Slow transfert in OpenmediaVault

O

Oro_Orion

Member
Sep 5, 2020
20
0
6
47
Hi, I recently tried to migrate my VM from esxi to proxmox. So far I am impress of Proxmox simplicity.

I am facing a strange issue by now....with my VM openmediavault.
I have HDD 2Tb Raid HW by server.
Passed to OMV direct from host. Basically, everything just went well.
But here is the issue, I can tranfer large folder to OMV OS virtual drive with a high transfert rate, but when transfering to the direct drive for data storage, speed is in between 1/10 with lots of errors.

What I could check now :/
 
I'd suggest to check Firmware status, eventually the drives.
Where do you see the errors?
In the VM? If yes, what does the host say?
How did you do the "direct pass through"?
Is it the lun? The raid controller itself?
 
tburger said:
I'd suggest to check Firmware status, eventually the drives.
Where do you see the errors?
In the VM? If yes, what does the host say?
How did you do the "direct pass through"?
Is it the lun? The raid controller itself?
Click to expand...

All firmware are up to date, both raid drives are almost new they were running fine under esxi.
Errors in my SFTP transfert client. I tried SMB too and result in a slow transfert and error of writing files sometimes.

I did that procedure to pass drives "https://pve.proxmox.com/wiki/Physical_disk_to_kvm" to VM.
Its the logical drive that containt the 2 HDD by the raid controller.
 
Damn I have messed that question up... I meant virtual disk type and controller (of your VM).
 
What kind of speed do you get from both disk types? Obviously there must be a difference...
Is the cache enabled on the HW raid controller?
I have had no bad experience with passing through raw disks, but never used a RAID device for it, just individual disks...
 
Seems to be fixed so far. I tried somes others scsi HDD controller and the default ( LSI 53C895A ) is the better ....
 
Oro_Orion said:
Seems to be fixed so far. I tried somes others scsi HDD controller and the default ( LSI 53C895A ) is the better ....
Click to expand...

Hmm, no luck, its not fixed, its transfering a large file at 113Mb/sec, and after 40% ( approx), speed drop near zero. Transfert failed, and target guest OS is running very slow, I have to reboot it !!
 
I guess the guest runs your io into memory as filesystem cache and once it needs to destage it drops out.
What version of openmediavault are you using?
Any chance to test a different is and see if it behaves the same?
 
tburger said:
I guess the guest runs your io into memory as filesystem cache and once it needs to destage it drops out.
What version of openmediavault are you using?
Any chance to test a different is and see if it behaves the same?
Click to expand...

Good idea to test it, but like a told its working well on a drive on regular storage. For the memory, while transfering, I monitor CPU and Mem usage, and nothing special. For me too it look like a cache that dont do its job somewhere. I tried all HDD cache mode and nothing changed.

But I will give it a try to test OMV5 instead of OMV4.
 
What you perhaps could try / check are the cache settings in the controller itself. If it destages earlier then eventually the system behaves more stabile/graceful.
I know that on esxi it worked for you, but the SCSI controllers there (also the PV ones) have a limit in queue depth. Can't tell what these values are on KVM but maybe the is just overwhelms the storage? I have seen the erratic behaviour on some SANs as well, whereas cache destage led to IO timeout which finally ended up in errors, because the controller was refusing/rejecting additional IO.
The p410 is definetly no small controller or toy but on the other hand it is also not very powerful (if I don't mix things up).
Your disks are not very fast either so maybe (MAYBE) this just comes from an overloaded storage subsystem...
 
You might be right, but I will be fix after some tests with an others OS, from an other native distro. Basically esxi is a special flavor of linux, so lets try and see...if issue still remains, I might have no choice to revert back to esxi.
 
Oro_Orion said:
Basically esxi is a special flavor of linux, so lets try and see...
Click to expand...
Well, that is a total misconception ;)
Esx never was Linux.
It used RHEL based Systems for the service console but that is long gone.
Esxi was a complete rewrite which used some open source tools (like busybox, dropbear, etc.) But it is no Linux at all...
 
Photon ( esxi ) is a linux container, thats what doc says as far as I know. But I am not a vmware expert anyway.

So I did few tests, and I got same result with ubuntu as guest and Proxmox as host.
I am redoing Esxi test, to clear HDD, but if test is positive, proxmox or guest might be the problem
 
I found it !!!!
My server is an HP DL 360 G7 with p410i raid controller. My 2Tb HDD are Sata. The only way HP can detect the drive is with a raid array whatever if its raid 0 or 0+1 with 1 or 2 drives in the array. I only encouter my issue with a sata drives as destination ( my Nas for storage ). If i try with a Sas drive....no problems !!!

So next test, put a sata Pcie controller in the server and plug the drives on it and redo the test !!!
 
You must log in or register to reply here.
Top Bottom