Re: [2.3.17] Poor NFS client handling of server death

Jeff Garzik (jgarzik@mandrakesoft.com)
Fri, 03 Dec 1999 23:17:11 +0000

Trond Myklebust wrote:
> The problem is that the NFS client doesn't know that the RPC is having
> trouble sending out requests. It therefore continues to send out new
> asynchronous write requests. As they start to time out, the above
> messages start to snowball.
>
> I've fixed this in NFSv3 and also the 'ac' series by adding a wait
> queue on which the process that is attempting to write can sleep while
> the server is down. It is then woken up once the existing backlog of
> requests has been cleared.
>
> I hope to find time to backport the improved RPC stuff to the standard
> kernel soon. I think it is good enough for 2.3.x.

Any progress on this for 2.2.x?

Running 2.2.10pre9 NFS server and 2.2.10pre10 NFS client, I can kill NFS
reproducibly with high traffic.

No NFS errors on the server, the following errors on the client:

rpc_release_client: nfs client already free??
rpc_release_client: nfs client already free??
rpc_release_client: nfs client already free??
rpc_release_client: nfs client already free??
rpc_release_client: nfs client already free??
nfs: task 65219 can't get a request slot
nfs: task 65221 can't get a request slot
nfs: task 65222 can't get a request slot
nfs: task 65218 can't get a request
slot 

-- 
Jeff Garzik              | Just once, I wish we would encounter
Building 1024            | an alien menace that wasn't immune to
MandrakeSoft, Inc.       | bullets.   -- The Brigadier, "Dr. Who"


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/