VMDK Corruption - But VM Boots Just Fine

[TheCiscoKid]

Hello Friends,

I'm running ESXi standalone - free (v6.7.0 Update 3) and am encountering a very strange issue. I realized for one of my VMs, I'm getting a read error "Read error from virtual disk (DietPi_VMware_EXSi.vmdk): Input/output error (327689)"

The strange thing is, this VM boots just fine and I can access the VM and read all the data on VM's disk just fine. I also attempted to run a vmkfstools --fix check and I receive the following response "Disk is error free".

Clearly something is amiss as I can't copy or move the -flat VMDK file but otherwise everything looks just fine.

This is a lab server so VM is stored on an NVMe disk which ESXi is also using to boot from. I'm at a loss here. Does anyone have any suggestions or ideas?

Thanks in advance!

[Luciano-Patrão]

Did you try to use vmkfstools -x repair parameter?

You can also try to clone that VM.

[TheCiscoKid]

It's very bizarre. vmkfstools reports that the disk is error free:

[root@ESX:/vmfs/volumes/5b1878bd-6e284c09-04ef-ac1f6b8a8936/VMs/Pihole] vmkfstoo
ls -x repair DietPi_VMware_EXSi.vmdk
Disk is error free

I can't clone or copy the disk either but VM is running just fine. ¯\_(ツ)_/¯

[Patrik Kernstock]

Hi . 

maybe the datastore itself does have some issues, like metadata corruption. Can you check /var/run/log/vmkernel.log on your ESXi host if you can see any log messages mentioning some (metadata) corruption? (e.g. grep -i "corruption" /var/run/log/vmkernel.*)

Regards,
Patrik

[continuum]

Clone the vmdk from inside the VM with ddrescue.
Everything else is a waste of time.

Ulli

[esxiathome]

Thank you Ulli! 

I've spent 2 days trying to recover from vmdk copying issue, then as you recomend did ddrecover to a new virtual drive from within the vm/os (literally ddrecover /dev/sda /dev/sdb ), it solved my issue.

Regards, 

Vitaly