[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | DEC Network Integration Server (DECNIS) |
| Notice: | Please read note 1 to use this conference effectively |
| Moderator: | MARVIN::WELCH |
|
| Created: | Wed Sep 18 1991 |
| Last Modified: | Thu Jun 05 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 3660 |
| Total number of notes: | 15082 |
3571.0. "I-ISIS Routing Convergence time" by CSC32::J_RYER (MCI Mission Critical Support Team) Fri Mar 14 1997 18:28
OpenVMS V6.1
|
--------------------------------------Ethernet----------------(166.37.5.0)-----
| |
NIS02 NIS05
| |
_____________________________________________
/ FDDI Ring (166.37.4.0) \
\_____________________________________________/
|
NIS04
\___ (wide-area link)
\ 166.37.3.0
NIS03
|
---------------Ethernet---------------------------------------(166.37.2.0)-----
|
Digital UNIX
Customer is using above configuration to do some timing tests of IP
failover, and is seeing longer-than-expected failover times between
the VMS and UNIX nodes. I think the IP failover is actually happening
in a timely fashion (about 3 seconds) and that the real issue is
the routing convergence time.
All four DECnis's are running V3.1-8 and are L2Routers, running the Link State
algorithm at both L1 and L2, and are all four in the same DECnet area.
IP standby is enabled on the Ethernet circuits on NIS02 and NIS05.
Routing timers are cranked down as follows on all four DECnis's
Minimum LSP Generation Interval = 1
DR ISISHello Timer = 1
Holding Multiplier = 2
ISIS Holding Multiplier = 2
Hello Timer on _all_ routing circuits = 1
ISIS Hello Timer on _all_ routing circuits = 1
(Note: customer had started with Hello Timer and ISIS Hello Timer adjusted
to 1 only on the Ethernet circuits on NIS02 and NIS05, and with them set
to the default values of 10 and 3 on all other circuits, but in the course
of the testing, he turned them all down to one second, with no effect on
the basic problem.)
Customer is testing failover time by starting a continuous ping on the
VMS node to IP address 166.37.2.2 (the UNIX box) and watching what happens
to it as he pulls the Ethernet cable from NIS02. (The VMS node is configured
for NIS02 to be its default gateway.) About three pings get missed completely
(indicating about a three-second IP failover time), and for about another eight
seconds, every other ping works and every other one fails. Customer is
surmising (and I concur) that it is taking eleven seconds for NIS04 to
realize that it no longer has a return path to the VMS node via NIS02.
(We believe that the ICMP replies had been path splitting between NIS02
and NIS05 on the return path.) After the eleven seconds, pings start
consistently working again.
The customer's questions are "why does it take so long for NIS04 to realize
that subnet 166.37.5.0 is no longer reachable via NIS02? Shouldn't NIS02
have sent out an LSP immediately indicating the change in reachable networks?
If so, what is causing the delay in NIS04 updating its routing tables to
reflect the change?"
We did get a routing circuit trace of the FDDI circuit on NIS02 while the
customer did the test three times. I can supply a pointer to the complete
CTF output if anyone is interested in looking at it. I will include below
all the L2 LSP's transmitted by NIS02 which have IP addresses in them
(they all show source address 08-00-2b-A6-C5-50.00.FE; there were also
other L2 LSP's with a different last two digits on the source address,
but they didn't look like they had anything to do with IP.)
Based on the fact that one of the LSP's showing 166.37.5.0 as reachable
was transmitted only .06 seconds before another one was transmitted with
that subnet NOT included in the LSP, I believe that NIS02 is in fact
sending out an LSP fairly promptly once the cable is pulled; the question
is, why does it take eleven seconds for NIS04 to quit sending packets
that are addressed to 166.37.5.2 via NIS02?
Thanks for any advice/comments,
Jane Ryer
MCI Mission Critical Support Team
22:51:52.98| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 50|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| |
| |
22:52:02.80| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 51|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| |
22:52:02.86| Tx| 87| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 53|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| |
22:52:24.84| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 54|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| |
22:52:39.98| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 60|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| |
22:52:49.95| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 61|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
22:52:49.98| Tx| 87| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 63|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
22:53:06.48| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 64|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| |
| |
22:53:07.07| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 68|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
22:53:19.98| Tx| 99| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 70|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.0 Mask 255.255.255.0 |
| |
22:53:28.98| Tx| 87| Type: L2 LSP |
| Protocol ID: 83, Length: 1B, Version: 01 |
|Src: 08-00-2B-A6-C5-50.00.FE, V1.0.0, Seq #: 72|
| Good Type: L2, Life: 1199, Phase: V, Overload: N |
| IP Address : 166.37.5.2 |
| IP Address : 166.37.4.2 |
| Costs: Default 0 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.5.2 Mask 255.255.255.255 |
| Costs: Default 40 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.3.0 Mask 255.255.255.0 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.1 Mask 255.255.255.255 |
| Costs: Default 20 Delay 128 Expense 128 Error 128 |
| Internal: Address 166.37.4.0 Mask 255.255.255.0 |
| |
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 3571.1 | lower CSMA-CD STATION FAILURE DETECTION TIME | MARVIN::HIGGINSON | Peter Higginson DTN 830 6293, Reading UK | Sat Mar 15 1997 13:59 | 23 |
|
Jane,
I think this is the Ethernet detect timer still being set to the default.
The basic problem is that on Ethernet there isn't a good way to detect
"cable out" and so we have to use "unable to send for some time t".
We at one point had t = 3 secs but we got problems from customers with
busy networks who kept getting false timeouts. The code went through
several revisions so they need to be using at least V3.1-7 and to look
at release note 9.2, which includes this text:
Now, a new characteristic, FAILURE DETECTION TIME, has been
added to the CSMA-CD STATION entity. This characteristic allows
a tradeoff between the time to detect a station failure and
the bandwidth consumed. Higher values result in less bandwidth
consumed but slower failure detection time.
With the hello timers set to 1 second, there will be very little extra
bandwidth consumed by lowering the detect time to 3 secs (or even lower).
Peter
|
| 3571.2 | but it doesn't take that long . . . | CSC32::J_RYER | MCI Mission Critical Support Team | Mon Mar 17 1997 17:57 | 14 |
| Thanks, Peter . . . that was indeed what we were missing.
One question . . . is that timer in seconds? If so, why when it was
set to the default (of 20), did the failover take about 11 seconds?
And with the failure detection timer set down to 5, the customer
was observing failover times for the pin somewhere around 3 seconds.
It almost seems as if the failure will be detected in about half
the time specified by the failure detection timer.
Any comments?
Thanks,
Jane
|
| 3571.3 | time is worst case detect time | MARVIN::HIGGINSON | Peter Higginson DTN 830 6293, Reading UK | Tue Mar 18 1997 03:26 | 19 |
| Jane,
>but it doesn't take that long . . .
Detecting cable-out isn't an exact science. It depends for a start on what
type of Ethernet interface is being used and how quickly the components
react. The detection software is also a "sort of" background task and how
quickly this reacts depends on the number of packets being transmitted from
the MPC, from other line cards and their sizes (the sizes can make a big
difference).
The timer value is the worst case detect time (eg if there is no other
traffic).
The timer sets the polling interval to be one quarter of its value, and
thus the minimum detect time is about half the timer value. With 1 second
ISIS LAN hellos being sent (which are big packets), you'll get the minimum.
Peter
|