yahoo-eng-team team mailing list archive

[OpenStack Infra <1316928@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.openstack.org/290315
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=bd508a5a0e2462292cb3ebfa3ba851e8a1a2a01f
Submitter: Jenkins
Branch:    master

commit bd508a5a0e2462292cb3ebfa3ba851e8a1a2a01f
Author: Giridhar Jayavelu <gjayavelu@xxxxxxxxxx>
Date:   Tue Mar 8 23:13:14 2016 -0800

    VMware: set service status based on vc connection
    
    When the vcenter (vc) session is disconnected, say,
    if the management service on vcenter is down, then
    the nova-compute service status does not reflect that.
    
    This patch sets disabled=True in nova.objects.Service
    when the vcenter session is broken.
    When vcenter service is accessible again, then oslo_vmware
    takes care of recreating a new session.
    
    Closes-Bug: #1316928
    Change-Id: I73a34eb6e0ca32d03e54d12a5e066b2ed4f19a61


** Changed in: nova
       Status: In Progress => Fix Released

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1316928

Title:
  VMware driver does not change compute node state to unavailable after
  disconnecting from vCenter

Status in OpenStack Compute (nova):
  Fix Released

Bug description:
  VMware driver does not change compute node state (from :-) to XXX)
  when it lost connection to vCenter.

  api_retry_count does not affect it at all.

  When connection to vCenter is lost and nova-compute is restarted it
  works. But it does not without restarting. After restart state is
  changed to XXX and nova-compute is still polling API connection (good
  behavior) but without restart nova-compute has still smile state
  despite of " SessionConnectionException: NV-06475AF urllib2 error in
  RetrieveServiceContent: : <urlopen error [Errno 113] EHOSTUNREACH>"

  Steps to reproduce:
  1. Configure nova-compute to connect to VMware vCenter using vmware community driver.
  2. Check state of compute-node, it should be available
  3. Disconnect vCenter - for example just disconnect network from it
  4. Check state of compute-node, it still be available

  Expected result:
  compute-node should be unavailable when vCenter is disconnected (maybe after api_retry_count?) without need to restart nova-compute.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1316928/+subscriptions