forced to run twice due to failing provider detection

[zeph]

err: /Stage[main]/Rabbitmq::Management/Rabbitmq_user[guest]: Provider rabbitmqctl is not functional on this host

u can checkout my https://github.com/zeph/mcollective-lnmp and type "vagrant up" to see it

u need to rekick puppet via "vagrant provision" to see it succeed

[jurgenweber]

run puppet a second time and it works fine, it is just the first time where it fails.

[bodepd]

@zeph

try reverting this patch and see if it works:

https://github.com/puppetlabs/puppetlabs-rabbitmq/commit/0c4ba9bff3f6c4306b9d4eedb6847ad0b9cc2c98

This patch looks like it would result in that behavior

[apenney]

Hmm, this is unexpected (and of course I can't reproduce it easily in testing). I can confirm what you're seeing via vagrant and it's definitely odd. We do a rabbitmqctl list_users -q and look for 'unable to connect to node'. If we get that we fail out the providers.

I can see that the service refreshed and is running prior to the first attempted load of the provider, so this should work. At least my understanding was that providers don't run the :confine block until they are actually called for the first time so it shouldn't have done it up front and cached that result past the start of the service.

I'm torn over the right thing to do here. We want the provider to fail if the service isn't up so users don't have to specify their own requires, but if this is going to fail on every first run then it won't do. I'll dig into this a little more and see if I can find out exactly what is happening.

[jurgenweber]

Yeah, I can confirm reverting that code resolved the issue as well.

Thanks

[apenney]

I'll give in and switch these over to just autorequires the service. I wanted to avoid that as I couldn't autorequires the service but it seems like a cleaner solution and should hopefully work for a first run.

Testing with your vagrant and the changes:

Notice: /Stage[main]/Roles::Middleware/Rabbitmq_user[mcollective]/ensure: created

I'll get a PR in to make these changes.

[zeph]

awesome! tnx @apenney & @bodepd 👍

[apenney]

@zeph #96 - give that a whirl, seems to work better for me. If you have success I'll merge it.

[jurgenweber]

Yeah, works great for me. Thanks.

[apenney]

Awesome, closing this for now then!