Daniel Borkmann
Daniel Borkmann
(Wrt 'broken' links, some also still point to https://github.com/cilium/ebpf.io location.)
I think you might need to set it the other way round.. configure Cilium with handle 0x2 (there's a hidden agent config: https://github.com/cilium/cilium/pull/18896) and the datadog agent as handle 0x1...
Two entries are being created in this flow, here's an example (in a host with a pod CIDR of 10.138.186.0/24) where 10.141.118.153:3051 is an endpoint of a service which lives...
> Hi @borkmann, thanks for chiming in :) > > Indeed that would solve the problem, that's the alternate solution I proposed above (different to the one in the MR...
We just had a discussion internally that this commit could already resolve the case around delete events (which then trigger the internal endpoint cleanup already): https://github.com/cilium/cilium/commit/54fef7acd341a4dd95848f9843773ba407c55499 Cc'ing @squeed to confirm....
> That will use the host's routing table, which might work. Could be worth a try, I'm not sure if netfilter would support MPLS (+ VLAN) + masquerading in the...
Hello, sorry for the delay due to conference travel.. digesting this thread a bit: [...] > In the Bandwidth Manager code, I can see that it [sets](https://github.com/cilium/cilium/blob/13a497a6b56d57061beb39ed9155389a17970501/pkg/bandwidth/bandwidth.go#L117) `net.core.default_qdisc` to `fq`...
> > I am still curious to understand the performance issues I saw with fq when the Bandwidth Manager is enabled since we are not going to migrate to 6.7...