Slow(er?) Preclear speeds

[Idaho121]

Hey all,

 

Built a server a few days ago. Precleared 3 8TB drives (shucked WD whites). Started preclear using the plugin. Had initial speeds in the 170-190MB/s range throughout. Dropped a little under 100 by the end. My understanding is that's normal.

 

Just added in 2 more 8 TB disks (shucked WD reds). Started preclear using the plugin. Initial speeds are right under 60MB/s.

 

Some differences (don't know if they're relevant):

1) The whites were fresh from the box. The reds have been in a QNAP box for a while.

2) With my initial setup, I had 1 of the drives on a SATA cable, and each of the others going through a separate Dell HBA, running on a separate cable. These drives, while split between the HBAs, are each now on a controller and cable shared by one of the initial drives.

3) Since the first preclear, I've built array/parity, added a cache drive, and started running Plex. I also turned reconstruct write on.

4) I'm just running the zero part of the preclear (since these disks are well tested).

 

Any ideas on what's slowing me down? And is there any way to fix it up?

 

Thanks!

Edited July 10, 2018 by Idaho121

[JonathanM]

7 hours ago, Idaho121 said:

4) I'm just running the zero part of the preclear (since these disks are well tested).

There is no need to run preclear on known good disks. The only reason remaining for preclear to exist at all is disk testing, the unraid functionality of preclear has long since been built into the OS.

[itimpi]

7 hours ago, Idaho121 said:

Hey all,

 

Built a server a few days ago. Precleared 3 8TB drives (shucked WD whites). Started preclear using the plugin. Had initial speeds in the 170-190MB/s range throughout. Dropped a little under 100 by the end. My understanding is that's normal.

 

Just added in 2 more 8 TB disks (shucked WD reds). Started preclear using the plugin. Initial speeds are right under 60MB/s.

 

Some differences (don't know if they're relevant):

1) The whites were fresh from the box. The reds have been in a QNAP box for a while.

2) With my initial setup, I had 1 of the drives on a SATA cable, and each of the others going through a separate Dell HBA, running on a separate cable. These drives, while split between the HBAs, are each now on a controller and cable shared by one of the initial drives.

3) Since the first preclear, I've built array/parity, added a cache drive, and started running Plex. I also turned reconstruct write on.

4) I'm just running the zero part of the preclear (since these disks are well tested).

 

Any ideas on what's slowing me down? And is there any way to fix it up?

 

Thanks!

 

If you do not have a cache drive the /mnt/cache is a location in RAM!    It would definitely interfere with unRAID mounting a cache disk at that location. It may be being created if you have a docker or VM that is running still pointing to that location.

 

i would suggest turning off docker and VMs for the time being and reboot the system.  You can then get the cache disk set up correctly before re-enabling docker and/or VMs.    If you want dockers to work correctly regardless of whether there is a cache disk present then set them to use /mnt/user/appdata and then make sure the settings for the ‘appdata’ share are correct (probably want the Use Cache’ setting to be ‘Prefer’.

[Idaho121]

4 hours ago, jonathanm said:

There is no need to run preclear on known good disks. The only reason remaining for preclear to exist at all is disk testing, the unraid functionality of preclear has long since been built into the OS.

 

Ah, good to know! I'm being a bit cautious since this is my first unRaid build, but I will skip that in the future and just add to the array, letting unRaid work its magic.

 

I'd still like to figure out what's slowing this down, though, or if this is within expected parameters.

 

4 hours ago, itimpi said:

 

If you do not have a cache drive the /mnt/cache is a location in RAM!    It would definitely interfere with unRAID mounting a cache disk at that location. It may be being created if you have a docker or VM that is running still pointing to that location.

 

i would suggest turning off docker and VMs for the time being and reboot the system.  You can then get the cache disk set up correctly before re-enabling docker and/or VMs.    If you want dockers to work correctly regardless of whether there is a cache disk present then set them to use /mnt/user/appdata and then make sure the settings for the ‘appdata’ share are correct (probably want the Use Cache’ setting to be ‘Prefer’.

 

I do have a cache drive - I added it and followed Spaceinvaderone's video. The only hiccup is that I can't get VMs turned off because it's giving me an error on that option (AMD-x not enabled, or something like that), so the one file associated with that is still on the drive. I had Plex running, but I turned that off, and am still only getting ~63MB/s.

 

Was there something in what I said suggesting that I had set up the cache drive incorrectly? I'm worried now that I messed it up. I know that appdata is on the cache drive, with it being set to Prefer, and that Plex has been writing my metadata to it.

Edited July 10, 2018 by Idaho121

[airbillion]

I am having the same slow preclear issue...

 

This week I precleared 5 x 8tb drives in usb enclosures and the avg speed was 150+mb/s...

 

Today I tried to preclear another drive in its usb enclosure and the preclear started at a little under 60mb/s...so I shucked it and tried again...

 

It started out around 60mb/s and has steadily slowed down under 20mb/s now...

 

Anyone have any ideas???

 

EDIT:

 

I think maybe the newest update 07-09-2018 might have changed something and made the performance terrible...

 

EDIT 2:

 

Rolled back to the previous version 05-03-2018 and the speeds are back up to 200+mb/s via sata...the newest plugin is having performance issues!

Edited July 13, 2018 by airbillion

[garys]

On 7/13/2018 at 3:22 PM, airbillion said:

 

EDIT 2:

 

Rolled back to the previous version 05-03-2018 and the speeds are back up to 200+mb/s via sata...the newest plugin is having performance issues!

 

I'm having slow write performance also.  Can you share the link to the version you rolled back to?

 

Thanks!

 

EDIT:

 

Found the following thread (see first post by Squid on page 85) and followed the steps to roll back to 05-03-2018 version ... fixed!

https://lime-technology.com/forums/topic/54648-preclear-plugin/?page=85

 

Edited July 16, 2018 by garys