Description of problem:
We ran into a situation with an OSP12 deployment (upgraded from OSP11) in which new instances were failing to schedule on a target compute node that was empty. Nova-conductor would throw a NoValidHost exception without any details, and there were no other error messages related to the failure. It turns out this was due to a number of stale resource allocations on the target compute node.
If nova-compute finds stale allocations, it logs a message at DEBUG priority that says:
There are allocations remaining against the source host that might need to be removed:
Since this is logged at DEBUG priority, the message will not typically be seen by operators in a production environment. However, stale allocations can prevent Nova from scheduling new instances on the compute node, leading to mysterious failures for which there are no obvious errors.
This message should be logged at WARNING priority so that it is visible.
Version-Release number of selected component (if applicable):
Bug filed by a Red Hatter with no customer case and is a simple logging level change, don't think we need doc text for this.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.