in daemon mode, master swallows errors during startup
|Affected Puppet version:||Branch:||https://github.com/puppetlabs/puppet/pull/621|
Because of the way that logging is handled / initialized during master startup, failures that occur before the logging is completely initialized seem to just disappear into the void. This makes it very tricky to diagnose problems that happen during startup, e.g. during acceptance testing on the CI nodes.
Would be extremely useful to find a way to get the master to keep STDOUT/STDERR alive until logging has been successfully initialized, or register a “bootstrapping” log destination for use during startup so that these errors don’t just disappear.