Slow parity check - reviewed with DiskSpeednext steps?

[deanpelton]

Hi All,

 

I had a slow parity check last week (around 24 hours for 4tb. 100% increase in time).
Also had 2 days where mover broke trying to move on to my SMR ST4000DM004 drive (#2 of 3 SMR drives). It was running at under 500kb/s for almost 12 hours before I killed it.

I suspect the cause of both my issues is displayed in the graph below, something seems to be not quite right with Disk2.

 

It's currently about 65% full, I have run many extended SMART tests on it, and the DiskSpeed check for it is below in black compared to the other SMR drives:

 

 

This happened once on Desk1 before when I was stupidly moving stuff between drives and wrote a lot more than the capacity to the drives, it was resolved by:

• Copying data from Drive to an external drive.
• Pulling the drive.
• Running a clean (zeroing entire drive)
• Copying data back to drive consecutively (this ran at a spiritedly 150mb/s, which is how I knew it fixed the issue)
• Removing parity, adding the drive, and doing a parity rebuild.

Note I did it this way because it was the easiest way to write the data consecutively onto the drive tracks, as far as I understand a parity rebuild of the drive would have returned the data to the exact same state (track by track).

Unraid post referring to this issue:
https://forums.unraid.net/topic/102913-solved-unbearably-slow-disk-to-disk-write-speeds-1mbs/

 

 

Questions:

1. Do we think thing this drive is dead? Has anyone seen a drive do this before?
2. Is it worth doing what I did above again to recover it?
3. Does anyone one run SMR drives with this issue?

 

Thanks 🙂

adam-htpc-diagnostics-20220316-1515.zip

Edited March 16, 2022 by deanpelton

[JorgeB]

It is a disk problem, it can also occur with CMR drives, full disk write might help, if if does an Unraid rebuild will do it, that test has nothing to do with contiguous data, and SMR by itself will not slow down reads, on a normally working disk. 

[deanpelton]

Thanks Jorge.

So you suspect the drive is failing? I wish there was something in the SMART data that actually could tell me this.

I know SMR should read the same speed as a CMR drive which is why I suspected something was off.

 

What do you mean by a full disk write?

[trurl]

13 minutes ago, deanpelton said:

What do you mean by a full disk write?

Rebuilding the disk from parity will write the full disk.

 

13 hours ago, JorgeB said:

Unraid rebuild will do it

 

[deanpelton]

Okay I rebuilt onto an IronWolf I had hot spare.

Precleared the drive (which went very fast!) then tried to write the 3TB back to it.

Got 1tb in then write speed tanked. RMA'ing the drive now.