[Starlingx-discuss] [ Test ] Meeting notes - 12/18/2018

Cabrales, Ada ada.cabrales at intel.com
Tue Dec 18 19:26:17 UTC 2018


Agenda for 12/18
Attendees: JC, JP, Cristopher, Jose, Numan, Maria, Fernando, Bruce, Elio, Bill, Richo
Sanity testing: coverage improvement - JC
-> Current sanity has 24 tests, covering 4 virtual configs: simplex, duplex, multinode with local storage, multinode external storage
  -  Proposal:
      * Recovery after dead office
      * Recovery after Power Down / Power Up controllers, computes and storage hosts
      * Lock / Unlock active controller reject and check alarms
      * Lock / Unlock standby controller, computes and storage hosts, check alarms and then SSH from controller to each host
      * Create an stack from a heat template
      * Controller swact with active instances
      * Controller swact triggered by host reboot with active instances
      * Launch instances from volume, from image and from snapshot, with different OS, and ping between them
      * Launch instances from volume, from image and from snapshot, with different OS, and perform suspend/resume, pause/unpause, stop/start, lock/unlock, reboot,    rebuild and resize
      * Launch instances from volume, from image and from snapshot, with diferent OS, and set/unset instances properties
      * Live / Cold migration of instances launched from volume, from image and from snapshot, with diferent OS
      * Kill instances and check they goes to hard reboot and recovery
      * Evacuation of instances by rebooting a controller host or compute host
      * Down management network in active controller, in standby controller, in computes and in storage hosts, then check alarms and check swact takes action
      * Change the MTU value of data interfaces in all hosts
      * Kill services managed by SM on active controller, then check service recovery                         
      * Display HA system service group, service list and state of such services
      * Check there are not new alarms generated at the end of Sanity testing
  -  Sanity should be fast, assuring the build is safe. 
         Recovery after dead office and recovery after power down/power up nodes -> this could be part of an extended sanity 
  - Exercise different types of VMs, with different types (as one with Windows)
  - There's no certainty on how long would it take to run,could be all night long
   - If Sanity fails, critical bug is created. If extended sanity fails, might not be a critical.
-> Items from the distro.openstack team - Bruce
   - Rebasing to OpenStack master on mid-January
   - 3 features: SRIOV
         SR-IOV/PT best effort scheduling policy with Queens feature
         Replace vswitch affinity with Rocky feature
         DB purge 
   - Request is to get help from testing team for making sure the 3 features are working correclty
   - Bruce to create the storyboards for the rebasing and the test creation
         FH> Adding some helpful links
         https://review.openstack.org/#/c/555000/3..3/specs/queens/implemented/share-pci-between-numa-nodes.rst
         https://review.openstack.org/#/c/541290/18/specs/rocky/approved/numa-aware-vswitches.rst,unified
         https://blueprints.launchpad.net/nova/+spec/purge-db
-> Update request to CENGN -  storage of the sanity logs - Ken
      Meeting next Thursday - update to be provided
-> Reminder - Meeting this Friday for checking PyTest
         Last meeting of the year, will resume on Jan 8
-> Opens:
         Fernando:  email sent to Numan asking for help on some security stuff. 
             Numan to get back with the information.

--

Ada


More information about the Starlingx-discuss mailing list