yahoo-eng-team team mailing list archive

Thread
Date

[Bug 1261631] Re: Reconnect on failure for multiple servers always connects to first server

To: yahoo-eng-team@xxxxxxxxxxxxxxxxxxx
From: Ihar Hrachyshka <1261631@xxxxxxxxxxxxxxxxxx>
Date: Mon, 31 Mar 2014 14:29:00 -0000
Reply-to: Bug 1261631 <1261631@xxxxxxxxxxxxxxxxxx>
Sender: bounces@xxxxxxxxxxxxx

** Also affects: glance
   Importance: Undecided
       Status: New

** Also affects: ceilometer
   Importance: Undecided
       Status: New

** Also affects: cinder
   Importance: Undecided
       Status: New

** Also affects: heat
   Importance: Undecided
       Status: New

** Also affects: ironic
   Importance: Undecided
       Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1261631

Title:
  Reconnect on failure for multiple servers always connects to first
  server

Status in OpenStack Telemetry (Ceilometer):
  New
Status in Cinder:
  New
Status in OpenStack Image Registry and Delivery Service (Glance):
  New
Status in Orchestration API (Heat):
  New
Status in Ironic (Bare Metal Provisioning):
  New
Status in OpenStack Identity (Keystone):
  New
Status in OpenStack Neutron (virtual network service):
  In Progress
Status in neutron havana series:
  Fix Committed
Status in OpenStack Compute (Nova):
  New
Status in Oslo - a Library of Common OpenStack Code:
  Fix Committed
Status in oslo havana series:
  Fix Committed
Status in Messaging API for OpenStack:
  Fix Released

Bug description:
  In attempting to reconnect to an AMQP server when a communication
  failure occurs, both the qpid and rabbit drivers target the configured
  servers in the order in which they were provided.  If a connection to
  the first server had failed, the subsequent reconnection attempt would
  be made to that same server instead of trying one that had not yet
  failed.  This could increase the time to failover to a working server.

  A plausible workaround for qpid would be to decrease the value for
  qpid_timeout, but since the problem only occurs if the failed server
  is the first configured, the results of the workaround would depend on
  the order that the failed server appears in the configuration.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ceilometer/+bug/1261631/+subscriptions