yahoo-eng-team team mailing list archive

[OpenStack Infra <1854212@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.opendev.org/696392
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=6f1e43f831a5418d6202eed9875d4284557a4969
Submitter: Zuul
Branch:    master

commit 6f1e43f831a5418d6202eed9875d4284557a4969
Author: Vladyslav Drok <vdrok@xxxxxxxxxxxx>
Date:   Wed Nov 27 20:27:02 2019 +0100

    Make removal of host from aggregate consistent
    
    It should allow the same conditions as addition of host allows, for
    example in case of HostMapping missing for the host being removed.
    
    Closes-Bug: 1854212
    Change-Id: I72c2c16ae0441ee605b834ad146513e98f84c632


** 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/1854212

Title:
  For ironic nova-computes that have no nodes associated aggregate
  operations are inconsistent

Status in OpenStack Compute (nova):
  Fix Released

Bug description:
  Consider a case of mixed hypervisor setup, there are computes doing
  QEMU instances, and there are computes providing possibility to boot
  ironic nodes. There are no ironic nodes commissioned at the moment,
  but QEMU computes are setup and ready to go. When no "available" for
  deployment ironic nodes are associated to the compute, running
  discover_hosts will not create HostMapping entries for compute hosts,
  as there are no hypervisor nodes they could be tied to. Here is how
  host list, service list and hypervisor list commands output look like
  (bmt hosts are computes with ironic virt driver):

  root@ctl01:~# openstack compute service list --service nova-compute
  +----+--------------+--------+----------+---------+-------+----------------------------+
  | ID | Binary       | Host   | Zone     | Status  | State | Updated At                 |
  +----+--------------+--------+----------+---------+-------+----------------------------+
  | 55 | nova-compute | bmt03  | foo_zone | enabled | up    | 2019-11-27T12:08:54.000000 |
  | 58 | nova-compute | bmt02  | nova     | enabled | up    | 2019-11-27T12:08:32.000000 |
  | 61 | nova-compute | bmt01  | nova     | enabled | up    | 2019-11-27T12:09:06.000000 |
  | 64 | nova-compute | cmp003 | nova     | enabled | up    | 2019-11-27T12:08:24.000000 |
  | 67 | nova-compute | cmp001 | nova     | enabled | up    | 2019-11-27T12:08:14.000000 |
  | 70 | nova-compute | cmp002 | nova     | enabled | up    | 2019-11-27T12:08:21.000000 |
  +----+--------------+--------+----------+---------+-------+----------------------------+
  root@ctl01:~# openstack host list | grep compute
  | bmt03     | compute     | foo_zone |
  | bmt02     | compute     | nova     |
  | bmt01     | compute     | nova     |
  | cmp003    | compute     | nova     |
  | cmp001    | compute     | nova     |
  | cmp002    | compute     | nova     |

  In hypervisor list you can see that no ironic nodes are present.

  root@ctl01:~# openstack hypervisor list
  +----+--------------------------------------+-----------------+--------------+-------+
  | ID | Hypervisor Hostname                  | Hypervisor Type | Host IP      | State |
  +----+--------------------------------------+-----------------+--------------+-------+
  |  7 | cmp003.bm-cicd-queens-ovs-maas.local | QEMU            | 10.167.11.17 | up    |
  | 10 | cmp001.bm-cicd-queens-ovs-maas.local | QEMU            | 10.167.11.15 | up    |
  | 13 | cmp002.bm-cicd-queens-ovs-maas.local | QEMU            | 10.167.11.16 | up    |

  Then the following test is run:
  tempest.api.compute.admin.test_aggregates_negative.AggregatesAdminNegativeTestJSON.test_aggregate_remove_host_as_user
  [id-7a53af20-137a-4e44-a4ae-e19260e626d9,negative]

  The test tries to add a host to an aggregate and then remove it,
  addtion seems to be ok but removal fails with the following:

  Cannot remove host bmt03 in aggregate 52 (HTTP 404) (Request-ID: req-
  25acad79-b820-4889-a2a3-71b4a7913a86)

  In nova api something like the following can be seen:

  2019-11-27 12:11:42,263.263 2762 INFO nova.api.openstack.wsgi [req-
  80f752d1-0ce6-446a-ba92-cf815bfedd78 b31b4e9171154b1d9d1fec5af0ed5880
  6a33e1d31b53496885e19d7cba4390dd - default default] HTTP exception
  thrown: Host 'bmt01' is not mapped to any cell

  The problem seems to be in an inconsistent behaviour during adding the
  host to aggregate and removing it from aggregate, as during addition
  it allows host mapping to be missing, but does not allow it during
  host removal.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1854212/+subscriptions