> Of course, I have a half dozen files that I would like to recover from the > crashed raid set. Before I switched to the kernel based method of raid > recognition, I seem to recall a ckraid --force-sync option, or some such, > that allowed me to convince the raid set that it was a valid raid set. I > can't find the equivalent option in the raidtools-.90 that is running on > this machine. Is there a way to convince a raid set that it's still > synced to itself? One uses: mkraid -force /dev/mdX or mkraid -force --dangerous-no-resync /dev/mdX for this. mkraid does not over write anything except for the raid superblocks themselves. So if you make sure to keep you /etc/raidtab up to date (if you have moved drives, or changed scsi id's), you can re-run mkraid with impunity and convince the raid system that yuour set is up to date. Note that this reconstructs the parity drive when run, so you better be damn sure all of the disks are in fact in sync. if you choose --dangerous-no-resync no resync will occur. In some situations you might have had an experience where one disk is kicked out, and an array runs in degraded mode for some time, and then another is kicked out. I have seen this with my temperamental ide raid system. In such a case, you really want to study the logs and figure out which disk failed first. Then edit your raidtab and replace that disks raid-disk line with one that says failed-disk in it's place. Then run mkraid as above, which will put your array in degraded mode. At that point you can mount (I suggest -ro at first) inspect to make sure everything is sane, e2fsck it, and if everything is still good, do a raidhotadd /dev/your_missing_disk to add it to the array and begin the resync. using failed-disk in this way allows you to choose the disks that make up the latest clean raid set... therefore not getting corruption from an out of date parity disk. Without using failed disk first, I don't believe that the reconstruction process can take this into account, and you may/will suffer corruption. When you're doing this kind of raid disk surgery, it's prudent to check twice and do once... that is, be very careful. It is easy to make mistakes during this process that can compound into total destruction of your file system. However, if you do it correctly and carefully there is really no (transient failure) situation that you can't recover from. Tom
