Progress!

[swestrup]

It seems like I DID have filesystem corruption after all. I thought that when 'fsck.reiserfs /mnt/system' returned immediately it meant that everything was fine. Nope. It means that fsck.reiserfs doesn't work on mounted filesystems. When I gave the same command on /dev/md5, it found all sorts of problems and I was forced to rebuild the file tree.

That has now finished and fsck.reiserfs claims it didn't find anything it couldn't fix, and that the filesystem is now clean. Yay!

I'm starting to get pretty damn tired now, so I'll leave it there. Besides, right now I don't think there'll be any trouble finishing the job in the morning, so my stress levels are low, and I should be able to sleep.

Were I to continue, and something ELSE were to crop up, I would not be willing to leave it alone and just go to bed, and who knows what damage I could inflict while too tired to work well, but too anxious to sleep...

So, at this point, sleep is the better part of valour.

Comments

[labyrinthman.livejournal.com]

Happy birthday, btw. (if LJ's birthday feature is reporting correctly, that is)

[swestrup]

Thank you!

[cpirate.livejournal.com]

Happy birthday!

You may want to double-check that the drive is still in good shape; one common source of ReiserFS corruption at work is bad drives. Of course, another large source was power outages (despite ReiserFS being journalled!) - we had to put in a big fat hack to force single-threading for Reiser to put an end to that.

[pphaneuf.livejournal.com]

Nah, we didn't leave the "big reiser lock" after all, we disabled tail packing (with the "notail" mount option), which removed most corruption problems on poweroff. There was a bug in the journalling code that dealt with tail packing.

[(Anonymous)]

Oh right, my bad. That'll teach me for posting before properly waking up.

[cpirate.livejournal.com]

It's 1 in the afternoon, and I'm still not sure I've woken up. That was me. *sigh*

[pphaneuf.livejournal.com]

I'd recommend you do a "dd if=/dev/md0 of=/dev/null" and monitor for errors in the kernel messages (dd might complain on it's stderr as well). Like cpirate says, disk errors are a very common cause for reiserfs corruption, and just reading the whole disk could expose badness.

See you later tonight for your birthday party!

[sps.livejournal.com]

...And I was worried when you said that you had started by mounting everything. Really, when the fs is corrupt, you do NOT want to mount it!

[pphaneuf.livejournal.com]

The trick part is also that with reiser, you might not necessarily want to run reiserfsck either.

[swestrup]

Yeah, but I had been getting bizarre errors from the raidtools and I wanted to check that the raid partitions that I had finally rebuilt were indeed the ones I was trying to build. I didn't want to run fsck on something that was built out of the wrong set of partitions...

[sps.livejournal.com]

Well, maybe Pierre knows differently, but I would always have said that a read-only fsck was the best way of finding out whether I had found an intact filesystem. A mount is supposed to *do* stuff, and it relies on fsck to have put things in order, first.