For historical reasons I am running SNAP (802.3) on the RS6000. I have cisco routers in the network. When I went to the new OS on the cisco's I started getting all kinds of data corruption. Portions of some individual's files would show up in another user's files.
Although initial impression is that it is all cisco's fault, I've retreated to an older cisco OS and discovered that we have been haveing the problem all along it just wasn't noticed because the higher network layers were covering the problem.
Part of the problem is because the RS6000 doesn't provide UDP checksums for the RPC packets (which are used by NIS/NFS) Part of the problem is the translation between my 802.3 nets and my standard DIX nets (which points at the cisco's).
We verified there is a problem by 'summing' several remotely mounted large files. The whole file has to be transported across the net to be summed. You have to use several to make sure the buffer spaces are emptied. Well, the result is that the sum varies so the file is being corrupted. I'd like some
confirmation from other system administrators that they also see the problem. This is the csh script we were using.
while (1)
set cksum = "`sum /home/page/bin/ngrph | cut -d' ' -f1`"
if ($cksum != 62684) then
echo ERROR: Checksum $cksum does not equal 62684
else
echo Checksum for ngrph is $cksum
endif
set cksum = "`sum /home/page/utils/xloadimage | cut -d' ' -f1`"
if ($cksum != 45172) then
echo ERROR: Checksum $cksum for xloadimage does not equal 45172
else
echo Checksum for xloadimage is $cksum
endif
set cksum = "`sum /home/page/bin/ntst | cut -d' ' -f1`"
if ($cksum != 00725) then
echo ERROR: Checksum $cksum for ntst does not equal 00725
else
echo Checksum for ntst is $cksum
endif
end
Any data will be appreciated. Also Email is preferred.
