nova-compute fails to start if vCenter cluster has no ESXi's

[Tatyana Dubyk]

Affects 6.0 and 6.1.

Nova-compute service cannot start if at least one vCenter cluster
specified in nova.conf doesn't contain any ESXi in it.

For 6.0 it leads to full unavailability of nova-compute because all vcenter clusters are specified in one nova.conf and are served by one nova-compute service instance.

For 6.1 each vCenter cluster are served by a dedicated nova-compute instance, so this bug will lead to unavailability of a particular nova-compute service.

Workaround:
For 6.0 manually remove broken cluster from nova.conf.

Original description:

1. create lab with RC3 #56 from here http://jenkins-product.srt.mirantis.net:8080/view/6.0/job/6.0.all/
2. Wizard parameters :
-CentOS or Ubuntu (no matter)
-hypervizor - vcenter
- storages - by default all settings

 settings for vcenter:
VCENTER_IP=172.16.0.254
<email address hidden>
VCENTER_PASSWORD=Qwer!1234
VCENTER_CLUSTERS=Cluster1,Cluster2

settings:
vcenter/ESXi IP:172.16.0.254
username:<email address hidden>
password:Qwer!1234
Datastore name: nfs
Datacenter name: Datacenter
Datastore Images directory: /openstack_glance2

Interfaces on slaves:
eth0 - admin(PXE)
eth1 - public
eth2 - management
eth3 - VM(Fixed) VLAN ID:103
eth4 - storage

3, env : 1 controller, 2 cinder

My lab contains snapshot with 2 clusters, where in one of them is absent hypervizor.

Expected result: one cluster must be available

Actual result: no one is available

[root@nailgun ~]# fuel --fuel-version
api: '1.0'
astute_sha: 16b252d93be6aaa73030b8100cf8c5ca6a970a91
auth_required: true
build_id: 2014-12-18_01-32-01
build_number: '56'
feature_groups:
- mirantis
fuellib_sha: 73332192a257ea02c40a39885c502ad1ebdf3eda
fuelmain_sha: 45caacadb878abfbd9d60e134d72229698b469c9
nailgun_sha: 5f91157daa6798ff522ca9f6d34e7e135f150a90
ostf_sha: a9afb68710d809570460c29d6c3293219d3624d4
production: docker
release: '6.0'
release_versions:
  2014.2-6.0:
    VERSION:
      api: '1.0'
      astute_sha: 16b252d93be6aaa73030b8100cf8c5ca6a970a91
      build_id: 2014-12-18_01-32-01
      build_number: '56'
      feature_groups:
      - mirantis
      fuellib_sha: 73332192a257ea02c40a39885c502ad1ebdf3eda
      fuelmain_sha: 45caacadb878abfbd9d60e134d72229698b469c9
      nailgun_sha: 5f91157daa6798ff522ca9f6d34e7e135f150a90
      ostf_sha: a9afb68710d809570460c29d6c3293219d3624d4
      production: docker
      release: '6.0'

[Oleksiy Molchanov (omolchanov)]

Please provide diagnostic snapshot and more logs.

[Davanum Srinivas (DIMS) (dims-v)]

Igor, can you please get me logs from nova compute for this bug?

[Davanum Srinivas (DIMS) (dims-v)]

Tatyana, can you please get me logs from nova compute for this bug?

[Igor Zinovik (izinovik)]

I created environment and trying to reproduce the bug;
will upload logs tomorrow.

[Igor Zinovik (izinovik)]

I created environment with fuel-6.0.1-99-2015-02-20_08-18-15.iso, added vCenter with two clusters (Cluster1, Cluster2).
Moved all hosts from Cluster2 to Cluster1, so Cluster2 became empty.

Environment has one controller that holds all nova services (api, scheduler, compute, network).

Attached logs (/var/log/nova/ directory from single controller).

[Stepan Rogov (srogov)]

I think, we should set "won't fix", and document this issue.
Because in 6.1 we have not this problem.

[Denis Meltsaykin (dmeltsaykin)]

Closing this as Won't Fix for 6.0-updates, as the issue is not customer-found and no fix is expected.