Limited benefits versus downsides of btrfs "RAID-1" mirrored cache pools?

[dev_guy]

I've seen several posts where people have had a cache pool drive fail that was in a "RAID-1" two slot mirrored configuration and have still had issues including having to start over with a new cache pool. I've searched but I can't find anywhere that documents how to replace a failed cache drive in a 2 drive pool and have it automatically rebuild the mirrored pair? This doesn't seem to be a GUI feature of Unraid?

 

I do understand one cache drive can fail and you hopefully still have your data on the other drive. But it's my understanding you still have to manually repair the cache? Further many argue btrfs has many performance and stability issues (I'm happy to provide references). It's also less natively compatible for recovering data from drives outside of Unraid compared to a single XFS drive.

 

I tried to mount a healthy btrfs mirrored Unraid cache drive in Linux and it wouldn't mount. There may be some magic obscure incantations to mount such a drive but it wasn't obvious to me. Linux insisted it was corrupt. It may have to do with Unraid not using the official RAID-1 format for cache pools?

 

So all of the above leads me to wonder if many systems wouldn't be better off with single drive XFS cache pools that are regularly backed up to the array (or elsewhere)? I get that if you're using the cache for critical applications where you can't afford to lose even a few hours of data the downsides of btrfs, non-standard RAID, and dual drives might be worth it. But even then if a cache drive fails it seems you still have a significant problem on your hands?

 

Am I missing something here or is the above more or less correct?

[JorgeB]

10 hours ago, dev_guy said:

how to replace a failed cache drive in a 2 drive pool and have it automatically rebuild the mirrored pair?

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=480419

 

10 hours ago, dev_guy said:

I do understand one cache drive can fail and you hopefully still have your data on the other drive. But it's my understanding you still have to manually repair the cache?

If it's a device failure you just need to replace it.

 

10 hours ago, dev_guy said:

I tried to mount a healthy btrfs mirrored Unraid cache drive in Linux and it wouldn't mount.

You must use the 'degraded' mount option:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=543490

 

 

 

[dev_guy]

11 hours ago, JorgeB said:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=480419

 

If it's a device failure you just need to replace it.

 

You must use the 'degraded' mount option:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=543490

 

Thanks for the answers! I found info on replacing/upgrading a single drive cache, that's obvious enough, but nothing promising on rebuilding a RAID1 cache with a failed drive. In fact, the advice I did find in this forum suggested moving everything off the remaining working drive to the array, replacing the failed drive, formatting the modfied RAID1 cache pair, and moving the data back. I don't believe this is in the Unraid Manual but probably should be as it's not at all obvious from the GUI what to do. And thanks also for the magic command line option to get a Btrfs RAID1 drive to mount outside of Unraid.

 

I still need to test the performance of Btrfs vs XFS on NVMe SSDs for Docker/VM use. I've seen benchmarks, like the one attached, showing Btrfs taking around twice as long as XFS for the same file operations. I don't know if that's the case with the current versions and within Unraid but is certainly something to consider. Some of my use cases include full text searches, de-duplication, and other file intensive operations running locally on the server.

[JorgeB]

10 hours ago, dev_guy said:

but nothing promising on rebuilding a RAID1 cache with a failed drive.

If the failed drive is still assigned the procedure is the same as replacing, if it's not still assigned, i.e., old device is completely dead, you just add a new one, no need to reformat. 

[Alex43]

Not to thread hijack, but is it possible in this scenario to replace it with a drive of slightly different size?  I have a 240GB and a 250GB SSD and I was thinking of putting them in as RAID1 cache drives.  Will that work or is some extra adjustment required?

[itimpi]

If you have 2 disks in a RAID1 pool with them being different sizes this will work with the available space being that of the smaller drive.