[6bone] [NOTIFY] XS26 service/peering outage

[Yasuhiro Ohara]

> You were probably never running zebra on router with 2048 interfaces, having
> 2k static routes redistributed into BGP, 10k internal BGP routes, about 200
> prefixes in IGP and about 300 external BGP routes.
> 
> The result: CPU time of ospf6d reached sometimes ~100%, zebra was unable to
> save config files, zebra sometimes freezed for 5 minutes or so making ospf6d
> and bgpd also freeze, sometimes something crashed and so on.
> 
> Zebra is not ready for production networks.

Ok, it will be the problem of ospf6d, not bgpd. Current Zebra ospf6d
lacks some support, that needed in this case may be multiple Router-LSA
origination by one router. ospf6d cannot originate it that way and cannot
calculate them originated from the other router that way.

Another concern I have is that in the SPF calculation, ospf6d uses
simple insert sort of sequential list. LSDB lookup is now using
patricia tree (lib/table.[ch]) and should not be a problem.
Both is related to the size of the network (not related to the interface
 number each router has).

I guess, bgpd has no problem in your case. My take is that either ospf6d
interferes the bgpd from working correct, or it it just a file descriptor
limit problem. File descriptor problem is the one we WIDE Project
experienced and it came just fine when we change the file descriptor limit
bigger in the OS configuration.

Also the problem you encounter will be none of OSPFv3 protocol problem.
But I'm interested in your IGP, is it link-state ?

I plan to rewrite ospf6d to have it full support of OSPFv3 and other
OSPF extentions by the end of Apr 2003. This time I'm serious ;p)

regards,
yasu