VMware Cloud Community
TheMustache
Contributor
Contributor

VM "space used" differs from guest operating system by double the size

Currently the customer has vmware vsphere 6.7 running on 2 HPE DL360 servers connected via SAS cables to an HPE MSA 1040i with 12 SATA HDDs and 4 SATA SSDs.  They have a VM that currently shows consuming 2 TB on the SSDs which is most of the space available on the datastore that resides on it.   The guest operating system running inside the virtual machine shows approximately 1.2 TB of space used.  The VM disks are thinly provisioned so the the amount of space they host sees should be very close to what the guest OS sees.   This not the case.  The host is reporting that the VM disks require consolidation.  When we try to consolidate the disk the host locks up.  When I try to using Storage vMotion to move the VM to a different data store ( either on the MSA or locally attached ) the system times out.  The same thing happens when the VM is powered off and I try to clone the VM to a different data store.  The VM currently is running business critical accounting software that must be available from 4AM to 7PM so I can't power it off during the day.   My resolutions to the problem have thus far failed with the following errors

My migration failed about 3 hours in with message (redacted)  " PM Alarm 'Migration error' on *** triggered by event 2010 'Cannot migrate *** from ***, Vol0002 to **, datastore1’.

Currently trying the storage vMotion with just one of the drives. There was an event that stated it was having problems accessing one of the disks so I'm avoiding migrating that disk until I free up more space on the datastore.

Any thoughts?

Reply
0 Kudos
1 Reply
StephenMoll
Expert
Expert

Sounds like the guest OS is not unmapping blocks when they are marked as unused. The storage will then not unassign those blocks from the VM that was using them.

I would expect the amount of disk space used by a VM to be higher than the guest reports. I would expect the SSDs to be tier-1 storage and probably RAID-1, so each block of data occupies two blocks on disk.

Reply
0 Kudos