zfs discuss - Mar 2006 - ON 20060327 and upcoming solaris 10 U2 / coreutils

So, I noticed that a lot of the fixes discussed here recently,
including the ZFS/NFS interaction bug fixes and the deadlock fix has
made it into 20060327 that was released this morning. My question is
whether we''ll see all these up to the minute bug fixes in the Solaris
10 update that brings ZFS to that product, or if there is a specific
date where no further updates will make it in to the product? I ask as
I''m trying to make a decision to go live with a Solaris Express build
or wait for Solaris 10U2 for SPARC systems.

Second, in testing out nexenta gnu/solaris, which happens to package
the coreutils revision of du, I''ve found a bug in that program. A du
on a snapshot or other .zfs entity causes this error:

du: fts_read failed: No such file or directory

fts_read in previous bug reports for linux et al shows that the
underlying filesystem is corrupted. I highly doubt that, but it raises
the question of what Sun may wish to provide to third party utility
developers, perhaps more proactively, to get ZFS to be well supported
by their tool chains. I know the fault is most certainly on those
external tools, but some assumptions on file system primitives, header
definitions, etc may need to be updated for "Solaris 10" as these
people may already think that Solaris 10 has been covered in their
port. See, I cleverly linked this to the U2 release mentioned above.
Just waiting for Sol10U2 and hoping that all these issues are fleshed
out before then.

On Wed, Mar 29, 2006 at 09:11:31AM -0800, Joe Little
wrote:
> So, I noticed that a lot of the fixes discussed here recently,
> including the ZFS/NFS interaction bug fixes and the deadlock fix has
> made it into 20060327 that was released this morning. My question is
> whether we''ll see all these up to the minute bug fixes in the
Solaris
> 10 update that brings ZFS to that product, or if there is a specific
> date where no further updates will make it in to the product? I ask as
> I''m trying to make a decision to go live with a Solaris Express
build
> or wait for Solaris 10U2 for SPARC systems.
Almost all the ZFS bugs you''ve seen fixed in Nevada will be part of
S10U2.  We (i.e. George) are working hard on backporting the remaining
few in time.
> Second, in testing out nexenta gnu/solaris, which happens to package
> the coreutils revision of du, I''ve found a bug in that program. A
du
> on a snapshot or other .zfs entity causes this error:
> 
> du: fts_read failed: No such file or directory
> 
> fts_read in previous bug reports for linux et al shows that the
> underlying filesystem is corrupted. I highly doubt that, but it raises
> the question of what Sun may wish to provide to third party utility
> developers, perhaps more proactively, to get ZFS to be well supported
> by their tool chains. I know the fault is most certainly on those
> external tools, but some assumptions on file system primitives, header
> definitions, etc may need to be updated for "Solaris 10" as these
> people may already think that Solaris 10 has been covered in their
> port. See, I cleverly linked this to the U2 release mentioned above.
> Just waiting for Sol10U2 and hoping that all these issues are fleshed
> out before then.
I''m not sure what the issue is, but we decided to ease the pain for
general consumers by switching the default for "snapdir" from
''visible''
to ''hidden''.  This should make tar & other backup
utilities behave
better in the future.

- Eric

--
Eric Schrock, Solaris Kernel Development       http://blogs.sun.com/eschrock

On 3/29/06, Eric Schrock <eric.schrock at sun.com> wrote:
> Almost all the ZFS bugs you''ve seen fixed in Nevada will be part
of
> S10U2.  We (i.e. George) are working hard on backporting the remaining
> few in time.
>
That''s good to hear. I hope S10U2 w/ ZFS is available soon!
> > Second, in testing out nexenta gnu/solaris, which happens to package
> > the coreutils revision of du, I''ve found a bug in that
program. A du
> > on a snapshot or other .zfs entity causes this error:
> >
> > du: fts_read failed: No such file or directory
> >
> > fts_read in previous bug reports for linux et al shows that the
> > underlying filesystem is corrupted. I highly doubt that, but it raises
> > the question of what Sun may wish to provide to third party utility
> > developers, perhaps more proactively, to get ZFS to be well supported
> > by their tool chains. I know the fault is most certainly on those
> > external tools, but some assumptions on file system primitives, header
> > definitions, etc may need to be updated for "Solaris 10" as
these
> > people may already think that Solaris 10 has been covered in their
> > port. See, I cleverly linked this to the U2 release mentioned above.
> > Just waiting for Sol10U2 and hoping that all these issues are fleshed
> > out before then.
>
> I''m not sure what the issue is, but we decided to ease the pain
for
> general consumers by switching the default for "snapdir" from
''visible''
> to ''hidden''.  This should make tar & other backup
utilities behave
> better in the future.
>
I was explicitly entering the directory of the snapshot and running du
there. Just some tools that one may use do not yet support ZFS well,
and I was raising that awareness.
> - Eric
>
> --
> Eric Schrock, Solaris Kernel Development      
http://blogs.sun.com/eschrock
>

Glad to hear the news about snapdir being off by default.  I shot myself in the
foot with that yesterday wondering why a tar was taking so long.  Truss showed
it chomping through the snapshots which I hadn''t intended.
 
 
This message posted from opensolaris.org