Description of problem: During the normal process of updating to 2.0.30 in PROD, we started getting these errors in the mcollective log on the ex-nodes. Error: I, [2013-07-18T15:01:48.235593 #3413] INFO -- : runner.rb:23:in `block in initialize' Reloading all agents after rec eiving USR1 signal W, [2013-07-18T15:01:48.250785 #3413] WARN -- : agent.rb:234:in `metadata' openshift.rb:23:in `<class:Openshift>': s etting meta data in agents have been deprecated, DDL files are now being used for this information. W, [2013-07-18T15:01:48.251170 #3413] WARN -- : agents.rb:95:in `rescue in activate_agent?' Agent activation check f or openshift failed: NameError: uninitialized constant OpenShift::Runtime::CartridgeRepository I, [2013-07-18T15:01:49.212931 #3413] INFO -- : runner.rb:23:in `block in initialize' Reloading all agents after rec eiving USR1 signal Restarting mcollective fixed this problem. This problem caused the broker cartridge cache to become invalid, which made it so people couldn't create apps. Version-Release number of selected component (if applicable): rhc-node-1.11.7-1.el6oso.x86_64 How reproducible: unkown, but we didn't see this in INT or STG. Steps to Reproduce: 1. Unknown Actual results: Rest API was broken Expected results: We should have been able to upgrade without downtime.
Dan McPherson asked me to add this line from mcolective.log to this bug: E, [2013-07-18T18:45:45.282900 #18584] ERROR -- : pluginmanager.rb:171:in `rescue in loadclass' Failed to load MCollective::Agent::Openshift: cannot load such file -- openshift-origin-node/plugins/unix_user_observer
The fundamental problem is we are updating our source which reloads our mco agent but the puppet changes haven't ran yet. So any config necessary to run isn't available. The solution to this is we can not reload the agent as part of rpm install.
*** This bug has been marked as a duplicate of bug 985514 ***