[Weekly btrfs_preempt_reclaim_metadata_space appears to be crashing the cache drives since 6.12.2?]

File system check did find and apparently fix issues.  Will see what happens in 6 days... Thanks!

[Weekly btrfs_preempt_reclaim_metadata_space appears to be crashing the cache drives since 6.12.2?]

Been running this system for years, with a few upgrades here and there.  Been very stable until recently, where every 6+ days I'd notice the Docker service was no longer running. Restarting would fix the issue until another 6+ days would pass.  Finally bothered to look at the diagnostics, and the crashing seems to be related to:
WARNING: CPU: 0 PID: 16956 at fs/btrfs/extent-tree.c:3061 __btrfs_free_extent+0x466/0xc02

...

Workqueue: events_unbound btrfs_preempt_reclaim_metadata_space

...

BTRFS error (device sdh1): unable to find ref byte nr 2845564928 parent 0 root 5  owner 40359587 offset 0
Jul 13 07:04:47 Storage kernel: ------------[ cut here ]------------
Jul 13 07:04:47 Storage kernel: BTRFS: Transaction aborted (error -2)

etc.

 

I see this in 6.12.2 and 6.12.3 logs (attached)

 

I'm going to try a BTRFS file system check next; the 2 SSDs that make up the cache drive are definitely old, but never had an issue until 6.12.2

Thoughts?
 

storage-diagnostics-20230715-1605.zip

storage-diagnostics-20230810-1055.zip

[2.5" HD vs 3.5" HD]

They make 15k 2.5" SAS drives, we have them in some of our Dell servers - but you're going to pay for them

 

http://www.seagate.com/www/en-us/products/enterprise-ssd-hdd/

[Preclear.sh results - Questions about your results? Post them here.]

So it's only the WD20EARS, right? I have some WD20EADS drives as well...I'm assuming these are not affected?

 

You need to look on the label and see if it states Advanced Format Technology.  I recently purchased 2 1 TB EADS drives and they have it.

[Re: preclear_disk.sh - a new utility to burn-in and pre-clear disks for quick add]

Personally, I don't run other tasks while important things like parity creation are occurring.  Can Unraid do it all at the same time?  Sure, but if you do hit a memory limit, or max out your PSU you chance interfering with the parity calculation.  And since parity protection is a primary reason why we use Unraid - I'd let the complete before doing the preclears.

[Re: preclear_disk.sh - a new utility to burn-in and pre-clear disks for quick add]

The advantage to preclearing is it allows you to perform the formatting outside of Unraid so your array remains online for that potentially extended period of time.  In addition you get SMART reports on the drives which will show you whether your drive has potential problems before ever trusting your data to it.