The bluepill cookbook does not reload services when the template is changed

I have added a :force_load action to the service provider as I am facing the same issue.

The reason I called it :force_load instead of :reload is because it doesn't check whether the pill is already loaded or running, it simply issues the load command and informs the update.

The code can be found in this branch of my private fork: https://github.com/cassianoleal/bluepill/tree/force_load

Please try it out and let us know if it works as you expected or if you find any problems.

Thanks.

I would expect that the solution would add a notify to the template resource to call load_command again.

The only difference between the :load and :force_load actions here appear to be that the latter loads even if the service isn't running, which doesn't sound like the solution to the OPs bug.

• Why do we not load the pill if the service is not running?

• Is there a reason to not have an execute resource that is notified by the template for the pill?

• Not being familiar with bluepill, do we then need to notify the service to reload after the pill is [re]loaded?

Hi Brian,

First of all, thanks for your feedback. I'm still learning Chef, and I benefit from any advice on how to write better cookbooks.

That said, I'm not sure I understand your suggestion.

Let me start with a bit of an explanation about how bluepill works (note that I'm not a bluepill expert either, but I do have some experience with it).

• You create a service descriptor (a "pill"), which is a configuration file that's used by bluepill to know what to do (a bit like a SysV init or upstart script);
• Bluepill reads the file, starts the described service and monitors it;
• If the service is already running and I issue a "bluepill load" command, it re-reads the configuration and updates the service behaviour.

The third bullet is where the cookbook falls short. Since the service is already running, the :load action does nothing and the service keeps running with the old configuration, even though the pill file has changed.

Answering your questions:

• We do not, we actually do the opposite. If the service is running, we do not load the pill;
• This what I'm not sure I understand. That was the point of :force_load – to be notified by a template change and then load the pill even though @current_resource.running == true;
• The bluepill [re]load command takes care of updating the service.

One alternative implementation that worked for me:

action :reload do
shell_out!(stop_command) if @current_resource.running
shell_out!(load_command)
new_resource.updated_by_last_action(true)
end

Is there a way to call other actions from within that action? If so, I could just call :stop and :load subsequently and it would have the same effect.

That change can be found on https://github.com/cassianoleal/bluepill/tree/reload

Why do we only load if the service isn't running? I think we can drop that, but someone more familiar with bluepill should confirm.

I ran a test and while bluepilld will exit when you load, the processes it manages won't.

$ ps ax | grep blue
22814 ?        Sl     0:00 bluepilld: chef-client                                                                      
22988 ?        S      0:00 /usr/bin/ruby1.8 /usr/bin/chef-client -d -P /tmp/chef.bluepill.pid
23286 pts/6    S+     0:00 grep --color=auto blue
$ sudo bluepill status
chef-client(pid:22988): up
$ sudo bluepill load /tmp/test.pill
Killing previous bluepilld[22814]
$ ps ax | grep blue
22988 ?        S      0:00 /usr/bin/ruby1.8 /usr/bin/chef-client -d -P /tmp/chef.bluepill.pid
23322 ?        Sl     0:00 bluepilld: chef-client                                                                      
23358 pts/6    S+     0:00 grep --color=auto blue
$ sudo bluepill status
chef-client(pid:22988): up

I'll ping akzhan on twitter to see what he thinks.

Yes, it's behavior of bluepill for now. It does not make sense to change, because most system administrators are accustomed to such behavior.

Hi Akzhan,

Could you please elaborate on that? Specifically, I'd like to understand what it means in the context of this ticket, since this is functionality that I need.

I personally don't understand why I would not want to reload the service when the definition changes, but the reload action would solve my problem so I'd be happy to go with it.

This hasn't been updated in a while so I took the liberty to change its state to "Fix provided" once more.

The fix is still the same, on https://github.com/cassianoleal/cookbook-bluepill/tree/reload

What motivated me to resolve it again even with the lack of attention given to the issue was that a duplicate has been opened, meaning that more people want this behaviour to be implemented: COOK-2091

Looks fine, none of us are bluepill experts though.

This is merged.

Released in v2.2.0