RabbitMQ Supervisor Structure
Global supervisor started after the
rabbit app has started.
-- rabbit_node_monitor_sup ---- rabbit_node_monitor Notifies other nodes in the clusters about its own node presence. Takes cares of the situation when other node dies.
-- msg_store_transient ---- rabbit_msg_store
-- msg_store_persistent ---- rabbit_msg_store
Both handle the message storage.
-- rabbit_tcp_client_sup See drawing on paper. Started during the networking boot step. See rabbit_networking:start/0
-- rabbit_direct_client_sup Started during the direct_client boot step. See rabbit_direct:boot/0
-- tcp_listener_sup <--- not registered. This supervisor structure is repeated for each network interface on which RabbitMQ listens too ---- tcp_acceptor_sup_IP:PORT ------ tcp_acceptor ---- tcp_listener
-- rabbit_amqqueue_sup ---- rabbit_amqqueue_process <- one child per queue?
-- rabbit_mirror_queue_slave_sup ----- rabbit_mirror_queue_slave <-- one child per queue? Used for HA queues.
-- rabbit_guid_sup ---- rabbit_guid Globally Unique Identifier server.
-- rabbit_memory_monitor_sup ---- rabbit_memory_monitor Monitors queues memory usage.
-- delegate_sup <-- instance of rabbit_sup.erl ---- delegate_x <-- x goes 0-15 according to app config argument: delegate_count. Used to parallelize calls to processes. For example when routing messages, the delegates take care of sending the messages to each of the queues that ought to receive the message.
-- rabbit_registry Keeps a registry of plugins and their modules. For example it maps authentication mechanisms to modules with the actual implementation. The same thing is done from exchange type to exchange type implementation.
-- rabit_log_sup ---- rabbit_log Delegates logging calls to the native error_logger module.
-- vm_memory_monitor_sup ---- vm_memory_monitor Monitors Erlang's memory usage to prevent it from crashing or swapping.
Handles event notification for statistics collection. For example when a new channel is created, then a notification like
rabbit_event:notify(channel_created, infos(?CREATION_EVENT_KEYS, State)) is fired.
-- file_handle_cache_sup <-- instance of rabbit_sup.erl
Manages file handles to synchronize reads and writes to them. See
file_handle_cache.erl for an in depth explanation of its purpose.
-- worker_pool_sup <-- instance of rabbit_sup.erl
The worker pool process manages a pool of up to
N number of workers where
N is the return of
erlang:system_info(schedulers). Used to parallelize function calls.