Fast disks as NVMe M.2 SSD devices are mostly used as cache pool disks. To replace such disks it's required to shutdown the server. After the replacement of such disks, and starting up the server, the array and the cache pool are started automatically with a missing cache pool disk message. The balance starts immediately too. The replacement disk is shown in Unassigned Devices.

 

With that procedure the remaining cache pool disk is written twice: During balance to single cache pool it's read and written. There's no way to stop that process. After successful balance one can stop the array, set the new disk and restart the array. Now both cache pool disks are read and written.

 

My feature request:

 

If Unraid detects a missing cache pool disk, don't start the array/cache pool automatically - as it's done with a missing array disk.

 

If Unraid does not start automatically with a missing cache pool disk, there's a chance to set the replacement disk before starting the cache pool. That way both cache pool devices are read and written only once. The required time for the replacement drops to 50%.

 

Thanks for listening.

 

Edited September 2, 20205 yr by hawihoney

I agree, and there's already a request for that, in the meantime I recommend having array auto-start disable when using pools, so you can always check everything is correct before starting the array.

Saw that request but within the first posts it mentions missing array devices. Later in that thread cache pool devices came into the game. Didn't see that.

 

Did vote for that as well. However the current array behaviour is ok for me. My problem is the cache pool. It took me 4 hours where it could have been 2 hours without automatic start. Not to mention the double stress on the remaining pool device.

 

Edited September 2, 20205 yr by hawihoney

1 hour ago, hawihoney said:

However the current array behaviour is ok for me. My problem is the cache pool.

Allowing array to start without pool would be even worse unless dockers/VMs were also not allowed to start.

Yes, that's what I tried to express. If any disk is missing - array or cache - don't start array and don't start cache pool.

 

The array behaviour is ok for me because it does that already. If an array disk is missing it doesn't start array and doesn't start cache pool. 

 

This should work identical for a missing cache pool disk.

 

The different behaviour is confusing and in case of a failed cache pool disk the current procedure leads to double stress on the remaining disk and the time to recover is doubled as well.