Hi !

On Mon, 15 May 2000, Anders Magnusson wrote:
[SGEC]
> > No, just ignore it, it's a bug i think. I found out, this comes more often with
> > UDP packets. If i copy a file over the network with NFS, i can be sure to have
> > to grab my VT to shut down the machine, because network is impossible,
> > especially the ypbind doesn't get any more answer. If i copy the same file via
> > ftp, the problem isn't there. Ok, this is not much, but a bit :-)
> > 
> The SGEC should be a quite good chip, so it's probably not a hardware
> problem. I haven't seen any problem with it myself (on 4000/90) but
> I should maybe try to kick a little bit harder on a 4000/300 and see if
> I can get any errors.

My environment here: My Sun as Server is connected with 100Mbps fullduplex to
a good Switch, the VAX is connected through AUI-RJ45 to this switch, so it runs
10MBps halfduplex. This should give no problems on the hardware side, except
the Server having a little bit more performance than the VAX.

If i compile a kernel via telnet, everything runs fine, the kernel sources
are on a local disk on the VAX. If i want to copy it to my homedir to put it
onto the bootserver from there, in 99.9% it sais this xmit logic died stuff.
Then, absolutely no telnet session is possible, because i have NIS running
here. If i log in as michael on the console, the box is absolutely finished, no
more reaction due to missing NIS replies. Only chance to get rid of this is
loggin in to the console as root, and shut the machine down. If i ftp the
kernel to my homedir, this problem doesn't exists. Ok, good to have a nice
.netrc with "macdef init" :-) So i think, excessive UDPing might make trouble.
I didn't try excessive ICMPing by doing a ping -f, this might be another nice
test. In pure TCP, the SGEC works wonderful, and has a nice performance. Maybe
i should try mounting the NFS stuff with TCP, this could improve stability.

...Michael

-- 
 Michael@camaronet.de, Rottweil/Germany, http://www.camaronet.de
 Visit the german VAXpage: http://www.vaxpower.de