Re: tc question about ingress bandwidth splitting

On 4/1/20 3:48 AM, Marco Gaiarin wrote:
> Mandi! Grant Taylor
>    In chel di` si favelave...
> 
> 
> Beh, something similar to what i do now for IFB:
> 
>   tc filter add dev eth1 parent ffff: protocol ip prio 50 \
>          u32 match ip src 0.0.0.0/0 \
>          flowid :1 \
>          action mirred egress redirect dev ifb0

ACK

Thank you.

> AH! Sure, i meant 'bridge', not bond, sorry...

;-)

> I suppose, throw away 'ifb' and use veth in place. ;-)

If it makes sense for the need at hand.

> With 'tc' command above, i 'pipe' ingress to ifb; surely i can create 
> a 'route' between phisical and veth interfaces, but clearly i have 
> to manage a bit of routing and so on...

I'd expect to.

> Can you provide me some examples? Thanks.

Sure?

Add a veth interface (pair), bring the local one up, add an IP & subnet 
to it, enable forwarding.  Then on your remote system, add a route to 
the new veth subnet via the eth0 IP.

The uncertainty above is that I doubt that this is what you're asking.

Please provide a hypothetical topology and I'll describe how it could be 
implemented with network namespaces and veth pairs.  (I don't know if 
you are asking for an ifb alternative or something else.)



-- 
Grant. . . .
unix || die