2

Three HPE dl380 GEN9 server

HBA MODE

On each server one physical disk for os and 6 physical disk for storage spaces direct

windows server 2016 and enable storage spaces direct three way mirror on it

recently three of my physical disks retired so i just replaced them with new physical disk but now i have the following error

transient error on three of my physical disks.

the result of get-physicaldiskenter image description here

the result of get-virtualdiskenter image description here

what should i do? i have done everything by the book but no result. enter image description here

enter image description here enter image description here

Improve this question
  • You're probably aware of this, but the Samsung 850 series are 'consumer grade' SSDs and as such not recommended by either Microsoft or HP for this use case. Particularly note that non-HP hard drives are known to behave unpredictably in proliant servers. – BlueCompute Jun 1 '17 at 9:02
  • yes i'm aware of them being consumer grade – Amir Jalali Jun 1 '17 at 9:58
5

Hello from what I am seeing, I assume that your original storage space is corrupt. I would recommend manual rebuilding.

By the way, because of these hiccups with S2D, we don't use it

Improve this answer
  • how do i do manual rebuild? and what do u use instead of s2d? – Amir Jalali May 31 '17 at 17:27
  • does manual rebuild issue any interruption? – Amir Jalali May 31 '17 at 17:58
  • If you're stuck now you won't rebuild... Get your data to backup destination, devastate your S2D pool, create new one from scratch, create new virtual disk from pool occupying your expected capacity and get your data back from backup destination. – BaronSamedi1958 Jun 6 '17 at 13:56
5

Does manual rebuild issue any interruption

You can interrupt manual rebuild, and continue after an hour, for instance. However I can't be sure about S2D proper functioning if you would stop the rebuild prosses.

And what do u use instead of s2d?

If you are building a compact 2-node HA cluster, there are better solutions available on market to ensure storage availability, for example, HPE Store Virtual VSA and Starwind VSAN. They are designed to handle multiple drive failures by utilizing replication along with hardware RAIDs for local redundancy. Both of these solutions in 2-node configurations allow you to lose a whole node plus a pair of drives on the other node and still be up and running.

Improve this answer
-3

Transient error could be a few things. Are these the old failed disks or the new one?

If you have replaced one and one. Added the new disk retired the old and let it rebuild inbetween replacing of disks?

Run a get-physicaldisk

Also turn off all vms and take the virtual disks offline and run repair-clusters2d -verbose This will repair s2d setup.

Is your virtual disks onlie bdw? Run a get-virtualdisk

Also check out my blog for som s2d troubleshooting. Jtpedersen.com

Regards Jan-Tore

Improve this answer
  • these are new one. i just replaced the retired physical disk with new one. i didnt do any manual rebuild – Amir Jalali Jun 1 '17 at 8:16
  • I added the result of get-virtualdisk and get-physicaldisk – Amir Jalali Jun 1 '17 at 8:24
  • Have you tried the repair-clusters2d command? Id recomend stopping all vm's that are running when running it. You can use -node to just do it on 1 node. If that does not help, id recomend getting another new disk. And replace only 1 disk untill it's finished rebuilding. Then take the disk you took out. Put it into another computer and run the clear-s2dconfig.ps1 script to clear any S2D config on it. You can get it here tinyurl.com/zndvqd7 Make sure to retire the old disk as well. So when you remove it it will have lost communication. – Jan-Tore Pedersen Jun 2 '17 at 2:12
  • Just noticed you are running samsung 850 drives. These do not have power loss protection. This might be the cause of your issues. Id recomend getting some other drives like the samsung sm863 or pm863. Sm863 for better write performance. To use them replace one and one disk. Takes a while but. – Jan-Tore Pedersen Jun 2 '17 at 9:11
  • Also you only have 1 virtual disk for hyper converged? Recomended is q virtual disk pr node. – Jan-Tore Pedersen Jun 2 '17 at 9:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.