Thanks for fixing the realmid passing around.

My original point about breaking up the patches into separate changes still
stands.

I see some whitespace changes in all the patches that don't need to be
there.

If you must reread the rt_realms file at the drop of a hat, I would prefer
the usage of the inotify_XXX system calls, to be called back when it
changes instead of rereading a file once a minute.

donald

On Wed, Jul 1, 2015 at 10:58 AM, Kaloyan Kovachev <[email protected]>
wrote:

> Hello,
> I have updated the patch to include support for IPv6 too and have removed
> the floating number of parameters of some functions.
> https://github.com/KaloNK/quagga/compare/master...realms
>
> About caching the rt_realms file: It should be enough if inside
> rtnl_rtrealm_initialize we do it once every minute instead of each call?
>
>
> On 2015-06-28 15:17, Kaloyan Kovachev wrote:
>
>> Hi!
>> Sorry for the late replay, I broke my leg, so won't be able to check
>> my emails often for while, but will try.
>>
>> On 2015-06-21 19:23, Paul Jakma wrote:
>> Hi Kaloyan,
>>
>> On Mon, 15 Jun 2015, Kaloyan Kovachev wrote:
>>
>> Hello devs,
>>
>> I am using the realms patch from Adrian Ban [1] for several years, but it
>> is still missing from Quagga, while it is very useful (not only for me I
>> hope). I have ported the patch for the latest version on GitHup and while
>> working on it found another small bug, which I have fixed [3] Is there
>> something else that needs to be done so they can be included in the
>> official source?
>>
>> It's an interesting patch. I went through it and wrote a longer commit
>> message for it.
>>
>> I think it needs work before we could commit it.
>>
>> It may be a while until I will be able to do it, but will do my best.
>>
>>
>> 1. On the bgpd side, it's basically an ID that can be set per-peer and
>> that can be set in a route-map action, but isn't significant for
>> route-selection.
>>
>> Setting it for a peer - can't you just use the peer ID? Is that
>> required?
>>
>> Further, it looks like the peer->realm only gets used for RSclients?
>> Never gets used to set a route realm otherwise? Is that the intention?
>>
>> The bgp_{export,input}_modifier hunks don't seem right. It's applying
>> the realm from the peer->realm regardless of the existing value, and
>> it's modifying the global attr. Shouldn't it be modifying the
>> rsclient specific attr?
>>
>> I'm guessing you don't use the "neighbour ... realm ..." form, and only
>> use it with route-maps?
>>
>>
>> I do use both (per peer and as route map) and yes it isn't significant
>> for route-selection, but route classification and shaping later,
>> please see the example below
>>
>> 2. On the zebra side it seems to only be honoured for IPv4 routes. So it'd
>> need to be extended to v6 too, no?
>>
>>
>> Yes, IPv6 and other protocols (not only BGP) can benefit from realms
>> support too.
>>
>> The above need answering/fixing before we can go further I think. Upon
>> which, we'd need to think about:
>>
>> 3. How do we integrate this generally? Is there a general use-case for a
>> route-map setable ID in bgpd?
>>
>>
>> Example of use:
>> There are several 'edge' routers accepting routes and forwarding them
>> to Slave/IGP peers after they are marked with a per-peer realm.
>> On each 'edge' router an IMQ qdisc allows shaping and prioritising the
>> traffic from each peer in both directions separately by matching on
>> it's realm number.
>> In addition a community is set inside a route-map, with the peer's
>> realm and AS number and also with the type of traffic (International,
>> National or Local/City peering)
>> When a Slave receives a route, it may set (inside route-map) a realm
>> based on the originating peer realm/AS or type of traffic and then
>> perform shaping/prioritizing of clients traffic based on that.
>>
>> In short the realm is used as MARK of the route to pass info to iptables.
>>
>> Thanks!
>> _______________________________________________
>> Quagga-dev mailing list
>> [email protected]
>> https://lists.quagga.net/mailman/listinfo/quagga-dev
>>
>> _______________________________________________
>> Quagga-dev mailing list
>> [email protected]
>> https://lists.quagga.net/mailman/listinfo/quagga-dev
>>
>
> _______________________________________________
> Quagga-dev mailing list
> [email protected]
> https://lists.quagga.net/mailman/listinfo/quagga-dev
>
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev

Reply via email to