tripleo

  1. Bug #1651704
  2. Comment #2

Comment 2 for bug 1651704

Revision history for this message
Lucas Alvares Gomes (lucasagomes) wrote : Re: Overcloud deployment fails when node introspection is enabled

Thanks for the report. Yeah I can see that one node got deployed correctly [0] but there's no trace from a second node being deployed in the Ironic logs, which indicates that it failed prior to that (when nova is scheduling) so "No valid hosts" makes sense.

One possible cause for this could be: When the nodes are updated in Ironic (after inspection for example) it takes a while for the values to get propagated to the nova resource tracker. This happens because nova runs a period task that to update it and, as you may know, periodic tasks in OpenStack can be slow (because they run one after the other in a greenthread). In the Ironic DevStack pluging we have a workaround for this: https://github.com/openstack/ironic/blob/master/devstack/lib/ironic#L1151-L1167

Perhaps we need something similar in OOO-quickstart ?

[0]

2016-12-21 06:50:39.867 29447 INFO ironic.conductor.task_manager [req-f77a2cef-c3e7-4863-b313-e8fee9d2a3c2 - - - - -] Node f06c26d1-8008-47a9-a687-d5ef05083e41 moved to provision state "active" from state "deploying"; target provision state is "None"
2016-12-21 06:50:39.869 29447 INFO ironic.drivers.modules.agent_base_vendor [req-f77a2cef-c3e7-4863-b313-e8fee9d2a3c2 - - - - -] Deployment to node f06c26d1-8008-47a9-a687-d5ef05083e41 done