Hi Quentin, Thanks again for your review.
> On Jul 29, 2023, at 1:10 PM, Quentin Armitage <[email protected]> wrote: > > I believe that section 8.3.2 - "Recommendations Regarding Setting Priority > Values" is > incorrect. > > 1. Second paragraph, "especially if preemption is set" should be deleted. > Only one Virtual > Router can be the address owner, and the "especially ..." implies that there > are > circumstances under which more than one Virtual Router can be configured with > priority 255. I agree. > > 2. Third paragraph. This paragraph states that uniformly distributing > priority values > "facilitates faster convergence". This is not correct; using higher > priorities results in > faster convergence, since the higher the priority the lower the value of > Skew_Time, and > hence Active_Down_Interval. > > In order to achieve the fastest transition of a Backup Router to Active > Router after the > original Active Router fails or shuts down, the priorities should be > configured to be as > high as possible. This needs to be tempered by the differences in Skew_Time > between the > various Backup Routers should be sufficiently large that the second highest > priority Backup > Router consistently does not transition to be an Active Router since it sees > the first > advertisement from what was the highest priority Backup Router before > Active_Down_Interval > expires on the second (and lower) priority Backup Routers. > > I believe the following could replace the third paragraph: > > "For the fastest transition of a Backup Router to Active Router after the > original Active > Router fails or is shut down, configured priorities should be as high as > possible, since > this reduces Skew_Time. It is important that the differences in Skew_Time > between the > Virtual Routers are sufficiently large that the highest priority Backup > Router transitions > to Active Router and sends an advert before Active_Down_Interval expires on > any other any > Backup Router, thereby ensuring that only one Backup Router transitions to be > an Active > Router. I’ve reworked this section with this as the last recommendation and softened. > > It should be noted that this is more critical with lower > Advertisement_Intervals, and > priorities that work with an Advertisement_Interval of, for example, 100 may > not work > reliably with an Advertisement_Interval of, for example, 10." > > > Equal priority Virtual Routers > ============================== > Whilst the VRRP protocol and procedures work with Backup Routers having equal > priorities, it > causes operational problems due to two, or more, Backup Routers transitioning > to Active > state simultaneously, and learning bridges updating their MAC address caches > following the > failure or shutdown of the previous Active Router. This will only be > corrected once the > Virtual Router with the higher primary IPvX address next sends an advert (I > have separately > proposed that if an Active Router receives an advert from a lower priority > (or equal > priority and lower IPvX primary address) Virtual Router, it should > immediately send an > advertisement rather than wait for Adver_Timer to expire, thereby speeding up > the recovery > from having two (or more) Active Routers). > > I therefore suggest adding the following paragraph at the end of section > 8.3.2: > > "In order to avoid two or more Backup Routers simultaneously becoming Active > Routers after > the previous Active Router fails or is shut down, all Virtual Routers SHOULD > be configured > with different priorities, and with sufficient differences in priority so > that lower > priority Backup Routers do not transition to Active state before receiving an > advertisement > from the highest priority Backup Router following it transitioning to Active > Router." I agree and have adopted this text. Thanks, Acee > > > With regards, > > Quentin Armitage > _______________________________________________ rtgwg mailing list [email protected] https://www.ietf.org/mailman/listinfo/rtgwg
