Comment 100 for bug 1789177

Reviewed: https://review.opendev.org/c/openstack/oslo.messaging/+/749193
Committed: https://opendev.org/openstack/oslo.messaging/commit/b2acc6663f6c3f60e07cdeb1eae97fd1210a4d81
Submitter: "Zuul (22348)"
Branch: stable/stein

commit b2acc6663f6c3f60e07cdeb1eae97fd1210a4d81
Author: shenjiatong <email address hidden>
Date: Fri Jul 3 15:51:21 2020 +0800

    Cancel consumer if queue down

    Previously, we have switched to use default exchanges
    to avoid excessive amounts of exchange not found messages.
    But it does not actually solve the problem because
    reply_* queue is already gone and agent will not receive callbacks.

    after some debugging, I found under some circumstances
    seems rabbitmq consumer does not receive basic cancel
    signal when queue is already gone. This might due to
    rabbitmq try to restart consumer when queue is down
    (for example when split brain). In such cases,
    it might be better to fail early.

    by reading the code, seems like x-cancel-on-ha-failover
    is not dedicated to mirror queues only, https://github.com/rabbitmq/rabbitmq-server/blob/master/src/rabbit_channel.erl#L1894,
    https://github.com/rabbitmq/rabbitmq-server/blob/master/src/rabbit_channel.erl#L1926.

    By failing early, in my own test setup,
    I could solve a certain case of exchange not found problem.

    Change-Id: I2ae53340783e4044dab58035bc0992dc08145b53
    Related-bug: #1789177
    Depends-On: https://review.opendev.org/#/c/747892/
    (cherry picked from commit 196fa877a90d7eb0f82ec9e1c194eef3f98fc0b1)
    (cherry picked from commit 0a432c7fb107d04f7a41199fe9a8c4fbd344d009)
    (cherry picked from commit 5de11fa752ab8e37b95b1785f4c71210bf473f0c)