Can snapshots cause a performance issue (as with ESXi)?

[ronejamesdo]

In a production ESXi environment one has to watch how many snapshots one has and how old they are allowed to get because performance for the VM will degrade (and I believe, eventually, for the other VMs, provided the VMs were busy enough to make a lot of changes since the snapshots were taken).

Does Proxmox have this too?

Thanks.

RJD

 

[bbgeek17]

In ESXi world it is called "VM stunning", i.e. when a snapshot initiation on what is effectively a file (vmdk) causes IO to pause in the VM.

Proxmox can suffer from a similar effect. It's most pronounced in a QCOW2/NFS combination.

However, it's not a hypervisor problem in either case. It's a slow storage problem. We have helped customers in both ESXi and PVE worlds solve this particular challenge.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[ronejamesdo]

Thank you.

 

[UdoB]

ZFS is copy-on-write. Always. I have not explicitly tested it now, but I would really expect that the write performance does not suffer from snapshots.

Surprisingly what possibly actually slows a local ZFS is deleting a large amount of snapshots. The multiple operations in the famous Merkle-tree trigger a vast amount of write activity.

Again: not tested --> just a guess.