Uploaded image for project: 'Camel'
  1. Camel
  2. CAMEL-14789

camel-rabbitmq - Automatic recovery of temporary reply queue is not handled correctly

    XMLWordPrintableJSON

    Details

    • Type: Bug
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: 2.25.0
    • Fix Version/s: 3.2.0, 2.25.1, 3.1.1
    • Component/s: camel-rabbitmq
    • Labels:
      None
    • Environment:
      • RabbitMQ server 3.8.3
      • Erlang 22.3
    • Patch Info:
      Patch Available
    • Estimated Complexity:
      Moderate

      Description

      When Remote Procedure Call (RPC) communication pattern is used RabbitMQ Camel component creates server-named auto-deleted temporary reply queue. Next, the queue is bound to the exchange with routing key equal to queue name. Likewise, ReplyManager's replyTo field is set with the value of the queue name. Reply queue created in this manner is then reused for subsequent RPC requests and message property rabbitmq.REPLY_TO is copied from the value stored previously by ReplyManager.

      When some network error suddenly appears causing connection failure, RabbitMQ's automatic recovery kicks in and tries to recover affected entities (assuming the connection was created with automatic recovery enabled, which is default.). As temporary quueues are auto-deleted, during recovery process a new temporary queue is created which has a new name that differs from the original one and here is where the problem begins.

      Creation of the new temporary queue is NOT detected by ReplyManager and therefore replyTo property is NOT updated. Also, routing key no longer matches queue name. This causes a problem when some implementations of RabbitMQ client, like. Spring AMQP, are used server-site. RPC service receives our request, processes it, and replies to default exchange with routing key equal to rabbitmq.REPLY_TO sent in our request. RPC service provider perceives no problem as response is sent successfully but RabbitMQ Camel Component keeps awaiting for the response to arrive to the original temporary queue which no longer exists due to connection failure and recovery. Eventually, Camel throws ExchangeTimedOutException.

      Example:

      1. After automatic recovery, a RPC request is sent to example-exchange with rabitmq.REPLY_TO=amq.gen-0lLvpnj4ZMlkhxZIcCPVpA
      2. After 20 seconds org.apache.camel.ExchangeTimedOutException is thrown:
      The OUT message was not received within: 20000 millis due reply message with correlationID: Camel-ID-hostname-1583401848872-0-1690397 not received on destination: amq.gen-0lLvpnj4ZMlkhxZIcCPVpA. Exchange[ID-hostname-1583401848872-0-1690391]
      1. Using RabbitMQ Management WebApp (rabbitmqctl) we check example-exchange's bindings and see that routing key amq.gen-0lLvpnj4ZMlkhxZIcCPVpA now corresponds to queue amq.gen-zaRCP-p-JbXeSzJmzSp83g

      Proposed solution is to add QueueRecoveryListener to notify when temporary queue name changes due to recovery. On event, replyTo field will be updated with the new temporary queue name and the queue will be rebound to the exchange so that routing key matches queue name again. The change will be made to org.apache.camel.component.rabbitmq.reply#createListenerContainer.

      Attached patch also contains integration test.

        Attachments

          Activity

            People

            • Assignee:
              acosentino Andrea Cosentino
              Reporter:
              rszczesiak Robert Szczesiak
            • Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Time Tracking

                Estimated:
                Original Estimate - Not Specified
                Not Specified
                Remaining:
                Remaining Estimate - 0h
                0h
                Logged:
                Time Spent - 40m
                40m