Package org.apache.cassandra.net

Class AbstractMessageHandler.WaitQueue

java.lang.Object
org.apache.cassandra.net.AbstractMessageHandler.WaitQueue
Enclosing class:
AbstractMessageHandler
public static final class AbstractMessageHandler.WaitQueue extends Object
A special-purpose wait queue to park inbound message handlers that failed to allocate reserve capacity for a message in. Upon such failure a handler registers itself with a AbstractMessageHandler.WaitQueue of the appropriate kind (either ENDPOINT or GLOBAL - if failed to allocate endpoint or global reserve capacity, respectively), stops processing any accumulated frames or receiving new ones, and waits - until reactivated. Every time permits are returned to an endpoint or global ResourceLimits.Limit, the respective queue gets signalled, and if there are any handlers registered in it, we will attempt to reactivate as many waiting handlers as current available reserve capacity allows us to - immediately, on the signal()-calling thread. At most one such attempt will be in progress at any given time. Handlers that can be reactivated will be grouped by their EventLoop and a single AbstractMessageHandler.WaitQueue.ReactivateHandlers task will be scheduled per event loop, on the corresponding event loops. When run, the AbstractMessageHandler.WaitQueue.ReactivateHandlers task will ask each handler in its group to first process one message - using preallocated reserve capacity - and if no obstacles were met - reactivate the handlers, this time using their regular reserves. See schedule(), Runnable.run(), AbstractMessageHandler.WaitQueue.Ticket.reactivateHandler(Limit).