Lyvim Xaphir wrote: >On Thursday 11 April 2002 03:42 am, you wrote: > >>Well I'm officially very ticked off. >> >>I had some problems with X. It crashed, hard. Mouse/keyboard, >>evertying >>stopped working. Somehow, through a miracle I think, I got it to >>reboot by >>hitting teh reset button or just banging keys. >> >>Needless to say, I lost all the mail I'd d/led today. I also lost >>some >>small bookmarks *No biggy there*. The mail is what I was ticked >>about. I >>couldn't figure out a way to save it. I am using Sylpheed. >> >>When I rebooted, I got "fails" on a lot of Modules that tried to load. >>Like "DevFs=Fail!".... this was my first clue my install didn't >>survive. >> >>Now, knowing Reiser is supposed to stop corruption of an FS, why did >>it >>fail this time? Is it me or something else? >> >>any ideas to avoid this unfortunate situation in the future? >> >>I did try to get to a term by ctrl-alt-Fx. No joy. So...couldn't >>shutdown >>gracefully. >> >>*sigh* >> >>I tried the rescue option. Ya that didn't work. >> >>I tried the "reinstall but don't format /home". I have only 3 >>partitions >>*root (/), swap, home* and have yet to figure out how to divide up >>more & >>get it to work properly. *Thats a diff issue*. However, on reboot I >>got >>teh same fails. Plus! Kde couldn't write/execute the .ICEauthority >>file. >>:\ I tried the "chmod" method of giving global access... another file >>needed it. >> >>At that point it seemed easier to just reinstall all & save my nvidia >>drivers I'd downloaded instead. >> >>*sigh* Help? Pls? >> >>I am at a loss as to how to avoid/fix this in the future if this >>happens. >>And why reiser failed. >> >>Femme >> > >Ok. Well, I'm going to step out here at the risk of sounding like a >broken record. First, this filesystem thing is kind of like religion, >it's easy to spark off a debate. All I can do is point out what I have >observed on this and the expert list; and I've seen peeps with problems >both in XFS and Reiser...not just you, either. On the other hand (and >Civ you correct me if I'm in error) I have yet to hear of an ext3 system >that went down with corruption. Of course that does not mean there >hasn't been one. I can just assure you mine has not been one of those. >(KNOCK ON WOOD....) > > >IF you are not experimenting and IF you are not testing the relative >performance levels of the various filesystems, my personal and Official >Wooden Nickel Recommendation (OWNR, tm) is to use ext3 journaling as >installed by default. This is new stuff; and until the buggies get >worked out, I'd take the safe road, the one where there's no bandits >hiding in the directory trees out in the file forest along the way. No >sense trying to be an icecutter ship when you're in a fiberglass boat. > >The primary concern I had when installing was integrity of data. I >moved everything I had over to this system. And that's alot of stuff. >I've got backups but they are usually five days old at least. If this >rig crashes I'm SOL. So I copied red hat's recommendations and have >been watching carefully other folk's experiences. > >So that's my best approach to your question about avoid/fixing it in the >future. Sorry I can't be much help; wish I had a magic Reiser Wand I >could wave to get your data back. :( > >Till next time, > >LX > > > > > >------------------------------------------------------------------------ > >Want to buy your Pack or Services from MandrakeSoft? >Go to http://www.mandrakestore.com > Actually, there are reports of ext3/NFS corruption, but it may be broken clock chips causing it, and also some corruption copying from CD to ext3, which could also be bad clocks under UDMA operations at elevated speeds, with spurious clock interrupts playing hell with direct memory access.
Frankly, use ext2 for home use, unless you carry HUGE partitions. The recovery from a journaling filesystem is faster, but the operation is usually slower, and the security is _no_ greater. ext2 is extraordinarily well-designed and the one area of weakness is 4 hours to check a _full_ 50G partition in a worst-case scenario, but its check/rebuild is no more likely to lose data than any other. Now on Reiser, use the rescue CD to boot, modprobe the reiser stuff then try reiserfsck ... reiserfsck --rebuild-tree To do a complete fix if journal playback doesn't do it. One thing I have noticed with Reiser back in 8.0 was that an error in the tree will fail to be detected by journal playback and will show up only subtly, like missing libraries on a compile which nevertheless show up with ldconfig -v. Because this behavior is a serious failure--failure to detect a problem, I stopped recommendeing Reiser, and I have seen insufficient evidence to begin recommending it again. Using scripts people showed me with mixed IDE and SCSI drives on their systems, I was able to reproduce corruption on XFS for 8.1. Using the same scripts on 8.2, I can no longer reproduce it. JFS I was able to crash shortly before the 8.1 release. Repeated resets seemed to take it out even for a new version, but again the version with 8.2 seems very stable. Evidence is positive, but not yet sufficient. So ext2--solid ext3-semi-solid XFS-improved, need more evidence JFS-looking awfully good, but needs defrag and needs more experience Reiser-A very complex system, at least one level of abstraction and code above the others, and will be great if ever it works fully. ALL--An unexpected down can wreak havoc on a disk or on a filesystem. ALL-- When things look bad, try alt-sysrq-s alt-sysrq-a alt-sysrq-b to flush the buffers before reaching for a reset switch... If it doesn't work, you have lost no more than you would lose, and it if does work, you will minimize your losses. ALL--Data not backed up is data lost. End of story. Civileme
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
