Kube 1.11 CoreDNS + Ubuntu Bionic / Linux 4.15 => parallel DNS A+AAAA query timeouts due to response packets getting NF_DROP per nf_conntrack_confirm
Related but different issues:
- https://www.weave.works/blog/racy-conntrack-and-dns-lookup-timeouts
- https://blog.quentin-machu.fr/2018/06/24/5-15s-dns-lookups-on-kubernetes/
AFAIK these describe the outging DNS query packets getting dropped per nf_conntrack_confirm collisions, but this is about the DNS response packets getting dropped due to nf_conntrack_confirm
returning NF_DROP
for the incoming response packet.
This increaes the duplicate get_unique_tuple
race window from µs for the outgoing packet path to ms for the DNS server's query RTT.