Rudder

Matthieu CERDA wrote:

Well, I do not see where an interference might be possible, I am only injecting the same variables the restart bundle uses with another parameter forcing the restart, could you please describe a complete use case ?

Sure. This is not related to variables, but to classes. In each CFEngine run, you call the generic_process_check_process bundle 4 times (once for slapd, once for jetty, etc). Each time, the bundle checks a process, and may try and restart it. It then sets one of the classes process_restart_ok or process_restart_error, which is then used for reporting in that same bundle. Classes defined as the result of a promise, like these ones (from a classes body), are global.

This means that the first time you call the bundle, for slapd, if there is an error restarting the process, the class "process_restart_error" will be defined, and reporting will say "result_error...etc...The slapd process couldn't be restarted". However, when the bundle is next called, for jetty, this class will still be defined, causing the reporting to say "result_error...etc...The jetty process couldn't be restarted", even if the jetty process is fine and other reports for it are printed. And so on...

About the permissions, I totally agree but I copied the permissions that are already in place afterthe installation. I am afraid to break Rudder with access denied/permissions errors with Apache, slapd ...

I understand. But if this is the case, it should be considered a bug! Now is the time to find out, while 2.4 is still in alpha stage. Please test with 600.

About the permissions, please see #2211, I proposed a solution.

About the restart classes, I'm OK with that. I will address this as soon as a compromise is found about file permissions.