Datastore out of space due to snapshots on thick provisioned vmdk disks

Products

• VMware

Issue/Introduction

Some environments use thick provisioned disks to guarantee space is available for VM's to write data. In this configuration snapshot disks have to take up free space in the datastore that is outside of the virtual machines configured limits. This creates a scenario where snapshots can consume all the free space in the datastore and depending on how the utilization is reported you may not be notified that your datastore is running out of space. 

 

Cause

This is caused by the way VM usage is reported. When you thick provision a VM to be 100GB it now takes up 100GB on the datastore. Now you put your data on the VM and use 20GB of the space. It still takes up 100GB on the datastore. Next you take a snapshot and after the snapshot write 5GB of additional data. That 5GB of data is held in a snapshot disk that is outside of your 100GB vmdk disk. Look at the space consumption of your VM in the guest OS and it is using 25GB, through the UI it is 100GB because it is thick provisioned and that is its hard limit. Take a look at your VM's space consumption from inside the datastore by adding the file sizes together and now you are actually using 105GB. You can see how this problem will scale with time and utilization. 

Resolution

The situation as outlined above: Your datastore is full, and that was caused by snapshots taken against thick provisioned vmdk disks. The options at this point are to expand the datastore, or shut down the VM and delete the snapshots, or in the case of orphaned snapshots not in the UI inventory, perform a snapshot disk consolidation. Even with the datastore full a consolidation will still run as long as there is sufficient free space inside the thick provisioned disks.