Comment 1 for bug 1632043

'cannot fork' is normally a symptom of the machine running out of memory - the OS is basically stating 'no way' when the new process creation is attempted resulting in the fork failing in this way.

The 2.0 rc's did have some memory leaks so its possible that this has already been resolved, 2.1 is improving the situation further (as 2.0.2 also has some residual leaking).

I'm going to mark this as 'Invalid' as I don't think its a charm specific issue - either the machine is under specified or something is consuming all of the memory over time.

If you're still seeing this problem, please open a task on this bug report against the juju project.

Worth nothing that RMQ memory usage will also grow based on a) the number of clients connection and b) the number of instances on the cloud - the regular syncs between hypervisor and controller are done with a good level of parallism, so the more instances you have, the more connections created on the RMQ message bus.