[KLUG Members] Is tar broken? KERNEL BUG?

John Pesce members@kalamazoolinux.org
Thu, 8 Aug 2002 14:28:30 -0400 (EDT)


Ok.
I'm using Redhat 7.3 on the new server2 with ext3 filesystems.
I'm using Redhat 7.1 on the old server1 with ext2 filesystems.

> > I initially tried:
> > > cd /work
> > > cp -R home /
> >
> > but this seemed to hang half-way through.
>
> Did you try "cp -a"?

yes. I repeated it with that option. Same effect, but different files.
Some of the files that were different the first were ok the second time,
but some of files were different.

The filesystem was not being used.
Both filesystems are on RAID 5 arrays.

> Any errors in any log files or dmesg for why it hung?

I didn't see any.

> Was the directory still available from NFS after the hang?

Yes.

> You might want to try a different block size on the NFS mount.

umm. I never touch that before. Any recommendations?

> > Then I tried:
> > > cd /work
> > > tar -cvf - home | tar -xvf - -C /
> > > diff -r home /home > /tmp/list
> >
> > and within /tmp/list I got a few lines that looked like this:
> >
> > Binary files home/user1/file.tar and /home/user1/file.tar differ

> Any chance a process on one of the two systems has the file(s) open?

no. These are old files that no one is using.

> Did you double check that with fuser?

no.

> What version of tar are you running?  This is a long shot, but there was

GNU tar 1.13.25


UPDATE: the new machine just produced the following message!
What the heck does this mean? Is the ext3 filesystem STABLE????


 Assertion failure in journal_revoke_R3ff8495d() at revoke.c:329:
"!(__builtin_constant_p(BH_Revoked) ? constant_te
st_bit((BH_Revoked),(&bh->b_state)) :
variable_test_bit((BH_Revoked),(&bh->b_state)))"
Aug  8 12:31:48 terminator kernel: ------------[ cut here ]------------
Aug  8 12:31:48 terminator kernel: kernel BUG at revoke.c:329!