Ascend Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

(ASCEND) Re: OSPF question




> I'm sorry to hear you guys are having OSPF troubles. I had the same kind of
> problems for ages and went through extensive troubleshooting with Ascend to
> try to fix it. We never knew why it got stable but it did. Here's what we
> went through. I suggest you try it, as it works great for us now:
> 
> 1. Clear non-volatile RAM. Do it whenever you update software:
> 
> fclear
> Clears non-volatile RAM of configuration saved by tload.
> 
> fsave
> Saves configuration to non-volatile RAM.
> 
> nvr
> Clears non-volatile RAM and resets. If you've just uploaded new code via
> tload, configuration was saved to flash RAM and will be retained.
> Otherwise, configuration will be lost. Doing an nvr after tloading new code
> is recommended.

Done I think, though admittedly not universally.
 
> 2. Make _sure_ you have no duplicate routes. We found that we'd
> inadvertantly assigned the same subnet to two clients, each handled by a
> different Max talking OSPF to each other.

Can't do - our network is triangulated. If a network connected
user drops of one Ascend and pops up on another, the first
Ascend keeps the route advertised for a while, or until
it sees the route appear elsewhere. Bizarre but true. Thus
there are always duplicated routes to some extent. Also with
a triangulated network you inevitably see two routes to the
same subnet.
 
> 3. Turn off OSPF on unused interfaces. If your routers are talking to each
> other over WAN lines, you don't need it running on your Ethernet port.

Done.
 
> 4. Run at least 5.0Ap16  on any 4000s. I have yet to try Ap20. Pre-1.3A
> versions seemed fine on the TNT, and 1.3A is also stable for me. Again, I
> have yet to try later versions.

Done. We were on p16 before, we are now on p20. Still happening.

> I really hope this helps. I was in utter hell when my Maxen were rebooting
> themselves. There's pretty much nothing worse in this business.

Thanks for the suggestions. Ascend now acknowledge there is a problem
and can repeat it in their labs on demand. Whether on not this
will result in them fixing it, and distributing the results in
a timely manner is of course an entirely different question.

Alex Bligh
Xara Networks



++ Ascend Users Mailing List ++
To unsubscribe:	send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd:	<http://www.nealis.net/ascend/faq>


References: