Mistral components don't start properly with Kombu RPC

[Renat Akhmerov]

Engine, executor and event engine don't start properly when using Kombu RPC.

This happens because the method KombuServer.run() implemented as an infinite loop rather than a thread.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/500504

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/500504
Committed: https://git.openstack.org/cgit/openstack/mistral/commit/?id=2daa3c0ab2429c6c412d4df357fa54c78e60ef2e
Submitter: Jenkins
Branch: master

commit 2daa3c0ab2429c6c412d4df357fa54c78e60ef2e
Author: Renat Akhmerov <email address hidden>
Date: Mon Sep 4 14:11:06 2017 +0700

    Fix Kombu RPC threading and use within multiprocess environment

    * Added a call to clean up RPC clients into WSGIService start()
      method to make sure that we reinitialize RPC in every child
      API process. Otherwise, replies for synchornous RPC calls
      don't get routed to the right client process (see bug
      description)
    * Mistral components engine, executor and event engine didn't
      start properly because KombuServer.run() was implemented
      w/o a new thread. This patch fixes it.
    * The methods to get RPC clients like
      mistral/rpc/clients.py:get_engine_client() wasn't synchronized
      on the client instances. As a consequence, sometimes Mistral
      created more than one instance of the same client (violation
      of Singleton pattern). This patch fixes it by adding required
      locking.
    * Other style changes

    Closes-Bug: #1714186
    Closes-Bug: #1714929

    Change-Id: I242c48df9811459321d1cd4b1d37ed35af0a013a

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/pike
Review: https://review.openstack.org/501678

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/501678
Committed: https://git.openstack.org/cgit/openstack/mistral/commit/?id=ec1227771f84d8ad706d26dcb5d3724b5be85205
Submitter: Jenkins
Branch: stable/pike

commit ec1227771f84d8ad706d26dcb5d3724b5be85205
Author: Renat Akhmerov <email address hidden>
Date: Mon Sep 4 14:11:06 2017 +0700

    Fix Kombu RPC threading and use within multiprocess environment

    * Added a call to clean up RPC clients into WSGIService start()
      method to make sure that we reinitialize RPC in every child
      API process. Otherwise, replies for synchornous RPC calls
      don't get routed to the right client process (see bug
      description)
    * Mistral components engine, executor and event engine didn't
      start properly because KombuServer.run() was implemented
      w/o a new thread. This patch fixes it.
    * The methods to get RPC clients like
      mistral/rpc/clients.py:get_engine_client() wasn't synchronized
      on the client instances. As a consequence, sometimes Mistral
      created more than one instance of the same client (violation
      of Singleton pattern). This patch fixes it by adding required
      locking.
    * Other style changes

    Closes-Bug: #1714186
    Closes-Bug: #1714929

    Change-Id: I242c48df9811459321d1cd4b1d37ed35af0a013a
    (cherry picked from commit 2daa3c0ab2429c6c412d4df357fa54c78e60ef2e)

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/mistral 6.0.0.0b1 development milestone.

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/mistral 5.2.0 release.