The new coredump file feature in vSphere ESXi 5.5 creates a file to extend the coredump partition. This usually happens when you upgrade from a previous version to ESXi 5.5, or you install ESXi to a USB drive or SD card. This coredump file is created on a random VMFS datastoore by using a smart selection algorithm. This random placement causes different kind of problems, for example when you want to delete a datastore. Another problem might occur in an EMC VPLEX deployment where you want to failover VMs to the secondary site in a PDL situation (VMkernel.Boot.terminateVMOnPDL=1 / Disk.AutoremoveOnPDL=0). Datastores that had a coredump file configured can't get mounted back when the volume leaves the PDL state after a site failure has been fixed.
To workaround this issue you might want to permanently disable the coredump file. You can't just disable it, as it gets created automatically after a reboot.
Prior to disable local coredump files you should consider to have a network coredump collector installed. A coredump collector software is included on every vCenter Server installation medium and can be easily installed beside the vCenter.
1. Configure each ESXi Host to use a network coredump collector
~ # esxcli system coredump network set --interface-name vmkX --server-ipv4 X.X.X.X --server-port 6500 ~ # esxcli system coredump network set --enable true
2. Disable and delete the currently configured coredump file
~ # esxcli system coredump file set --enable false ~ # esxcli system coredump file remove --force
3. Disable VMkernel.Boot.autoCreateDumpFile
This is an Advanced Setting on the ESXi Host (Configuration > Software > Advanced Settings)
You can also use PowerCLI to disable autocreate of a dumpfile:
Get-vmhost esx01.virten.lab | get-AdvancedSetting -name vmkernel.boot.autocreatedumpfile |Set-AdvancedSetting -Value 0
4. Reboot the Host
5. Verify that the corefile is still disabled after reboot
~ # esxcli system coredump file get Active: Configured:
Pingback: Cannot remove datastore * because file system is busy. | Virten.net
I did what you described in p.2 "Disable and delete the currently configured coredump file" and p.3 "Disable VMkernel.Boot.autoCreateDumpFile", but the coredump files still get auto-created after reboot. Is redirecting them to the network coredump collector part of the trick ? Or you were able to disable coredump files without setting up the fake networ collector ?
Pingback: Unmount Datastore nicht möglich
Any idea where this setting resides in the profile ?
I'm not sure if the file is configurable in the host profile. Only the partition:
Coredump Partition Settings > Fixed Coredump Partition Policy
What happens if the ESXi Core Dump Collector (vCenter VM) is running on the host that crashed? Will you still get a coredump?
Is there a way to do it without esxcli commmands? Issue is every time I try to start mgmt services on a specific ESXi 6.0u2 host, hostd service crashes writing a coredump that takes up 100% of the partition on the SD card. Hostd will not stay started, not sure if it has to do with the coredump filling up that partition or the backtrace in the hostd.log referring to vsphere replication.
Since I can't get hostd to stay started I cannot run any esxcli commands. I was hoping to find a txt file in /etc or something.
Is it only useful to configure a network coredump collector if you're hosts are booting off SAN? Otherwise, wouldn't you just configure the coredump file location to be set to the local disk of some sort?
Does this issue (with VPLEX datastores) still persist in vSphere 6.x?