On Thu, 16 Dec 1999, Romain Kang wrote:
> We're using a central NFS server (a mutant SVR4.0+NFS3 variant) to keep
> multiple platform sources and we've historically been able to compile
> our source trees without a significant performance hit from NFS.
>
> However, we've recently been prone to errors in the vein of "RPC
> timeout", "connection refused", "request slots full", or "task %d can't
> get a request slot", as logged by Linux, SVR4.0, and Solaris NFS
> clients. To me, these sound like server bottlenecks, but the Linux
> clients seem to suffer worse performance than the others. For example,
> "ls -l" in a directory with about 20 entries can take up to 10 seconds
> on a RH6.0/Linux 2.2.12 client.
>
> Does anyone have suggestions how this NFS bottleneck might be alleviated,
> either on the client or server side? Or perhaps pointers to some NFS
> brains who might be willing to advise?
How do you know if it's NFS? It could be a overloaded server, a overloaded
network, bad routing, etc.
--
\ \/ / _ |~\ _ In God We Trust. All Others Pay Cash.
> < / \|\ /|+-< | | "The world is a comedy to those that think,
/ /\ \\_/| \/ ||__)|_| a tragedy to those who feel." - Horace Walpole
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
