Thank you for your time and your interest in solving this issue.
Your test added light to the issue, I will forward to Mikrotik the information 
received from you, maybe they will care and do something in the future updates.

Could you please share the patch for FRR so I can replicate the issue with FRR 
too?

> On 1 Jul 2023, at 17:20, Jörg Kost <[email protected]> wrote:
> 
> Well, the weekend is here, and I was able to test and recreate it.
> 
> With MLX, the behavior is like with CER
> -> BGP DOWN (Attribute Flags Error)
> 
> On SLX, the behavior is only a warning.
> -> received invalid AGGREGATOR attribute flag (0xd0).
> -> However, the actual route is installed, and the session stays up.
> 
> To recreate this, I had to actively change the BGP source code of FRR as peer 
> so that it is forced to set an extending length flag for AGGREGATOR Attribute 
> and writes a 16-bit value accordingly.
> 
> That's why I give the following verdict :-)
> In the name of interoperability when introducing new products and since the 
> Internet consists of many different manufacturers and devices and when it 
> comes to peers, the lowest denominator can sometimes be the best, it would be 
> appropriate to let Mikrotik know to persuade them to do without flagging and 
> the 16 bits in the aggregator. That's also better before people switch to the 
> new Mikrotik version and crash a lot of sessions with NetIrons or maybe other 
> manufacturers. Since the AGGREGATOR attribute has a length 8, an extending 
> length is unnecessary.
> 
> However, because I also have support for MLX and SLX, I will submit this as a 
> request when I get a chance. I am sure Extreme will adjust the SLX (different 
> log?) if necessary, but probably not the behavior for NetIron.
> 
> 
> On 1 Jul 2023, at 13:58, Bogdan Rotariu wrote:
> 
>> Thank you, I will not argue anymore as there is not so much interest from 
>> Mikrotik.
>> Yes, RR is an option, but a bad RR implementation is worst than no RR.
>> 
>> Mikrotik last answer is related to the BIRD/FRR issue from 2021:
>> 
>> "That bird problem is completely unrelated. They did not use 2bytes for the 
>> length field when extended length bit was set. RouterOS does not do that, 
>> length is encoded properly.”
>> 
>> Unfortunately we do not have any support for the CERs and ICXs we have 
>> within our network since Brocade got split in many pieces so we cannot ask 
>> Extreme Networks anything regarding CER. We do have 2+ years sessions up the 
>> CER’s, I cannot say nothing wrong about Netiron.
>> 
>> In the last days I’ve been testing Mikrotik’s CHR 6 release, Junipers vRR, 
>> Cisco IOS xRV, none of these generate issues to the CER and neither are 
>> affected by the Mikrotik. The only software in my testing that was affected 
>> by this is the Netiron, Dell 4032F prints the error and discards the prefix.
>> 
>> Thank you again for your great interest in this issue.


_______________________________________________
foundry-nsp mailing list
[email protected]
http://puck.nether.net/mailman/listinfo/foundry-nsp

Reply via email to