i'm wondering what are the common elements here:
is the EEpro 100 ethernet card common to all clients
that exhibit this problem? NFSv3 over UDP? what is
the nature of the network flood -- ICMP packets, or
UDP? does "top" on the client show whether a particular
process or processes is spinning?
a tcpdump or at least "netstat -s" on the client after
the network is unplugged might be helpful.
> -----Original Message-----
> From: Cooke, Alan [mailto:[email protected]]
> Sent: Friday, May 17, 2002 9:15 AM
> To: [email protected]
> Subject: RE: [NFS] RedHat 7.3 crashed NetApp?
>
>
> While running the 2.4.18-4smp kernel (RedHat), and testing
> NFS3, our NetApp filer running OnTap 6.1.1R2 experienced
> massive incoming traffic from the Linux client in question.
> The Linux client required rebooting. We have been able to
> repeat this behavior, and are suspicious of the network
> driver on the Linux client which is using the eth0 driver. A
>
> Sound familiar?
>
> Alan Cooke
> Hewlett-Packard Company
> ISS/EDA
> Office: 281-518-9775
> Fax : 281-518-8015
> Email : [email protected]
>
> "If everything seems under control, you're just not going
> fast enough."
> -- Mario Andretti
>
>
>
>
>
>
>
>
> -----Original Message-----
> From: H . J . Lu [mailto:[email protected]]
> Sent: Thursday, May 16, 2002 5:39 PM
> To: [email protected]
> Subject: [NFS] RedHat 7.3 crashed NetApp?
>
>
> Has anyone seen
>
> http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=65069
>
>
> H.J.
>
> _______________________________________________________________
>
> Have big pipes? SourceForge.net is looking for download
> mirrors. We supply
> the hardware. You get the recognition. Email Us:
> [email protected]
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs
>
> _______________________________________________________________
>
> Have big pipes? SourceForge.net is looking for download
> mirrors. We supply
> the hardware. You get the recognition. Email Us:
> [email protected]
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs
>
_______________________________________________________________
Hundreds of nodes, one monster rendering program.
Now that?s a super model! Visit http://clustering.foundries.sf.net/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
[a few data points from the bugzilla entries]
On Fri, May 17, 2002 at 03:17:53PM -0700, Lever, Charles wrote:
> i'm wondering what are the common elements here:
>
> is the EEpro 100 ethernet card common to all clients
> that exhibit this problem?
eepro, tulip, 3com have all shown problems with 2.4.18 (at least
serious performance issues).
> NFSv3 over UDP?
Yes. A number of people have now reported that forcing v2 operation
does not result in slow performance, whereas in the 2.4.9-ac12 based
kernel, v3 was fine.
-ben
_______________________________________________________________
Hundreds of nodes, one monster rendering program.
Now that?s a super model! Visit http://clustering.foundries.sf.net/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
On Fri, 2002-05-17 at 20:17, Benjamin LaHaise wrote:
> [a few data points from the bugzilla entries]
>
> On Fri, May 17, 2002 at 03:17:53PM -0700, Lever, Charles wrote:
> > i'm wondering what are the common elements here:
> >
> > is the EEpro 100 ethernet card common to all clients
> > that exhibit this problem?
>
> eepro, tulip, 3com have all shown problems with 2.4.18 (at least
> serious performance issues).
>
> > NFSv3 over UDP?
>
> Yes. A number of people have now reported that forcing v2 operation
> does not result in slow performance, whereas in the 2.4.9-ac12 based
> kernel, v3 was fine.
>
from the discussion in bugzilla it seemed like a fair number of people
were suffering from the nfs defaults going to nfsv3, upd and 32768 wsize
and rsize/ 32768 bytes. That size of packet is going to be hard for even
REALLY well maintained networks to manage.
It might be worthwhile to try out the nfsv3 client over tcp and see if
the same problems occur.
-sv