I am running CentOS7 on a virtual instance at Softlayer. There is a block storage device available to the instance via multipath/iscsi
I have successfully had this block device mounted as XFS. I decided I would give ZFS a try. When I manually issue commands to 'mount' this ZFS volume, it works fine.
zpool import -f zfs-data
There are two issues that I am having with ZFS on CentOS7.
- a.) my zpool isn't automatically mounted on boot
- b.) the system hangs on reboot
I suspect the solution for (a) could also fix (b).
Lets have a look at (a) first.
There is this discussion on the github page for zfsonlinux CentOs 7. Not mount/import pool after reboot
It appears that the /etc/hostid file is not created when ZFS is installed via yum. So as suggested on this page I run the following command to create it.
dd if=/dev/urandom of=/etc/hostid bs=4 count=1
Then it suddenly occurred to me that it cannot mount automatically because as the machine is booting, ZFS is being loaded before the iSCSI stuff.
from /var/log/boot.log - first column is the line number
81 Starting Import ZFS pools by cache file...
147 Starting Open-iSCSI...
How do I change the load order?
(b) hangs on reboot
If I import the pool and don't write anything to it, I can reboot fine. However as soon as any data is exchanged on the mount, the reboot/shutdown log contains these lines..
Stopping Import network configuration from initramfs...
[ 550.096199] end_request: I/O error, dev dm-0, sector 5514152
^^ repeated ten times with different numbers.
WARNING: Pool 'zfs-data' has encountered an uncorrectable I/O failure and has been suspended
[...]
Unmounting /zfs-data...
[...]
A stop job is running for /zfs-data
This never finishes and a HARD reboot is required to get back into the instance.
Any insight would be appreciated.
No comments:
Post a Comment