

If you want to try resizing anyway I would test that first on a non-production system with e.g. a GParted Linux live boot cd or even maybe works this way: VMware KB: Using the partedUtil command line utility on ESXi and ESX