Machines are converging twice about chef-provisioning

[jkeiser]

Yeah, two converges currently happen (will have that fixed soon). This is because the default machine_batch converges the machine, and then chef still runs the machine resource. The second converge, however, shouldn't really do anything--are you encountering errors?

To turn this off for now, you can put this just before your first machine declaration: with_machine_batch 'thenewbatch', :action => :nothing . I'll make sure it doesn't happen in the general release.

[mwrock]

Adding the with_machine_baytch line did pevent the second convergence but also fails since the machine conversion uses a different strategy for retrieving the driver. The machine resource instantiates a fresh driver from the driver URL and the client config. Since I retrieve credentials at the LWRP and add to the runcontext, those credentials are not not present.

The machine_batch convergence is now eliminated but that was the one that had been succeeding because it gets the driver from the run context that hd been previously stuffed with the appropriate credentials.

I would expect that credentials set in the recipe would override any in the config and therefore it should be acceptable to set creds in the recipe and have them absent from the config. if creds are missing from both, then that would be the error condition.

My PR seems to follow that logic. If a driver is constructed in the recipe and olaces in the runcontext that is used and if it is not declared in the recipe, then it falls back to the driver url and config data.

[jkeiser]

In those cases, the right thing to do is probably to cache the driver for a given URL, so that when you first specify it, all subsequent people will get the same driver instance. That way all callsites will get the same response.

You don't need with_machine_batch anymore to avoid multiple converges, FYI. That's fixed.

[jkeiser]

er, that was not intentional

[jkeiser]

Machines are no longer converging twice.