kris buggenhout wrote:
> Russell Cattelan wrote:
>
> > >
> > > I was filling the filesystem with an ftp client, just plain mget 6Gbytes,
> > > got to +1 Gig of data,... then i get a write error ... and the fs hangs
> > > ... cant
> > > umount ... -> no shutdown possible -> hard reset !
> > >
> > > then trying to get it repaired :
> > >
> > > [root@winnetou /home]# xfs_repair /dev/sde
> > > xfs_repair: warning - cannot set blocksize on block device /dev/sde:
> > > Invalid
> > > argument
> > > Phase 1 - find and verify superblock...
> > > Phase 2 - using internal log
> > > - zero log...
> > > - scan filesystem freespace and inode maps...
> > > freeblk count 2 != flcount -2048 in ag 0
> > > bad agbno 4294965248 for btbno root, agno 0
> > > bad agbno 531244720 for btbcnt root, agno 0
> > > - found root inode chunk
> > > Phase 3 - for each AG...
> > > - scan and clear agi unlinked lists...
> > > - process known inodes and perform inode discovery...
> > > - agno = 0
> > > Bus error (core dumped)
> >
> > This looks consistent with the problem we are working on right now.
> >
> > When the file system gets close to being filled it runs around finding all
> > the
> > small left over extents. At some point one of the writes goes to far and
> > scribbles
> > over
> > the the ag structures. We are getting close to resolving the bug... give us
> > a day
> > or
> > so.
> >
> > The core dump on the repair is not a good thing although... do you have a
> > stack
> > trace?
>
> not anymore ...
>
> but I checked out this morning the cvs .. and did a reduild ... it all
> compiled
> clean...
>
> but now I have a worse problem ... if I try to write data : input/output
> error ...
> umount and xfs_repair do work now ... but I cant write a bit of data , can
> create
> directories ok ... but cant write data
>
> [root@winnetou /xfs_filesys]# echo 1 > tttt
> [root@winnetou /xfs_filesys]# cat tttt
> cat: tttt: Input/output error
>
> Start mounting filesystem: sd(8,64)
> Ending clean XFS mount for filesystem: sd(8,64)
> I/O Error Detected. Shutting down filesystem: sd(8,64)
> Please umount the filesystem, and rectify the problem(s)
> PCD: pagebuf_bmap error -1010 pb_flags 0x10010002
> Start mounting filesystem: sd(8,64)
> Ending clean XFS mount for filesystem: sd(8,64)
>
> Should I be worried the disk is broken ? ... the filesystem that was
> preveously there
> ... seemed to be alright ... ??
Unclear.
Do you see any error messages on the console from the disk driver?
If xfs_repair runs without reporting IO errors then it probably isn't a disk
problem.
Assuming xfs_repair cleaned up the FS, does xfs_check run without errors?
That pagebuf_bmap error does seem to indicate some problem.
It this file system really close to being full?
-Russell
|