tjt@cirrus.com (Tim Tessin) (06/27/91)
Subtitle: Gosh darn it Cisco, ya build these damn things too robustly! Y'all might find this interesting: Background: 1 AGS, 1 CGS in two different cities, connected by 2 parallel T1's routing IP, Appletalk, and bridging Decnet. Last weekend I upgraded the AGS to 8.2(4). Monday rolls around and all the Appletalk between cities is down. Appletalk routes between ethernets but not via the serial lines. So I call cisco support, I try upgrading to 8.2(4) at the CGS end, still no good. The Problem: crossed serial lines (Thanks Cisco Support!) ------ ----- t1 ----- ------ | |--- ---|CSU|------------|CSU|---------| | | | \ / ----- ----- | | |AGS | x |CGS | | | / \ ----- t1 ----- | | | |--- ---|CSU|------------|CSU|---------| | ------ ----- ----- ------ Who would ever suspect that this stuff would run for over 9 months and 2 upgrades when the two ends of the serial lines aren't connected to their respective networks. The only reason it even broke at all was the fact that in 8.2(4) the software pays attention to the appletalk addresses on the serial interfaces and the mismatched addresses prevented routing information from being transmitted. Hey Cisco, to heck with my expensive T1's. I'm gonna try plugging an interface card in to the 110v socket. PG&E rates are much better that Pac Bell. Hmmm, Cisco routes over 15,000 watts per second.... Tim Tessin - Cirrus Logic Inc. tjt@cirrus.com uunet!cirrus.com!tjt -- Tim Tessin tjt@cirrus.com uunet!cirrus.com!tjt
fortinp@bwdls56.bnr.ca (Pierre Fortin) (06/29/91)
In article <1991Jun26.193714.26266@cirrus.com>, tjt@cirrus.com (Tim Tessin) writes: |> |> Subtitle: Gosh darn it Cisco, ya build these damn things too robustly! |> [stuff deleted] |> The Problem: crossed serial lines (Thanks Cisco Support!) |> |> |> ------ ----- t1 ----- ------ |> | |--- ---|CSU|------------|CSU|---------| | |> | | \ / ----- ----- | | |> |AGS | x |CGS | |> | | / \ ----- t1 ----- | | |> | |--- ---|CSU|------------|CSU|---------| | |> ------ ----- ----- ------ |> I had a similar situation here for a short time but my topology was a little more complex: |--R4----------------R5--| |--R1--\ /-----------R3--| | | X | |--R2--/ \------------------------------------R6--| Mind you, there were actually multiple Ethernets involved and many more serial links, but this triangle managed to get the packets through!!!! The clue that something was wrong was a complaint about throughput in this section relative to the rest of the network (all T1 in the metro). I found that some packets were traversing the same link up to four times (twice in each direction) until they managed to figure out how to get to the ultimate destination. So I agree: too damn robust... :^) |> Who would ever suspect that this stuff would run for over 9 months and |> 2 upgrades when the two ends of the serial lines aren't connected to their |> respective networks. The only reason it even broke at all was the fact that |> in 8.2(4) the software pays attention to the appletalk addresses on the serial |> interfaces and the mismatched addresses prevented routing information from |> being transmitted. Well, *I* believe you... |> Hey Cisco, to heck with my expensive T1's. I'm gonna |> try plugging an interface card in to the 110v socket. PG&E rates are much |> better that Pac Bell. Hmmm, Cisco routes over 15,000 watts per second.... Hey! You weren't supposed to say anything about that... cisco is still trying to figure out how to handle the influx of user generated watts which, in some jurisdictions, may be flowed into the power grid... The routing of these watts is still not fully understood in the Opens and Shorts Power Flow (OSPF) protocol. Ooops, I think we both blew our non-disclosures. |> |> -- |> Tim Tessin |> tjt@cirrus.com |> uunet!cirrus.com!tjt -- Cheers, Pierre Fortin fortinp@bnr.ca (613)763-2598