[vCenter] Launch instance from a snapshot has failed

[Ilya Bumarskov]

Fuel 8.0 ISO #322

Enviroment:
Compute: QEMU + vCenter
Networking Setup: Neutron with VMware DVS (ML2 plugin)
Storage backends: default
Additional services: all by default

Settings:
VMware vCenter/ESXi datastore for images (Glance)

Nodes:
Controller
Controller
Controller
Compute, Cinder
Cinder-vmware
Compute-vmware

Steps to reproduce:
1. Get existing image by name.
2. Launch an instance using the default image (flavor m1.tiny)
3. Make snapshot of the created instance.
4. Delete the instance created in step 1.
5. Wait while instance deleted
6. Launch another instance from the snapshot created in step 2.

Observed behavior:
Error: Failed to perform requested operation on instance "test_vcenter3", the instance has an error status: Please try again later [Error: No valid host was found. There are not enough hosts available.].

[Alexander Arzhanov (aarzhanov)]

Some logs.

[Alexey Galkin (agalkin)]

attached log uninformative, please attach fuel diagnostic snapshot

[Kairat Kushaev (kkushaev)]

Alex, Ilya, there is nothing useful in glance-api.log except CommunicationError:
2015-12-24T09:42:28.010888+00:00 err: Communication error sending http PUT requestto the url /folder/openstack_glance/fbcf78d5-172e-4c41-9302-cdf9449ad2f2%3FdcPath%3DDatacenter%26dsName%3Dnfs.
Got IOError [Errno 32] Broken pipe

Could you please attach fuel logs and VMware vCenter/ESXi datastore logs to the bug?

[Mike Fedosin (mfedosin)]

This seems to be due very low NFC timeout value used in VC.

Here is the workaround:

1. Make a backup copy of the vpxd.cfg file.
2. Open the vpxd.cfg file in a text editor.
3. Add the following lines between <vmacore> and </vmacore> stanza:

  <http>
      <readTimeoutMs>600000</readTimeoutMs>
  </http>

For example, a complete section appears similar to:

<config>
  <vmacore>
    <http>
      <readTimeoutMs>600000</readTimeoutMs>
    </http>
  </vmacore>
..
</config>

4. Stop and restart the VirtualCenter Server Service. This ensures it picks up the updated parameters in the vpxd.cfg file.

[Alexander Arzhanov (aarzhanov)]

Workaround (readTimeoutMs 600000) help with this bug.
Docs team it is necessary to add known issues.
Link to upstream bug - https://bugs.launchpad.net/glance/+bug/1498645, because upstream bug in status "Won't Fix", I put the bug in status "Won't Fix".

[Michele Fagan (michelefagan)]

Added tag "area-docs" so that this workaround can be added to the 8.0 Release Notes.

[Mike Fedosin (mfedosin)]

Hi, I think we can assign this bug for puppet team, because the fix is to change the configuration of vmware store (comment #4)