[sheepdog-users] can't umout disk after unplug

Valerio Pachera sirio81 at gmail.com
Fri May 3 16:17:11 CEST 2013

This happens only if the cluster is started using md.
I explain it better:

My first test was with nodes using a single disk.
sheep daemon was run using only a folder:
  sheep /mnt/sheepdog

After md plug a second disk,"ps aux | grep sheep" still shows
  sheep /mnt/sheep
no matter how many other disks we plugged in.

After I added a second disk on a node I stopped the cluster to test how to
start it using md, so
  sheep /mnt/sheepdog,/mnt/sheepdog/obj/,/mnt/sdb1

Cluster strats fine.
Also hot plug works fine
  collie node md unplug /mnt/sdb1
but if I try to unmount sdb1 I get

root at test004:~# umount /mnt/sdb1
umount: /mnt/sdb1: device is busy.
        (In some cases useful info about processes that use
         the device is found by lsof(8) or fuser(1))

root at test004:~# lsof /mnt/sdb1/
sheep   3056 root   17u   REG   8,17  4198968  173

root at test004:~# collie node md info
Id      Size    Use     Path
 0      195 GB  4.9 GB  /mnt/sheepdog/obj

root at test004:~# ps aux | grep sheep
root      3056  0.1  2.3 136456 45740 ?        Sl   15:58   0:01 sheep
/mnt/sheepdog/ /mnt/sheepdog/obj,/mnt/sdb1
root      3057  0.0  0.0  83408   764 ?        Ss   15:58   0:00 sheep
/mnt/sheepdog/ /mnt/sheepdog/obj,/mnt/sdb1
root      3289  0.0  0.0   7848   840 pts/0    S+   16:16   0:00 grep sheep
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wpkg.org/pipermail/sheepdog-users/attachments/20130503/ee6c0152/attachment-0004.html>

More information about the sheepdog-users mailing list