On Tue, 20 Jul 2004, nathan r. hruby wrote:
duplex mismatch? bad network cable? shifty DNS service? oddball firewall config?
Here's how I'd try to debug this: - Replace cables on both server and client - Drop firewall on server - Drop firewall on client - stick client on same switch or preferably a hub with server, try to eliminate all VLANing, etc.. - check to ensure nameservice works (either LDAP, NIS, DNS, Hosts - removing things you are not using from /etc/nsswitch.conf, then disable nscd) - Check to ensure time between machines is in sync (eg: ntp is important :) - Do a staitic mount to the NFS server, remove autofs / amd from the loop.
I also forgot, - check duplex and speed with ethtool / miimon
- If cisco try the "portfast" option (I'm not a cisco d00d so this is just
repeating with I've heard other say is a common problem :)
- Try upping the rsize and wsize to 8192 int eh mount options and see what happens.
-n
I've seen shoddy network gear dropping IP fragments do this. In that case *reducing* the rsize and wsize to 1024 fixes the problem. This is less than the 1500 packet MTU and a multiple of 512 for NFS. If fragment loss is a problem, you can usually also demonstrate this using ping with a larger packet size, like ping -s 4096 -c 1000 somehost This type of problem is more likely to show up when theres significant traffic on the net. -Bob Arendt