github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2025-07-23 19:56:01 +00:00
Code Issues Projects Releases Wiki Activity

Merge pull request #39030 from shyamjvs/kubemark-systemd-bug

Browse Source 
Automatic merge from submit-queue

Fixing 'systemd restart docker' command in kubemark master

We try to restart docker after assembling it's flags in kubemark master using `systemd`, while it doesn't exist on the gci image. Changed it to `systemctl` instead.
This means that docker wasn't getting restarted in gci after changing DOCKER_OPTS all this while and hence not flushing the nat iptable didn't really matter.
But now, since we have the flag `--ip-masq=false` in DOCKER_OPTS taking effect, flushing the nat table makes sense, in order to make both of them consistent wrt masquerading.

Also, do you think that this might have anything to do with the failing kubemark tests (if there were some containers that did not want to allow iptable modification / masquerading)?

@kubernetes/sig-scalability-misc @wojtek-t @gmarek
This commit is contained in:
Kubernetes Submit Queue 2016-12-21 04:19:32 -08:00 committed by GitHub
commit d1d3feb6c8
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
test/kubemark/resources/start-kubemark-master.sh
View File

@ -141,7 +141,7 @@ function assemble-docker-flags {
# TODO(shyamjvs): Incorporate network plugin options, etc later.
echo "DOCKER_OPTS=\"${docker_opts}\"" > /etc/default/docker
echo "DOCKER_NOFILE=65536" >> /etc/default/docker # For setting ulimit -n
systemd restart docker
systemctl restart docker
}
# A helper function for loading a docker image. It keeps trying up to 5 times.
@ -225,6 +225,9 @@ function start-kubelet {
local -r kubelet_bin="/usr/bin/kubelet"
create-kubelet-conf "${kubelet_bin}"
# Flush iptables nat table
iptables -t nat -F || true
# Start the kubelet service.
systemctl start kubelet.service
}