Dell ME4024 and Linux KVM/Qemu

Hi,

Did you do physical volume data alignment? page 26 https://dell.to/3345nFy

Do you have any issues with corruption if you don’t make a thin provisioned lun? Are you using xfs or ext4? Let us know if you have any other questions.

Hi,

Yes I create the LVM with —dataalignment=1M. 
With the Linux VMs I use ext4. 
The volume doesn't seem to have corruption when I don’t use the discard/unmap option, but the performance becomes much worse. I don’t know if I can create a LUN which is not thin provisioned from the ME4 interface when I use virtual storage. 

Thank you for the details.

Would you confirm the which model drives you're using?

Let us know. 

Thanks.

Toshiba AL15SEB18EQY and 

Seagate DL2400MM0159 mixed with ADAPT Raid. 

2* Toshiba kpm5xrug960g as SSD cache. 

Is the drive firmware up to date? https://dell.to/3xChgjQ

https://dell.to/3aYJmMA

Yes I think so, EF06 firmware for Toshiba HDDs, ST58 for Seagate HDDs and B01C for Toshiba SDDs.

The hard drives do look current on their firmware.

What SAS HBA do you use? Is it current on firmware and driver?

Is the EMM  up to date?

https://www.dell.com/support/home/en-us/product-support/product/powervault-me4024/drivers

I use the Dell 405-AAEV, the firmware is 16.17.01.00. On Linux the driver is mpt3sas version 36.100.00.00. I also tried with version 35.101.00.00.

EMM has firmware 526E.

Thanks, Is the device under warranty? This may be something we need to have phone support look at or escalated.

Yes, the device is very recent and under warranty, I called Dell Support here in France, they told me that it looks like a configuration issue, that if I want help, I must subscribe to Dell ProDeploy, and they recommended to look at the official documentation. It wasn't very helpful because I already read it... It is strange to get corruption with rather standard settings.

I would be grateful if there is something you can do about it.

Can you private message me the service tag? I an look into it but I may not be able to overrule what they said.

I did new testing today, I removed the ADAPT RAID and created a RAID6 storage with only the 1.8 TB Toshiba drives and one SSD cache, the problem still happened. I think that it is a software or firmware issue because the drives seem good. I compared with local storage, where the issue didn't happen.

Also I tested with Linear Storage and the corruption did not happen. But without virtual storage I cannot use SSD cache, snapshots etc.

I am still looking into it but I haven’t been able to find any known issues or any reason why this is happening. What connection method is from the server to the storage?

I use SAS cables. I configured multipath.conf by copy pasting from the "Dell ME4 and Linux" document. When I do multipath -ll, the multipaths are showing correctly. I create a new PV with pvcreate dataalignement=1m 

I create new VG. Then finally I create a new LV for KVM/Qemu.

I add the storage as raw, discard=unmap, detect_zeroes=unmap (otherwise TRIM doesn't seem to be working).

The storage gets detected by the VM. I copy dummy files with dd to fill the virtual disk. I delete the files, recreate them. After 1 or 2 iteration the disk gets corrupt. If I lower the RAM the corruption seem to occur even faster. The server uses ECC RAM with no error detected...

The corruption occurs with Ubuntu, Debian and Proxmox. I am currently running tests with Windows Server and Hyper-V. It seems much slower. So maybe the actual speed with Linux is wrong.