Comment 16 for bug 1717590
Revision history for this message
| Ryan Beisner (1chb1n) wrote Re: percona-cluster unit constantly fires hooks, blocks hacluster unit from being setup | #16 |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
513534c
demo site
(Get the code!)
Hi @jason-hobbs - I think there are 3 things happening here.
Dmitrii's comments #10 + #14 represent an engineering assessment that the telegraf charm looks to have a scale/efficiency and/or potential design issue. That should likely be a separate bug. @dmitriis, can you raise that bug?
We agree there is a separate bug re: #11 (LP 1718247). Indeed that condition could be handled more gracefully, and we're talking about a way to guard that. But that won't cause a lower-level juju agent not to initialize.
While those ^ may be compounding this, the lack of a juju agent on one of the units in question is the essence of this bug as reported and observed. If all agents are good and this issue still occurs, that's naturally a different position.
From a Juju model perspective, the principle (mysql aka pxc) is all good from an agent and application state perspective, which means the juju agents for the subordinates should be clear to init as well.
However, one of the subordinates doesn't have a juju agent, so the charm code isn't a factor at the time the condition is observed.