Proxmox VE - Slow migration

[Andrei_W4TL]

Hello,

I have seen there are a lot of threads on this topic, but none of them have helped me solve this. I have tried the often mentioned insecure option:



But I'm still getting only a weirdly ~300Mbit (30MB/s) copy speed. I'm migrating from ESXi, both machines are connected over Gigabit Ethernet

This is the Proxmox VE:



and this is the ESXi Host:


What may I be missing? Do I need to restart the Proxmox VE Host, so the insecure option gets applied?

I need to transfer a whopping 5TB, so speed is crucial for this migration.

Thanks in advance.

 

[tcabernoch]

I encountered similar limitations. Speeds sound about the same.

I could very well be wrong about this, but I think ESXi is protecting its ability to provide good service to all the other guests.
It's going to make some quality of service judgments and only deliver enough speed to the API calls from Proxmox to where it doesn't impact things.
ESXi may be throttling migrations.

That's a guess.

 

[Azunai333]

Just to be sure: Did you delete all snapshots on the ESXi side for the VM you want to migrate?

 

[Andrei_W4TL]

I encountered similar limitations. Speeds sound about the same.

I could very well be wrong about this, but I think ESXi is protecting its ability to provide good service to all the other guests.
It's going to make some quality of service judgments and only deliver enough speed to the API calls from Proxmox to where it doesn't impact things.
ESXi may be throttling migrations.

That's a guess.

It is the only (remaining) guest on the Host, but yeah that could be a case.

Just to be sure: Did you delete all snapshots on the ESXi side for the VM you want to migrate?

We did not use any snapshots on the ESXi side, but why would it make a difference regarding the speed? It would make a difference in the total data needed to be transferred I guess?

Also for the sake of giving out all the details, the ESXi Version is 6.5.0 and the host is a quite old HP ProLiant DL380e Gen8

Screenshot of the NIC in ESXi:


I have read about changing MTU size, didn't tr that out yet, the ESXi has 1500 set. Maybe I'll give it a try after office hours with a MTU size of 9000.

 

[sw-omit]

If you're running with snapshots, it needs to first calculate each total block, based on the base-file and all snapshots plus memory, so it's about 5-10 times as slow (no kidding, had this happen during my migration, and yes there was only 1 snapshot without memory) as without any snapshots.

Also, as I see that the last VM is the fileserver, if your "issue" is that it is too slow to do during a maintenance-window + you have the storage on a separate drive from the OS, maybe take a look at this thread here for an alternative route to consider/take (in which it doesn't really matter all that much how slow it is)
https://forum.proxmox.com/threads/c...roxmox-with-multiple-hdds.150986/#post-682983

 

[tcabernoch]

I second the idea of breaking up the migration into OS and Storage.

Or perhaps its time to move beyond this VM.
Setup a new VM in the new environment.
Use rsync to move your data.
Do a hard cutover one day.