[Bug 1870619] Re: rabbitmq-server startup does not wait long enough

Eric Desrochers eric.desrochers at canonical.com
Tue Apr 7 18:16:18 UTC 2020


IIUC, this is your own patch (you being the author).

* Does upstream maintain the debian/ folder ? 
 ** If yes, can you please make sure to submit a PR upstream ?
* Can you file a bug and submit the patch to Debian as well ?
 ** https://wiki.debian.org/BugReport

Thanks !

-- 
You received this bug notification because you are a member of Ubuntu
OpenStack, which is subscribed to rabbitmq-server in Ubuntu.
https://bugs.launchpad.net/bugs/1870619

Title:
  rabbitmq-server startup does not wait long enough

Status in OpenStack rabbitmq-server charm:
  New
Status in rabbitmq-server package in Ubuntu:
  Confirmed
Status in rabbitmq-server source package in Bionic:
  Confirmed
Status in rabbitmq-server source package in Eoan:
  Confirmed
Status in rabbitmq-server source package in Focal:
  Confirmed

Bug description:
  [Impact]

   * Rabbitmq-server has 2 configuration settings that affect how long it will wait for the mnesia database to become available
   * The default is 30 seconds x 10 retries = 300 seconds
   * The startup wrapper rabbitmq-server-wait will wait only 10 seconds
   * If the database does not come online within 10 seconds the startup script will fail despite the fact that rabbitmq-server is still waiting for another 290 seconds.
   * This behavior leads to falsely identified failures in OpenStack for example when a Rabbitmq cluster is restarted out of order (LP: #1828988)

  [Test Case]

   * Create Rabbitmq cluster and create a queue with "ha-mode: all" policy
   * Shut down nodes one by one
   * Restart the node that was shut down first
   * This node will fail to start because it was not the master of the queue
   * Note that the startup script (SysV or systemd) will fail after 10 seconds while the rabbitmq-server process is still waiting for the database to come online

  [Regression Potential]

   * This change potentially increases the time the rabbitmq-server service takes to start up which might lead to failures down the dependency chain of startup services.
   * This change potentially changes the result of starting the rabbitmq-server service in case the mnesia database takes more than 10 seconds to come online. Without this change, the service will incorrectly fail while it will succeed with this change.

To manage notifications about this bug go to:
https://bugs.launchpad.net/charm-rabbitmq-server/+bug/1870619/+subscriptions



More information about the Ubuntu-openstack-bugs mailing list