Add ntp to etcd image

[danielschlegel]

I get the folowing error on etcd in kubernetes environment:
rafthttp: the clock difference against peer a5307ab2af371d91 is too high [5.028292803s > 1s]

[LLParse]

We've typically left time synchronization up to the end-user to configure on their hosts. Many OS images in cloud environments come with it by default. There is no silver-bullet configuration for this (someone running in an air-gapped environment might not have access to time servers, etc).

[cloudnautique]

@danielschlegel configuring time is done at the server host level, and not at a container level. It would be outside of the scope of the application to to change the host time, though possible.

The best approach to fix that would be to configure the hosts properly with a config management tool.

[LLParse]

It would be cool, however, if somebody built a template to run ntpd in a privileged container using label: io.rancher.scheduler.global: 'true'...that would be a super slick, reusable template for synchronizing time across an environment where ntpd is non-existent.

This sounds like a great excuse for some community engagement!

[danielschlegel]

Thanks for the answers. @LLParse A container with label: io.rancher.scheduler.global: 'true' would be great yes. Im using rancher with kubernetes and will try to do something like this but first i need to dive a little deeper into my rancher setup.