Subject: Re: NFS writes hang
To: None <firstname.lastname@example.org>
From: Michael van Elst <email@example.com>
Date: 05/28/2004 09:18:55
firstname.lastname@example.org (Urban Boquist) writes:
>of) the standard things to try for nfs hangs is to decrease the
>read/write sizes. try adding something like ",-r=4096,-w=4096" to the
>options list of client side fstab.
Decreasing read/write sizes is supposed to avoid buffer overflows
on the receiver. Some systems can't even handle two back-to-back
packets and reducing further to 1024 (i.e. < 1500-X) is necessary
to create enough overhead between adjacent packets.
>Dave> Interestingly, rpcinfo -p client when done from the
>Dave> client also hangs.
Now this says there is a different problem. rpc probes do not
use large packets that are fragmented. Even with a saturated
link it is unlikely that no probe request will come through.
Michael van Elst
"A potential Snark may lurk in every tree."