VM Memory Snapshots takes longer time on vSAN
Article ID: 326722
Updated On:
Products
Issue/Introduction
Symptoms:
When taking a memory snapshot of a VM on vSAN, the operation can take longer than it takes when VM is on VMFS Datastore.
Environment
VMware vSAN 6.0.x
VMware vSAN 7.0.x
VMware vSAN 8.0.x
Cause
The reason memory snapshot of the VM on vSAN takes longer time is, vSAN doesn't implement the same per-handle IO scheduling that is available in VMFS and NFS.
Since IO shares can be set on the snapshot's memory file handle (.vmem), we instead rate limit at user level (the IO issuing thread is sleeping between IOs).
The way checkpoint memory I/O works is that it tries to use low priority I/O by creating a file handle to the vmem file and setting a low priority scheduling policy on it. This is done to make the memory saving I/O have less of an effect on actual guest I/O.
However, vSAN does not support this sort of I/O scheduling, so the checkpoint process falls back to throttling by sleeping between writes, again to try and not impact guest I/O.
Resolution
This is essentially by Design and applies to all releases of vSAN.
If further assistance needed on this issue, please contact Broadcom Support.
Additional Information
Additional info: Overview of virtual machine snapshots in vSphere
Feedback
