Hopefully someone out there knows something about this: I have a machine for a server that I am setting up. I have had both linux and 2k server on there. My Question has to do with unexpected shutdowns caused by power interuptions. I have a UPS, but I dont feel like I should completely rely on just that. When I run linux and there is an unexpected shutdown due to power and the system restarts, linux wants to check the disks for errors and sometimes comes up with an error that it asks me if I want to fix. If I say yes it says that I "could lose data" and I have lost it in the past if I let it fix it. On the other hand I dont wanna leave a messed up filesystem(EXT3). Running XP/2k with an unexpected shutdown from power, I have never lost a file unless it was open and unsaved. I dont feel comfortable having linux (even though i love it) as a server if I am going to lose files like that. Is there something I am missing? Does anyone have any questions/comments? Thanks so much for your time and input

Why do you call EXT3 "messed up?" EXT3 has journaling, just like NTFS and many other FSs, so you shouldn't lose data. I've never had problems with EXT3. Normally, however, I use ReiserFS becuase it performs better in many situations and has less limitations. It also uses journaling so you won't lose data (I never have). Perhaps you were confusing EXT3 with the older EXT2, which is the same thing but without journaling.

Maybe I misunderstand, but doesn't journaling make a filesystem more prone to dataloss if a computer is shutdown unexpectedly? doesn't journaling load the drive index into memory and if the computer isn't shutdown correctly and the drive unmounted, the index never gets written back to the disk, right? Or do I misunderstand?

Regarding ext2... You "could lose data". This means, that a portion of the file structure is missing or damaged, has to be recreated or deleted, and may have linked to data. This data is usually temporary files. "I have never lost a file unless it was open and unsaved." If you do not link the file to your file structure and save it somewhere other than the swap partition (or volatile memory, also a no-save), then it is a candidate for the abyss. Are you sure you have ext3? You should not need to run e2fsck. The journaling part of ext3 means that the journal is what is checked after each unclean shutdown. http://www.redhat.com/support/wpapers/redhat/ext3/

Journaling keeps track of the filesystem status somewhere within the filesystem. When you boot after a crash, it uses the journal to restore the filesystem to a healthy state, but unsaved files are still lost. NTFS, EXT3, ReiserFS, XFS, JFS, and others all use journaling. Personally, I prefer ReiserFS because it normally performs better than EXT3, with less limitations, and is very stable on Linux.

Journaling writes a metadata log to the disk *before* changes are made. Older disk checks had to walk through the entire filesystem's metadata, seeing if everything was correct. If something needed to be corrected, there often was not sufficient metadata to do so, causing the linked data to be placed in lost and found (or completely deleted). Because journaling records before actions are made, often those actions that are interrupted can be completed upon reboot. There is a slight performance penalty for journaling, but it is definitely worth it on the modern PC.

Hi, thanks for all of your input... one more question if you dont mind... I have 2 boxes next to eachother (1 linux, 1 XP). How come it seems that when they both experience a power loss the linux box loses data or its corrupt while that never happens to that windows machines (or almost all the other windows machines i've worked with save a few flukes.) It seems like this one linux box gets more corrupted data due to unattended shut downs than all of the windows machines i've ever delt with combined. Any ideas? THanks again