3

I'm working with a Supermicro SMC2208 raid controller.

To increase the capacity of a large RAID array, I decided to add two additional disks using storcli. storcli /c0/v0 start migrate type=raid6 option=add drives=2:12,2:14

The problem is, I did not expect that this would disable the "read ahead" read policy, the write-back cache and the pdcache, which all together slow down the system considerably.

So the first question is already: Is this cache turnoff expected after the above command? I never saw this mentioned.

Is there a way to stop the migration/expansion process, to delay it, pause it or anything? It would take very long to complete and with the read ahead cache off, it becomes too slow. At least it would be good to pause it during some hours. I'm aware of the migraterate parameter, but even when that is set to 1, the performance is slow.

I tried removing these disks again by storcli /c0/v0 start migrate type=raid6 option=remove drives=2:12,2:14 but it won't stop the current migration:

-------------------------------------------------------------------------
VD Operation Status ErrCd ErrMsg                                         
-------------------------------------------------------------------------
 0 MIGRATE   Failed    50 device state doesn't support requested command 
-------------------------------------------------------------------------

Pausing also doesn't seem to be an option storcli /c0/v0 pause migrate

-----------------------------------------------------
VD Operation Status ErrCd ErrMsg                     
-----------------------------------------------------
 0 Migrate   Failed   255 Pause Migrate not possible 
-----------------------------------------------------

And during the migration/reconstruction it seems impossible to use the read ahead cache policy: storcli /c0/v0 set rdcache=ra

----------------------------------------------------------
VD Property Value Status ErrMsg                     ErrCd 
----------------------------------------------------------
 0 rdCache  RA    Failed reconstruction in progress    29 
----------------------------------------------------------

Unfortunately the full rebuild would take 12 days according to the controller's estimate and it would be very annoying to keep the server this slow for that long.

2
  • 1
    I think you've covered all the obvious bits. You're probably screwed. Aug 27, 2019 at 19:01
  • @MichaelHampton That has to be the worst thing one can be told on SO
    – Jarmund
    Jul 10, 2020 at 22:27

1 Answer 1

3

I'm aware of the migraterate parameter, but even when that is set to 1, the performance is slow.

Changes to migraterate (aka Reconstruction Rate) only take effect after reboot. Rebooting is safe, the RAID controller will continue where it left off (but of course you always need to have backups in place, just in case).

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.