Problems with a Microserver Gen 10

[Jetro]

Hello everyone,

My name is Davide and after an upgrade to a client I just got for free an HPE Microserver Gen 10, the worst one, with this hardware configuration:

• AMD Opteron™ X3421 APU @ 2100 MHz
• 32GB PC4-19200E 
• 4 x 1TB Segate Ironwolf 7200rpm

I tried installing proxmox in it, but the CPU is really awful, so I'm going to try Unraid and use it for backup purposes. 

I'm quite familiar with enterprise solutions, I'm also using OMV but never used Unraid.

To improve SATA  performances I also installed an LSI MegaRAID PCIe Card which pass directly drives to OS.

 

System boot, array start, everything seem fine until I try to use it: it copy around 3 GB then stops for a while, then restart and while at the end of the transfer Windows throw out "Error 0x8007010B The Directory Name is Invalid".

 

Do you have any ideas on how to solve this?

 

I know even a raspberry might be more powerful than this useless box, but it's free and I'd like to keep it on for something. Just data backup would be fine, for VMs there's the Proxmox cluster

 

Thank you

[JonathanM]

10 minutes ago, Jetro said:

To improve SATA  performances I also installed an LSI MegaRAID PCIe Card which pass directly drives to OS.

Can you try without that card installed?

[Jetro]

I've already tried: same behaviour.

 

Only difference is that without card it tops out near 80 MB/s, with card about 106 MB/s (limited by gigabit connection i guess)

 

Tried also from a linux system and got the same problem: stop after 3 GB

 

Copy of smaller files seem fine. Is there a way to see any log and investigate?

Edited October 28, 2023 by Jetro

[trurl]

Attach diagnostics to your NEXT post in this thread.

[Jetro]

Here it is,

 

Thank you

 

test-microg10-diagnostics-20231028-1414.zip

[Jetro]

Edit: Testing from linux allow me to copy the file without errors, but it's really slow:

- Copy start at Gigabit speed, then stop at around 3 GB with 2 core (random one, real or HT) at 100%

- After a minute or so copy restart at gigabit speed and copy around 2 GB, then stops with 2 core (out of 4) at 100%

- The copy restart after a minute and copy the last 3-400 MB.

 

From Windows behaviour's the same but copy end with that error. I'll do a test also with NFS 

[JorgeB]

Also test with a different filesystem, there's currently a known performance issue with zfs when used on the array (not on pools), but it should still copy without errors, so possible not the only problem.

[Jetro]

Recreated the array with xfs: copy starts at 103-105 MB/s then goes down to 44-50MB/s after 4-5 GB transferred to the end of the transfer, which is successful with xfs.

 

I used 55 GB of video files.

 

Transfer from the same PC to another NAS starts at 112-113 MB/s and end without variations.

[JorgeB]

8 minutes ago, Jetro said:

Recreated the array with xfs: copy starts at 103-105 MB/s then goes down to 44-50MB/s after 4-5 GB transferred to the end of the transfer, which is successful with xfs.

Try enabling turbo write:

https://forums.unraid.net/topic/50397-turbo-write/

[Jetro]

With turbo write I'm now at 103-105 MB/s for the whole transfer.

 

I'm gonna do some other test (i.e. with onboard controller)

[Jetro]

Edit: replaced the 4x1TB Ironwolf with 4 x 4 TB WD Red Pros.

 

Added a 2 TB Samsung 870 EVO for containers/VM

 

Tried to create a new array: parity sync at 250 MB/s, but the old Fujitsu D2507-D11 only permit to use 2.2 TB out of 4 TB drives, so I switched back to integrated Marvell controller and it's syncing at 120 MB/s.

Ordered an LSI 9207-8i controller because all I have is really old controllers which doesn't support 4 TB disks or overkill controllers like Dell Perc HBA330 w/BBU. I'll keep you updated.

 

Plans are to use this as a Plex server (tried transcoding and seem acceptable)

[Jetro]

Edit: installed the Supermicro branded LSI 9207-8i and it works like a charm.

 

With 32GB of RAM it's okay as home media server now, and I'm planning to install some more containers in it.

 

So the big differences was:

- Switching back to btrfs instead of using zfs (which prevent files copy)

- Enabling turbowrite (eliminated the stalls after 3 GB of copy)

- Installing a dedicated PCIe SAS/SATA controller (improved general transfer speed)