From: perkinsr@xxxxxxxxxxxxxx
Date: Thu Feb 08 2001 - 17:29:53 GMT-3
In the following scenario:
|TR0
----------O----------
FR| R1 FR|
O-R2 O-R8
Gig| Gig|
O-R3 O-R7
FR| R4 Gig|
----------O----------
R6 |
O--|Gig
O-R5
|E0
There are two paths from R1 to R5, P(ath)1=R1->R2->R3->R4->R5 or
P2=R1->R8->R7->R4->R5. Each frame relay circuit has delay of about 30ms. A
ping using P1 there and back should take about 120ms, using P2 there and
back should take 60ms.
I know that R1 sees P2 as the best path and R5 sees P1 as the best path.
I thought the pings generated by R1 should take P1 to R5 and then P2 on the
way back, while pings generated by R5 should take P2 there and P1 back.
Either way there should be a round trip delay of about 90ms or crossing 3 FR
circuits once. I am assuming the time moving across the Ethernet could be
neglected.
What I see is R1's ping go take P1 there and back for a round trip of 60 ms,
and R5's pings are taking P2 there and back for a round trip of 120ms.
When I do a trace I see (paraphrased):
>From R1:
1 R8 68 msec 68 msec 64 msec
2 R7 64 msec 64 msec 68 msec
3 R4 65 msec 64 msec 64 msec
4 R5 G0 64 msec 64 msec 68 msec
5 R5 E0 68 msec * 64 msec
>From R5:
1 R4 0 msec 0 msec 0 msec
2 R3 52 msec 56 msec 52 msec
3 R2 52 msec 52 msec 56 msec
4 R1 132 msec *
R2 52 msec
I don't know why that second packet on the last hop is always dropped
either. Somewhere on the way from R5 to R1 there is an equal cost path and
OSPF is load balancing, but I think this gives the jist of it all.
My question is, why is a ping one way taking the same route there and back?
If R1 sees P2 as the best path it should route packets that way if they are
ping replies or pings and if R5 sees P1 as the best path it should always
route packets that way.
What do you all think?
Ray Perkins
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:28:42 GMT-3