OK, first a big thanks to everyone who contributed to this one as well
as those watching from the sidelines. I believe we're very close to
having this one figured out and here's how I've come to this conclusion:
I recreated Chuck's experiment with 3 routers on an ethernet segment. I
also simultaneously ran a sniffer on the segment to determine what's
going on for the 40 seconds or so while the DR is unavailable--that does
seem like an awfully long time. So I ran "debug ip ospf events" on the
BDR and ovserved the output. Things looked normal--regular send/receipt
of Hellos--then I removed the cable connecting the DR from the hub and
observed the output. Just what you'd expect...in 40 seconds an election
took place..BDR became the DR and the only DROther was now the BDR.
Well one odd thing ...the box made a note to "remember" the old DR as
indicated by this event msg:
03:16:25: OSPF: Remember old DR 192.2.2.1 (id)
I found this odd and I really have no explaination for it??
But the key point is that review of the sniffer capture showed that 5
packets after the DR stopped sending hellos the BDR began sending acks
to the DROther on LSAs--as best I can tell, at this time it was still
the BDR.
CONCLUSION: The process of DR/BDR election after a DR is declared dead
appears to be completely independent of the BDR "backing-up" the DR in
performance of it's duties. It seems even though the BDR has assumed
the "duties" of the DR it remains the BDR until the dead timer expires.
While I still don't know the timeframe precisely I must admit the 0.5
second answer given earlier seems like a reasonable SWAG. Any comments
are appreciated...Frank
"Shaw, Winston Mr." wrote:
>
> Chuck,
> Thanks for testing. Now my curiosity is at an all time high. Is it possible
> for you to do the test with 4 routers ?
> My theory is that 2 or maybe even 3 routers are not enough.
> Here is why: 2 routers on a broadcast net will always have a DR and BDR. If
> the DR goes away, the DR will know it is alone on a broadcast net. It will
> become a DR in its own right. It will not be taking over any duties of the
> former DR. The time to realize that it is alone will be 40 secs(default). A
> 3 router net might have the same issue. The absence of the DR will trigger a
> promotion of the BDR. The remaining router will become the BDR. This seems
> to be consistent with the broadcast net rules.
>
> Although the code obviously selects BDRs and DRs very early, maybe the
> duties of DR flooding do not start until there is at least one other DRother
> on the net.
>
> I only have two routers available and my test produced essentially the same
> results as yours. Something I did notice is that the Network type 2 LSAs
> never aged.
> If a 4 router test produces the same results then it will be difficult to
> say whether LSAs or hello packets determine when the BDR takes over for the
> dead DR. I hope they never ask this question on any test.
>
> Winston.
>
> -----Original Message-----
> From: Chuck Larrieu [mailto:[EMAIL PROTECTED]]
> Sent: Saturday, November 11, 2000 8:39 PM
> To: Matthew Herman; David Armstrong; [EMAIL PROTECTED]
> Subject: RE: Some OSPF Questions
>
> Just to put in some empirical data, I set up two routers on an Ethernet
> link, in the classic OSPF broadcast scenario. Hello time is 10 seconds. Dead
> time 40 seconds ( 4xhello )
>
> I determined which of the two routers was the DR, and which was the BDR
>
> I then plugged into and monitored from the BDR, using repeated "show ip ospf
> neighbor" commands
>
> I then unplugged the DR from the ethernet
>
> I then repeatedly reissued the show ip ospf neighbor commands
>
> I watched.
>
> The result of the show command was that the neighbor state FULL/DR remained
> in effect until the dead time was reached. After that, there was no
> neighbor.
>
> I also plugged it the first router back into the ethernet and repeatedly
> issued the commands. After a few seconds the first router showed up as a
> FULL/BDR
>
> Of some interest - the debug ip ospf hello and debug ip ospf events were
> silent immediately after unplugging the DR. It was only after the expiration
> of the dead time that debug ip ospf events indicated the election of a new
> DR, to whit, the router I was monitoring.
>
> Where this 0.5 second / half second thing comes from I cannot say. But using
> Cisco's defaults in a quick and dirty lab, it is safe to say that this is
> not what happens.
>
> Chuck
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of
> Matthew Herman
> Sent: Friday, November 10, 2000 9:56 AM
> To: David Armstrong; [EMAIL PROTECTED]
> Subject: RE: Some OSPF Questions
>
> I'll throw my hat in..
>
> 1. .5 seconds (50 msec) (Chapter 7, p142 exam cram acrc)
> 2. yes, there will be only one DR and its your single point of failure as
> well 8->.
> 3. doh...I have set up multiple as's on one router when I had multiple
> customer and redistributed into my AS. It worked ok but I am not saying
> that was a good way of configuring the router.
>
> 2/3 = 66 percent.... Still not enough to pass the ccie...
>
> matt
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of David
> Armstrong
> Sent: Friday, November 10, 2000 8:46 AM
> To: [EMAIL PROTECTED]
> Subject: Some OSPF Questions
>
> Last night at our BSCN study group meeting in Dallas we had some questions
> about OSPF that we weren't able to resolve. If someone or ones could answer
> these it would clarify some areas we're a little fuzzy on. Also, if you're
> iin the Dallas Ft. Worth area and would like to attend, we'd love to have
> you join us..
>
> Thanks for any help,
>
> David Armstrong
>
> 1) What is the default time period that the BDR waits when listening to
> LSA's from the DR before it decides that the DR is down and promotes itself
> to DR. All the literature we could find simply said that the BDR waits for
> the specified time period but never said what that period is.
>
> 2) In a Point-to-Point network in which the router in Area 0 is connected to
> FR, ISDN, X.25 or ATM branch offices (networks), how does convergence and
> updates take place? From what we've found a DR and BDR is not elected in a
> strictly Point-to-Point network.
>
> I think an example would explain this question better: We have one 3620
> router in our Ft. Worth office connected to an office in Houston (via FR),
> and office in Kansas City (via FR), an office in the DFW area (via ISDN) and
> the owner's home (via ISDN). The 3620 is behind a firewall (Pix 520) and the
> firewall is connected to a 1720 going to the Internet. I'd like to implement
> OSPF on our network simply for the experience. However, I don't have 2
> routers internally on our Ethernet LAN that can be configured for Area 0 and
> elected to DR and BDR. All other routers connected to that router are via
> NBMA Point-to-Point connections. Since I only have one router on the
> Broadcast Multiaccess network (the 3620) and routers connected via PtoP
> don't participate in DR and BDR elections, how would updates occur? Can
> their only be one DR (in this case the 3620)?
>
> 3) The books and tutorials all state that "router ospf 6" defines ospf on
> the router with a process ID of 6. They then all say that you shouldn't
> define more than one process. Does that mean that you can have a router with
> the following:
>
> router ospf 6
> network 10.100.0.0 0.0.255.255
>
> router ospf 7
> network 10.200.0.0 0.0.255.255
>
> If this is an allowed configuration, what kind of instances would it be used
> for? Also, exactly what is the process ID used for?
>
> _________________________________
> FAQ, list archives, and subscription info:
> http://www.groupstudy.com/list/cisco.html
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
> _________________________________
> FAQ, list archives, and subscription info:
> http://www.groupstudy.com/list/cisco.html
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
> _________________________________
> FAQ, list archives, and subscription info:
> http://www.groupstudy.com/list/cisco.html
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
> _________________________________
> FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
