Possible Kernel issue (?): "handed out a duplicate delegation" errors

[Nathan Patwardhan]

Looks like the issues we're seeing have been fixed in the 2.6.22
kernel (i.e. our Ubuntu systems do not see these errors) although have
not been addressed in patches before then.  Thanks, git!

Will try to work w/Novell to see if these changes can be rolled back
into the 2.6.16 kernel that's included w/SLES 10.

On Tue, May 6, 2008 at 3:55 PM, Nathan Patwardhan <noopy.org at gmail.com> wrote:
> Hello,
>
>  This issue has suddenly cropped up under SLES 10 w/the following specs:
>   - x86_64
>   - kernel 2.6.16.53-0.16-smp
>   - nfs-utils-1.0.7-36.26
>   - nfsidmap-0.12-16.17
>
>  Under certain circumstances (that we're still currently trying to
>  determine), the system will end up under a tremendously high load
>  (over 350!) and will no longer be able to function.  Reads, writes,
>  and killing processes will all fail and the only recourse is to
>  reboot.  When this happens, we see the following in /var/log/messages:
>
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   sql[4874]: segfault at 000000006320201c rip 00000000f7e2d9a9 rsp
>   00000000ffb3ff78 error 4
>   sql[4881]: segfault at 000000006320201c rip 00000000f7ebd9a9 rsp
>   00000000ff975da8 error 4
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   nfs_inode_set_delegation: server 172.17.4.50, handed out a duplicate
>  delegation!
>   decode_op_hdr: reply buffer overflowed in line 2071.<4>decode_op_hdr:
>   reply buffer overflowed in line 2071.<4>call_verify: auth check failed
>   call_verify: auth check failed
>
>  Has anyone else seen this before?  The message seems obvious, but I
>  want to make sure that it's not a red herring.
>
>  If the fix is to install updated libraries or a kernel,  we prefer not
>  to install custom RPMs for SLES 10 and instead to wait for packages to
>  be installed via online_update.
>
>  Note that we've only seen this problem under SLES 10 and not under
>  Ubuntu Gutsy or Debian Etch.
>
>  --
>  Nathan Patwardhan
>



-- 
Nathan Patwardhan
"In 1976, (the) Scorpions released Virgin Killer. The album's cover
featured a nude prepubescent girl covered by broken glass, was
designed by the German branch of RCA Records, their label at the
time."