Use redhat init instead of "init" init #248
Conversation
|
I should've actually tried this patch on my box. Looks like it doesn't work because of config.pp line 82. I commented out the provider line in run_service.pp. Is the provider necessary? i.e. Commenting it out on CentOS 6 works, but isn't clean for all the other platforms. Any reason not to trust the default puppet service provider? |
|
Yea, we basically can't make everyone happy. See #125 |
|
@solarkennedy - #125 also introduces a break for Amazon linux. |
|
init provider is a poor choice, it does not support enable. https://docs.puppet.com/puppet/latest/reference/type.html#service-providers Perhaps #125 should have only introduced changes for debian via an if block? |
|
Well, if we know we are on redhat then I guess it should be safe to use the redhat provider (baring systemd) I'm convinced. @benpillet I'll merge this in as soon as you fix the tests and rebase. |
|
fixed problems for me as well on CentOS 6.8 |
https://docs.puppetlabs.com/puppet/3.8/reference/types/service.html#service-provider-redhat
I had CentOS 6 nodes in an autoscaling group. When they would get scaled down/terminated, consul would show them as failed. I had leave_on_terminate set to true, but I was still seeing the error. When I ran chkconfig on the box, the consul service wasn't listed. So, the K20consul script was not getting called in /etc/rc6.d and the node wasn't leaving the cluster. Please correct me if I'm wrong, but I suspect this provider is the right one for the operatingsystem entries listed. Also, is this even necessary as puppet is supposed to pick the "best" init_style for a service anyway. Should most of these elsif entries just be removed?