Messages in this thread | | | Subject | Re: Horrible NFS performace | From | Trond Myklebust <> | Date | 30 Apr 1999 12:09:42 +0200 |
| |
Phil Blecker <tmwg@inxservices.com> writes:
> I hope that someone can shed some light here, or give me a way to > investigate. 3 computers, 2 are 2.2.6, 1 is 2.0.36. All ethernet. 2.2.6 has > PCI vortex, 2.0.36 has Intel ether express 16. Lots of the following in logs: > > Apr 29 14:45:59 mwg kernel: nfs: server inx not responding, still trying > Apr 29 14:45:59 mwg kernel: nfs: server inx OK > Apr 29 14:46:19 mwg kernel: nfs: server inx not responding, still trying > Apr 29 14:46:20 mwg kernel: nfs: task 768 can't get a request slot > Apr 29 14:46:20 mwg kernel: nfs: task 769 can't get a request slot
This is just a sign of congestion in the RPC layer. If the server is not able to treat the NFS requests fast enough, then a backlog of requests builds up in the RPC layer, and as they time out, they fire off the above messages.
I must admit, however, I'm not yet sure about all the causes of these errors. I've been seeing some under 2.2.6, which seemed to point to a problem with sockets: typically, a UDP socket returns -ECONNREFUSED without having transmitted anything to the server. Perhaps somebody with better understanding of the UDP layer has some suggestions as to what is going on in these cases?
Cheers, Trond
- 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/
| |