Re: NFS, got bad cookie, and Bonnie++

From: Charlie Watts <cewatts_at_frontier.net>
Date: Tue, 16 Dec 2003 09:59:18 -0700 (MST)
Nevermind. I've just found:
http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/57696

Summary:

 ... readdir() is entitled to fail if the directory has been changed
 underneath it ...

 ... This is a known problem in the nfs code ...

(The "got bad cookie" messages *are* informational; Bonnie just handles
readdir() failling poorly ...)

But this appears to be a not-issue with Solaris, Linux, etc. I'm not clear
what the *correct* behaviour is, but Bonnie++ succeeds on those platforms.

On Tue, 16 Dec 2003, Charlie Watts wrote:

> When using FreeBSD as a client, bonnie++ over NFS fails.
> Server software does not appear to matter. Bonnie dies if the server is
> another FreeBSD machine, localhost, or a Linux-based SNAP appliance.
>
> Note that bonnie++ with Linux as a client does *not* fail.
>
> backup# bonnie++ -u 0 -s 2 -r 1
> Using uid:0, gid:0.
> Writing a byte at a time...done
> Writing intelligently...done
> Rewriting...done
> Reading a byte at a time...done
> Reading intelligently...done
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...Bonnie: drastic I/O error (rmdir): Directory not empty
> Cleaning up test directory after error.
>
> And the Bonnie.pid directory is left behind with 8000+ empty files in it.
>
> These (and many more) are logged on the client the moment "Delete files
> in sequential order" is printed by Bonnie:
> Dec 16 09:34:43 backup kernel: got bad cookie vp 0xc298d920 bp 0xc7723c78
> Dec 16 09:34:44 backup kernel: got bad cookie vp 0xc298d920 bp 0xc76964e0
>
> There is a PR that discusses this "got bad cookie" issue:
> http://www.freebsd.org/cgi/query-pr.cgi?pr=33203
>
> Matt Dillon suggests that the "got bad cookie" messages are informational.
> Nonetheless, Bonnie++ dies with FreeBSD as a client but not (at least)
> Linux.
>
> Is this a Bonnie++ anomoly? It is hurting our confidence in FreeBSD NFS in
> a big way.
>
> Any insight would be greatly appreciated.

-- 
Charlie Watts
cewatts_at_frontier.net
Received on Tue Dec 16 2003 - 07:58:51 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:34 UTC