Migrating from iSCSI to NFS

[alf.riga]

Hi,

i have a proxmox cluster of 2 nodes, connected via a 10Gb network to a QNAP Nas. Currently, all my vm disk are stored on that NAS using iSCSI.

I would like to move from iSCSI to NFS. So, i created the NFS datastore on proxmox and all just works. But, regarding the migration, i currently using the "Move Disk" functionality on each VM and its a very tedious and long task. Two question:


1) It's normal that the moving task (from iscsi to nfs on the same nas) its very slow? I'm talking about a 300mb/s speed against the 10Gb link i have.

2) There is a way to perform a bulk migration instead of moving each disk individually?


Thank you

 

[bbgeek17]

1) It's normal that the moving task (from iscsi to nfs on the same nas) its very slow? I'm talking about a 300mb/s speed against the 10Gb link i have.

Both the NFS and iSCSI are located on the same device. They share the same network interface. You need to read the raw iscsi data, convert it, write it back to the same storage in different (more chatty protocol). The speed of migration will depend on many things, in addition to network. For example: qnap cpu, disk speed, datastore organization (compression, encryption, deduplication), your CPU availability on hypervisor side.

2) There is a way to perform a bulk migration instead of moving each disk individually?

Use CLI (qm) and script it.

---

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

 

[alf.riga]

Thank you for support.
Could there be any performance improvement if I ran the "move disk" task with the virtual machine turned off?

 

[bbgeek17]

Could there be any performance improvement if I ran the "move disk" task with the virtual machine turned off?

Its impossible to say what, if anything, could help, because we dont know what the bottleneck is.

I believe Qnap has a few helpful performance graphs, did you check them out? Look at storage IO RT (response time). The one shown on the storage side is what storage thinks it takes to put the IO on the wire, ie before its transmitted and processed by client. If that time is in high millisecond, or even seconds, then the storage is just slow

---

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