On 9 January 2013 10:11, steve-ALUG@hst.me.uk wrote:
I'm not an expert at this!
More than I am though!
http://www.tcpdump.com/kb/os/linux/starting-and-stopping-raid-arrays.html
says "mdadm --assemble --scan" but I think that'll scan everything, and as you have one that's already working, I think you can do
mdadm --assemble /dev/md1 --scan
If that doesn't work, the above website lists how to find the UUID of the array and restart the array using that. HTH
It didn't quite work, as the array was in use, and also --query didn't work as expected: $sudo mdadm --query /dev/sdf1 /dev/sdf1: is not an md array /dev/sdf1: device 0 in 4 device unknown raid5 array. Use mdadm --examine for more detail.
However, this got me the UUID: $sudo mdadm --examine /dev/sdf1 /dev/sdf1: [...] Array UUID : e6d6be87:9fb1d819:d44509f4:46512dd5
I added that to /etc/mdadm/mdadm.conf: ARRAY /dev/md1 UUID=e6d6be87:9fb1d819:d44509f4:46512dd5
Then, as I still couldn't re-assemble the array (because it was in use), I got the following steps to get me back online: $sudo umount /dev/md1 # Unmount the array $sudo mdadm -S /dev/md1 # Stop the array $sudo mdadm --assemble /dev/md1 --scan # Reassemble the array $sudo mount /dev/md1 # Re-mount the array (based on fstab config)
I assume those four steps will suffice in future now that the device UUID is in mdadm.conf, but it looks like I have to wait another 19 days to find out...
I'm guessing the something that caused the error is the USB disconnect. I think all the rest of the log is just showing how the system is trying to recover. Does anything before give a clue as to why it disconnected? Lose wire, prying fingers? Dunno - grasping at straws!
No lose fingers, cables seem OK, so no clues. I wonder if the cheap ("eBay special") USB3 card may be part of it.
How vulnerable is the array to corruption from this process? I'm not too concerned about occassionally having to go through these steps, but I would worry a lot more about losing the data on the array.
As far as I know write caching is disabled, which would be a pretty bad thing under the circumstances (how do I check?), but otherwise is mdadm/RAID5 and ext4 pretty safe?