I performed an Unclean shutdown...what now?

[HiSoC8Y]

Ok

 

So i came home and found that the unRAID is unresponsive.  I had no choice but to shutdown the VM (unRAID VM) and start it again.

 

Kindly note that I'm not running a parity drive.

 

1. How can I see what went wrong that the array stopped functioning?

2. Now after I started the system, what do I need to do to make sure the system is fine?

[dgaschk]

See here: http://lime-technology.com/forum/index.php?topic=9880.0

[SSD]

Ok

 

So i came home and found that the unRAID is unresponsive.  I had no choice but to shutdown the VM (unRAID VM) and start it again.

 

Kindly note that I'm not running a parity drive.

 

1. How can I see what went wrong that the array stopped functioning?

2. Now after I started the system, what do I need to do to make sure the system is fine?

 

I think that dgaschk covered #1. It looks like he has that answer on speed dial.

 

Starting the array after a dirty shutdown normally kicks off a parity check. But with no parity drive, it will simply mount the drives. RFS is a journaled file system, meaning that writes occur to the journaling area located in the fastest part of the drive, and are then written to their final resting place. If there is an abnormal shutdown, when the disk is mounted, the FS will detect that the drive was not unmounted cleanly and "replay" the journaled transactions. I can say with some confidence that RFS is outstanding at recovering from a dirty shutdown. I assume XFS is similar but have no first hand knowledge. With RFS, you will see the number of replayed transactions logged in the syslog (again, maybe same for XFS, just have never had it happen).

 

Replaying transactions can take some time. One of the reasons it may take a long time is that unRAID hurriedly starts a parity check. So the replay is competing with the parity check which is reading at the highest possible rate. Under these circumstances, a replay can take minutes - even as many as 20 minutes if I remember correctly. Since you have no parity, hence no parity check, I would expect the replay to be much quicker, but there may be a noticeable a delay. Do not get impatient - give the mounting all the time it needs to complete. If after an hour it is still showing "Mounting ...", capture a syslog and post it. My guess is that all should be fine, except files that were literally being written as the server crashed/powered down. Such files will need to be recopied. Using md5 to compare recently copied (or attempted to be copied) files might be in order.