ST2000LM015 - Disk busy

[ecarjat]

I've been running FreeNAS for over a year now and despite following a number of test and burnin posts I can't figure out why my Seagate BarraCuda 2To (ST2000LM015) is having bouts of very high usage/high latency when the other disks in the the system don't.

In order to narrow it down, I have build a system with 4 Western Digital 2TB (WD20SPZX) and one Seagate to create a single RAIDZ1 Pool.
In order to avoid TRIM issues on the WD drives trim is disabled on the system (vfs.zfs.trim.enabled = 0)
The pool has an SLOG drive running on a 128G SDD drive.

All of the disks have been tested with short and log SMART tests and show no errors.

The performance is monitored via graphite/influxdb/grafana running in a jail. In the grafs below ada1 is the seagate drive.
There is no scrub operation running during this period.

This is the details of what's happening during that time with ada1

Here is what is happening at the same time with one of the WD drives

Any pointers towards a solution to improve this would be greatly appreciated. I could obviously remove this last Seagate drive and replace it with a WD but I have another 5 of them that I would like to use if possible ...

 

[Alecmascot]

There are all SMR drives and are not recommended for use with ZFS
Maybe the ST2000LM015 does not manage its housekeeping as well as the WD drives.

 

[ecarjat]

Thanks, that would make sense. A quick read on the matter seems to show that there is no work around correct ?

 

[Alecmascot]

If your NAS is mainly read-only then maybe you can live with it, but it will struggle or even fail if you have to do a resilver.
You could test a resilver, by simulating a drive swap, and see what happens. Will need some data to be realistic.

 

[ecarjat]

I have done a few resilvers they take ages but they succeed The main issue is the fact that during that housekeeping the disk is extremely slow which makes the entire NAS not usable even for read only usage.

 

[Apollo]

I have done a few resilvers they take ages but they succeed The main issue is the fact that during that housekeeping the disk is extremely slow which makes the entire NAS not usable even for read only usage.

I experienced Busy issues with my first set of SMR drives (Seagate 8TB Archive drives) and posted on the forum about it, out of the entire pool, only one was showing such behavior. I have tried multiple resilvering without incident. The drive was sent back to Seagate under RMA. The replacement drive is an exos drive I think, but the remaining SMR drive from the pool are still working perfectly.

I believe SMR drives are still strong candidate for supporting ZFS, but as stated by Seagate before, they are intended as Archive drives.
Not so much with Western Digital as it appears.

 

[ecarjat]

I experienced Busy issues with my first set of SMR drives (Seagate 8TB Archive drives) and posted on the forum about it, out of the entire pool, only one was showing such behavior. I have tried multiple resilvering without incident. The drive was sent back to Seagate under RMA. The replacement drive is an exos drive I think, but the remaining SMR drive from the pool are still working perfectly.

I believe SMR drives are still strong candidate for supporting ZFS, but as stated by Seagate before, they are intended as Archive drives.
Not so much with Western Digital as it appears.

In my case I have 5 Seagate 2TB and all of them have busy issues. It seems quite clear that for time being steering clear of SMR drives for NAS purposes is probably best. That's until we have SMR capabilities in ZFS but my rapid scan of the forums was inconclusive on that.