We require that your system have a certain percentage of each VMFS datastore free at all times. The datastore overhead percentage depends on the total capacity of each datastore.
Refer to the following minimum recommended datastore overhead values:
- For datastores with capacity up to 2048 GB, the overhead is 15%.
- For datastores with capacity between 2049 GB and 4096 GB, the overhead is 10%.
- For datastores with capacity greater than 4096 GB, the overhead is 7%.
- However: Where any VMs in the vSphere Cluster are being backed up with VM Recovery and/or replicated with VM Replication (Zerto), the overhead is 15% for all datastores in that cluster. This is regardless whether the backed up or replicated VMs reside on that particular datastore.
If the datastore is subject to multiple criteria listed above, the criterion with the highest percentage takes priority.
We require datastore overhead for the following reasons:
- Some files and features (for example, snapshots, including snapshots made automatically by VM recovery) might consume datastore space rapidly and without your awareness.
- If a datastore runs out of free space, it can cause downtime. A VM’s performance can be negatively impacted when the datastore is critically low on space.
- Certain VMware features, such as snapshots, might no longer function without sufficient datastore space.
- Powered-on VMs have a vSwap file equivalent in size to the amount of vRAM assigned to the VM. For example, a VM with 128 GB of RAM has a 128 GB vSwap file. vSwap files only exist for VMs that are powered on. Powering on a VM or increasing the amount of vRAM in your VM consumes additional space on the datastore.
- A VM might not power on due to insufficient space on the datastore.
- All VM log files are stored within the VM container folder on the datastore. A typical log file is a few MB in size; however, log files can grow to a few GB in size depending on VM activity and DRS migrations.
- VM Recovery requires additional space because the VM takes snapshots frequently, potentially on multiple VMs at the same time.
Updated 8 months ago