Mailing List Archive

Seeing the same thing - shut a peer, still seeing those routes via
3356 across multiple route-servers & looking glasses.

So, do we disconnect 3356 and suffer the blackhole in hopes that it
will eventually withdraw those routes, or leave it on to prevent
blackholing but suffer massive packet loss to other carriers? Wow.

On Sun, Aug 30, 2020 at 6:28 AM Lukas Tribus via Outages
<outages@outages.org> wrote:
>
> As previously mentioned by Stephen Flynn, 3356 does not WITHDRAW stale
> bgp routes, can be confirmed with AT&T's route server at (telnet
> route-server.ip.att.net).
>
> Stale routes from 1 hour + are still announced by 3356.
>
>
> This is causing blackholing.
> _______________________________________________
> Outages mailing list
> Outages@outages.org
> https://puck.nether.net/mailman/listinfo/outages
_______________________________________________
Outages mailing list
Outages@outages.org
https://puck.nether.net/mailman/listinfo/outages

I tried bringing two circuits back up hoping to not have blackholing, one never left idle, the other received <2000 routes, which I know from prior outages to mean that the entire region (Tampa Bay) has been BGP isolated from the rest of their network. Good times...

?On 8/30/20, 8:43 AM, "Outages on behalf of randal k via Outages" <outages-bounces@outages.org on behalf of outages@outages.org> wrote:

Seeing the same thing - shut a peer, still seeing those routes via
3356 across multiple route-servers & looking glasses.

So, do we disconnect 3356 and suffer the blackhole in hopes that it
will eventually withdraw those routes, or leave it on to prevent
blackholing but suffer massive packet loss to other carriers? Wow.

On Sun, Aug 30, 2020 at 6:28 AM Lukas Tribus via Outages
<outages@outages.org> wrote:
>
> As previously mentioned by Stephen Flynn, 3356 does not WITHDRAW stale
> bgp routes, can be confirmed with AT&T's route server at (telnet
> route-server.ip.att.net).
>
> Stale routes from 1 hour + are still announced by 3356.
>
>
> This is causing blackholing.
> _______________________________________________
> Outages mailing list
> Outages@outages.org
> https://puck.nether.net/mailman/listinfo/outages
_______________________________________________
Outages mailing list
Outages@outages.org
https://puck.nether.net/mailman/listinfo/outages

_______________________________________________
Outages mailing list
Outages@outages.org
https://puck.nether.net/mailman/listinfo/outages

Watching closely, after applying ^3356$ on our inbound routes, I can
see that they are bouncing sessions and slowly adding in prefixes -- I
have 17x 3356-originated routes in PA, and 893x in CO. And they have
reset the PA BGP session numerous times.

On Sun, Aug 30, 2020 at 6:33 AM randal k <rkohutek+outages@gmail.com> wrote:
>
> Seeing the same thing - shut a peer, still seeing those routes via
> 3356 across multiple route-servers & looking glasses.
>
> So, do we disconnect 3356 and suffer the blackhole in hopes that it
> will eventually withdraw those routes, or leave it on to prevent
> blackholing but suffer massive packet loss to other carriers? Wow.
>
> On Sun, Aug 30, 2020 at 6:28 AM Lukas Tribus via Outages
> <outages@outages.org> wrote:
> >
> > As previously mentioned by Stephen Flynn, 3356 does not WITHDRAW stale
> > bgp routes, can be confirmed with AT&T's route server at (telnet
> > route-server.ip.att.net).
> >
> > Stale routes from 1 hour + are still announced by 3356.
> >
> >
> > This is causing blackholing.
> > _______________________________________________
> > Outages mailing list
> > Outages@outages.org
> > https://puck.nether.net/mailman/listinfo/outages
_______________________________________________
Outages mailing list
Outages@outages.org
https://puck.nether.net/mailman/listinfo/outages