Right now we are suspecting that this is caused by issues with RabbitMQ. If a message for some reason isn't delivered to the compute (e.g. broken bindings) the instance will never be built and indefinitely stuck in BUILD.
Even worse is that there is nothing in the logs, and there is nothing in the api that would indicate the compute with issues. It would be helpeful to set the compute before the message is sent to the compute, or maybe change the RabbitMQ message to be an RPC, so that it will timeout.
Another idea would be to make the message sent mandatory.
Right now we are suspecting that this is caused by issues with RabbitMQ. If a message for some reason isn't delivered to the compute (e.g. broken bindings) the instance will never be built and indefinitely stuck in BUILD.
Even worse is that there is nothing in the logs, and there is nothing in the api that would indicate the compute with issues. It would be helpeful to set the compute before the message is sent to the compute, or maybe change the RabbitMQ message to be an RPC, so that it will timeout.
Another idea would be to make the message sent mandatory.