kubernetes/test/images/regression-issue-74839
Davanum Srinivas 50bea1dad8
Move from k8s.gcr.io to registry.k8s.io
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
2022-05-31 10:16:53 -04:00
..
.gitignore
BASEIMAGE Move from k8s.gcr.io to registry.k8s.io 2022-05-31 10:16:53 -04:00
Dockerfile e2e test regression image multiarch 2020-11-22 10:46:38 +01:00
main.go run hack/update-netparse-cve.sh 2021-08-20 10:42:09 +02:00
Makefile test images: Adds Windows support (part 1) 2020-02-21 02:09:49 -08:00
OWNERS add ipv6 support to the image issue-74839 2020-10-07 18:30:47 +02:00
README.md
tcp_test.go add ipv6 support to the image issue-74839 2020-10-07 18:30:47 +02:00
tcp.go add ipv6 support to the image issue-74839 2020-10-07 18:30:47 +02:00
VERSION add ipv6 support to the image issue-74839 2020-10-07 18:30:47 +02:00

Reproduction of k8s issue #74839

Network services with heavy load will cause "connection reset" from time to time. Especially those with big payloads. When packets with sequence number out-of-window arrived k8s node, conntrack marked them as INVALID. kube-proxy will ignore them, without rewriting DNAT. The packet goes back the original pod, who doesn't recognize the packet because of the wrong source ip, end up RSTing the connection.

Reference

https://github.com/kubernetes/kubernetes/issues/74839