Crash dump with standalone OS X package

[michaelklishin]

See this thread.

Even though RabbitMQ starts and stops successfully, something still produces a crash dump (I'm attaching one).

Steps to reproduce:

• Run ./sbin/rabbitmq-server -detached with the standalone OS X package
• Observe erl_crash.dump in the current directory.
• ./sbin/rabbitmqctl status suggests the node is fine
• Running various quick tests suggests the node is fine
• ./sbin/rabbitmqctl stop stops the node successfully

[videlalvaro]

There's no need to start/stop RabbitMQ to reproduce this issue. Just running rabbitmqctl will generate the crash dump

[videlalvaro]

As explained in our internal bug system here https://bugzilla.rabbitmq.com/show_bug.cgi?id=26441#c3 we need to pass the boot file explicitly when using the standalone release in every place where we call erl.

The code here https://github.com/rabbitmq/rabbitmq-server/blob/master/src/rabbit_nodes.erl#L207 doesn't pass this option, therefore Erlang crashes since it can't find the boot script.

Adding: "-boot", "/path/to/start_clean" to the list of arguments passed to open_port solves the problem. I'll try to fix this once I'm back.

[vitaly-krugl]

Good point @videlalvaro, I just noticed that simply running rabbitmqctl status produces a crash dump.

[michaelklishin]

The erl call that fails is supposed to force start epmd (if needed) because of certain platform-specific issues. So this is completely harmless.