Pool on USB3 seem unstable (1.3.1-r2)

Developer discussions.

Pool on USB3 seem unstable (1.3.1-r2)

Postby mortengh » Sat Aug 29, 2015 12:44 am

Hello!

I have been using ZFS on Mac for some time now without any problems - using a one-disk pool on internal storage.

OS: MacOSX Yosemite 10.10.5
OpenZFS: 1.3.1-r2

I now wanted to try something fun. I wanted to create an array in an external box.
I bought an Inateck external 2.5" USB3 inclosure which also have a 3 port USB3 hub.
(http://www.inateck.com/portable-2-5-inc ... b-3-0-hub/)

In the inclousure I added an Intel 320 SSD (40GB) and for each of the 3 USB3 ports I inserted an 128GB Sandisk Ultrafit USB stick.
(http://www.sandisk.com/products/usb/drives/ultra-fit3/)

So I created a raidz1 of the 3 usb sticks. Created 3 partitions on the 40GB SSD - one for slog, one for cache and another one for data.
This gives me the following two pools:
Code: Select all
   NAME        STATE     READ WRITE CKSUM
   USBSSD      ONLINE       0     0     0
     disk2s2   ONLINE       0     0     0

   NAME                                            STATE     READ WRITE CKSUM
   USBFlash                                        ONLINE       0     0     0
     raidz1-0                                      ONLINE       0     0     0
       disk3                                       ONLINE       0     0     0
       media-BD1EBFB3-4B7B-134A-A875-269D5C5E38B3  ONLINE       0     0     0
       media-1DB8E8CA-4868-2D4E-B82E-464C0A687DAE  ONLINE       0     0     0
   logs
     disk2s4                                       ONLINE       0     0     0
   cache
     disk2s3                                       ONLINE       0     0     0


The USBFlash pool is set to run compression 'gzip-9' on all filesystems.

But when copying large amounts of data (say 30GB) to the pool the copy stops. It seems like all IO to that pool is suspended.
'zpool status USBFlash' just hangs.

I tried to insert an USB2 hub in front of the external case to force it to slowdown (D-Link DUB-H7).
Then it seems like everything was stable (but somewhat slower of cause).

When IO is suspended the 'mds_stores' process takes up full CPU time for two (of the 4) kernels.
Is there any other information I can give to help fix this issue?

Best regards,
- Morten Green Hermansen, Fanitas
mortengh
 
Posts: 4
Joined: Thu May 22, 2014 5:04 am

Re: Pool on USB3 seem unstable (1.3.1-r2)

Postby MacFormatik » Sun May 01, 2016 7:29 am

This is an assumption of mine:

I belive to have read somewhere...
you can't have Disks that are used as logs or caches on one Pool be used on another Pool since this creates race conditions. Whatever that means.
User avatar
MacFormatik
 
Posts: 2
Joined: Sun May 01, 2016 7:25 am

Re: Pool on USB3 seem unstable (1.3.1-r2)

Postby Brendon » Sun May 01, 2016 1:44 pm

Seems that there are rumors that Apple busted USB 3 in 10.11.4. I'm no expert but I think that there is a chance you are simply overloading the USB bus.
Also weird hardware choices, I have trouble trusting such a configuration.

I'd also suggest updating your software.

Cheers
Brendon
Brendon
 
Posts: 286
Joined: Thu Mar 06, 2014 12:51 pm


Return to OpenZFS on OS X Development

Who is online

Users browsing this forum: No registered users and 15 guests