From: Volkov, Dmitry (Toronto - BCE) (dmitry_volkov@xxxxxxxxx)
Date: Mon Aug 05 2002 - 21:48:55 GMT-3
Hello,
Thinking about secondaries in IGRP as way to redistribute classles to
classfull routing protocols,
I came to intersting conclusion:
Let say we have the following:
1) case r8(s0)-----igrp---/24----(s0)r4(s1)-----ospf--/29---- or even vice
versa:
2) case r8(s0)-----igrp---/29----(s0)r4(s1)-----ospf--/24----
several methods exist:
1) ospf piggy back on r4 - drawback - second routing process
2) tunnel between r8 & r4 with the same mask as ospf - drawback - additional
interfaces
3) secondary addresses between r8 & r4 ro reflect ospf prefix lengths -
drawback - split horizon issue
4) making r4 ABR and use area range - not always possible...(ABR must be
connected to area 0...)
5) summarize somewhere inside ospf if it possible...- not always possible or
permitted
6) static routes - usually not allowed in the lab.
Usually in such scenarios we do mutual redistribution igrp <--> ospf...
Taken into consideration that after 12.1(3) redistribute connected will not
work, because ospf is running on s1(r4),
we can do the following : assign secondary address on s0(r4) ONLY !!!
We can put many secondary there actually.
Igrp is running on this secondary address, redistributes it to OSPF, it
appears as LSA 5,
after that we can put summary-address with any mask we want - shorter or
longer than original ospf network -
depending on which case we have 1) or 2) , no discard route external if You
want, and this summary will be
redistributed to IGRP. Now R8 will know required network.
Care should be taken to avoid black holes on R4.
I think this way more clean than using tunnels or piggy back...
I tried it using 12.1(15) - works fine.
What do You think about it ? Did I miss something ??
Thanks,
Dmitry
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:48:17 GMT-3