Reviewed: https://review.openstack.org/216049 Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=47181ae3ebcd1533c22378ee31a4b1f0848926d6 Submitter: Jenkins Branch: master
commit 47181ae3ebcd1533c22378ee31a4b1f0848926d6 Author: Ludovic Beliveau <[email protected]> Date: Wed Nov 18 11:52:39 2015 -0500 Allow saving empty pci_device_pools in ComputeNode object Prior to this patch, saving a ComputeNode with a pci_device_pools attribute that has no objects specified in it (empty PciDevicePool list) would result in the change not being saved. Object of type PciDevicePoolList are evaluated like a list, thefore a conditional statement like 'if pools' will always evaluate to False even if 'pools' is not None. Without this fix, if 'pci_passthrough_whitelist' is cleared in the configuration, nova scheduler still think a compute node has PCI devices available and can still trigger scheduling an instance with PCI devices on the node. Change-Id: Ib3c19d569b9b3b23a293ad55dd9023291435d5a6 Closes-Bug: #1487451 ** 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/1487451 Title: Stale pci_stats in the DB after PCI reconfiguration Status in OpenStack Compute (nova): Fix Released Bug description: Changes in PCI device configuration is not reflected in the database (pci_stats). After nova reboot, pci_stats still hold stale data. This happen only when the compute as at some point interfaces configured with PCI SR-IOV or passthrough (in the pci_passthrough_whitelist) and then all those interfaces are removed. Steps to reproduce: 1. Configure SR-IOV on an interface and edit nova.conf/pci_passthrough_whitelist accordingly. 2. Start nova on the compute. 3. Remove the SR-IOV interface and it's configuration in nova.conf. 4. Restart nova on the compute. 5. Validate that pci_stats still hold the PCI device information by looking at the SQL database. This behavior cause the scheduler to still try to schedule an instance on the compute that had PCI configured even since no PCI device are availalble. To manage notifications about this bug go to: https://bugs.launchpad.net/nova/+bug/1487451/+subscriptions -- Mailing list: https://launchpad.net/~yahoo-eng-team Post to : [email protected] Unsubscribe : https://launchpad.net/~yahoo-eng-team More help : https://help.launchpad.net/ListHelp
