hi: 8.1 with xfs is bad, unclean shutdown broke normal files. but 8.2 seems good to me. i shutdown my pc unclean several times a day, and only files *supposed to be broken* are broken. eg: files that not written to disk may be broke. but normal files won't broke under 8.2 in my case...
> Bryan Whitehead wrote: > >>On Tue, 2002-04-02 at 04:04, Guy Zelck wrote: >> >>>You obviously haven't followed my lead : read the 'w >>>ich is better choice ext3 or ...' thread. A lot of detail is in there. >>>I can reproduce it any time by just doing un unclean shutdown or when >>>I have to when the system hangs. But maybe it has something to do with >>>my SCSI card which is not 100%. I'm getting closer. >>> >>>Guy. >>> >> >>I should have mentioned I read the thread. We have had problems on >>machines that shutdown cleanly.... Files end up with null's in them >>instead of data. I can reproduce the error on basically any of our >>machines (around 20 Dell machines with onboard scsi). The problem never >>comes up with other FS's. (well, at least ext2) On very heavily loaded >>machines we've had entire directories disappear. Running any of the xfs >>tools does not restore missing directories or fix files that are full >>of null characters. >> >>If anyone would like instructions on how we trigger the problem about >>30% of the time (for null characters), I can give detailed instructions >>and some scripts. It would be nice to have this resolved. We've had to >>switch back to ext2 since the problems were discovered. >> > Strange isn't it how important problems seem to be ignored sometimes. > I've got the same problems in md8.1 kernel 2.4.8. I don't need scripts > to provoke it, any unclean shutdown does it. > Although my system disk is on an ide channel I also had a chain of scsi > disks. I took out the scsi adapter alltogether thinking maybe there > was a conflict and hoping the pb might disappear but no, it's not that > either. I'm now on the verge of trying the the latest xfs release or > the latest kernel 2.4.18. > I took a look on the www.sgi.com site about xfs and there's nothing > about this, strange. Also SGI is a serious company which makes it hard > to believe the code would be so broken. It must be a combination of > factors. Greetings, > Guy.
