Slow performance after add disk to a stripe

[andrema2]

I had a stripe with 3 disks and I added a 4th disk using the GUI. I follow the procedure showed in the Wiki.

Took some time but the volume shows the total space as of the 4 disks and all data still there. The issue now is that copying files is really slow, like on kb/s from time to time.

I check the status of the volume via ZFS commands and it says that it is ok with no errors.

I was wondering if some rebuild process is going on in the background and while it is happening the performance will be bad.

How can I check it ? Has anyone done that before ?

 

[mrwillz]

I am a Solaris admin and have worked with ZFS from the Solaris side of things pretty extensively. I would run "zpool status" to verify that all disk are online and the resilvering is done. You can run "zpool iostat <interval> <count> ex. "zpool iostat 1 100" would run zpool iostat every second 100 times. Do that command while you are copying data to your zpool and see what the write speed is. You can run "zpool scrub" to verify your data, but keep in mind that this will take a while. Also verify that you are running the latest version of zfs by running "zpool upgrade -v". I am not sure of your system setup but zfs is very memory intensive and I have noticed slow writes with less then 6GB of RAM. And you may want to take a trick from windows and reboot freenas.

 

[matthewowen01]

What disks did you have and what did you add. How are the connected to your computer.

The output of 'zpool status' would also help.

 

[andrema2]

Hi All,

I'm running 8.0.1 beta 4 64 bits.

The ZFS version is 15.

The configuration was 3x1.5tb 7.200 disks in stripe mode, 4k forced and ZFS. My machine has only 2.5 gb RAM, but using dd command I was able to write at 86MB/S and read at 94MB/S. When I was copying via LAN I was getting 70mb/s.

The problem started after I had copied almost 3tb of data and I decided to add a 4th disk. Doing a dd copy the performance when down to 2kb/s. The 4th disk is a 320gb 7.200. All of them are SATA.

The zpool status says

Code:

%zpool status
  pool: NAS
 state: ONLINE
 scrub: none requested
config:

	NAME                                          STATE     READ WRITE CKSUM
	NAS                                           ONLINE       0     0     0
	  gptid/6158960e-c066-11e0-8d48-001ec91e029e  ONLINE       0     0     0
	  gptid/6214e75a-c066-11e0-8d48-001ec91e029e  ONLINE       0     0     0
	  gptid/62d1f9e1-c066-11e0-8d48-001ec91e029e  ONLINE       0     0     0
	  gptid/48cf5144-c101-11e0-aad5-001ec91e029e  ONLINE       0     0     0

errors: No known data errors

Something went wrong when I added the disk. I can see also that the permission of the NFS is also wrong and when I tried to correct via GUI, there a message saying some error occurred and the change is not implemented.
Any ideas ?

 

[matthewowen01]

Zfs hates to be filled and perfoms badly when it is.

Your original drives are still filled. Reading and writing will still be slow until you balance the files out. Only adding the 320 GB shot you in the foot. You need much more space than that. 50% free is best for good performance.