Fry-IT

Reliable journaling filesystem in Linux

By: Marcus von Scotti, 28th of September 2006

28^th of September 2006

We had problems with ext3 filesystem in the past, when bad sectors started corrupting the filesystem. Out-of-the-box ext3 is not fully configured to minimize the risk of creeping file-corruption. Here are a couple of simple steps to beef up the protection quite a bit.

The steps to getting a reliable ext3 journaling file-system are:

• fsck.ext3 -f /dev/xxx; sleep 5

  for any production filesystem, one needs to do a full -f(orced) fsck clean-up to avoid error propagation followed by a short sleep if you run this in a script



• tune2fs -O dir_index /dev/xxx

  create directory indexing



• tune2fs -O has_journal -o journal_data /dev/xxx

  make sure the journal is setup correctly



• tune2fs -e remount-ro

  prevent the file-system from being gradually trashed in case of errors being detected



• sleep 5; fsck.ext3 -D -f /dev/xxx

  clean up, re-index and sleep before this to avoid messing things up

Delaying the steps via sleep is very important to avoid really screwing things up (in case you want to stick all this into a clever little script)!

Just in case one can also add

data=journal,errors=remount-ro

to fstab options (I'm not completely sure if the data=journal would apply based on the stuff in superblock otherwise, I have vague recollection that I've seen it picking up data=ordered instead).

It's best to do all of this when you setup your ext3 file-system, but we have managed to convert existing production filesystems by a script outlined above as well. We can obviously give no guarantees, but the above tuning steps have made a difference to us (especially after unintended power-cuts).

Comment