vDNS for non-active VMs

[tom murray]

this is with Contrail release 1.20/63

the Contrail vDNS only works for VMs in the "ACTIVE" state. VMs that are shut down cannot be resolved.

For example (10.163.43.4 and 10.163.43.12 are my controllers):

root@gngsvm009d:/opt/contrail/utils# nova list --all-tenant |grep tmtest
| f301ccaf-343f-4d31-84fd-c284fecb12c7 | tmtest1 | ACTIVE | - | Running | svec-private=192.168.0.11, 10.163.0.12 |

root@gngsvm009d:/opt/contrail/utils# nslookup tmtest1.svec - 10.163.43.12
Server: 10.163.43.12
Address: 10.163.43.12#53

Name: tmtest1.svec.juniper.net
Address: 10.163.0.12

root@gngsvm009d:/opt/contrail/utils# nslookup tmtest1.svec - 10.163.43.4
Server: 10.163.43.4
Address: 10.163.43.4#53

Name: tmtest1.svec.juniper.net
Address: 10.163.0.12

root@gngsvm009d:/opt/contrail/utils# nova stop f301ccaf-343f-4d31-84fd-c284fecb12c7
root@gngsvm009d:/opt/contrail/utils# nova list --all-tenant |grep tmtest
| f301ccaf-343f-4d31-84fd-c284fecb12c7 | tmtest1 | SHUTOFF | - | Shutdown | svec-private=192.168.0.11, 10.163.0.12 |

root@gngsvm009d:/opt/contrail/utils# nslookup tmtest1.svec - 10.163.43.12
Server: 10.163.43.12
Address: 10.163.43.12#53

** server can't find tmtest1.svec: NXDOMAIN

root@gngsvm009d:/opt/contrail/utils# nslookup tmtest1.svec - 10.163.43.4
Server: 10.163.43.4
Address: 10.163.43.4#53

** server can't find tmtest1.svec: NXDOMAIN

A DNS client should not be dependent on a VM run state.

[Hari Prasad Killi (haripk)]

We follow Dynamic DNS - the DNS entry is added to the server when contrail-vrouter-agent receives VM interface info. In this case, the VM didnt get spawned and hence there is no entry in the DNS server for the same. Standard DDNS adds DNS entries upon DHCP.

I propose to invalidate the bug, as this is the model we use.

[tom murray (tmurray-a)]

I think the response above is only valid when the VM had never been started. If the VM was ever made active it would have been assigned an IP.

As a client of DNS i should not be dependent on the VM run state. In fact, nova still reports the IP for the VM regardless of the state. for example:

root@gngsvm009d:/opt/contrail/utils# nova show 0cb1d9ab-573a-47a2-9a15-80aa91f274c5
+--------------------------------------+--------------------------------------------------------------------------------------------------+
| Property | Value |
+--------------------------------------+--------------------------------------------------------------------------------------------------+
| OS-DCF:diskConfig | MANUAL |
| OS-EXT-AZ:availability_zone | nova |
| OS-EXT-SRV-ATTR:host | gngsvc018d |
| OS-EXT-SRV-ATTR:hypervisor_hostname | gngsvc018d |
| OS-EXT-SRV-ATTR:instance_name | instance-00000f25 |
| OS-EXT-STS:power_state | 4 |
| OS-EXT-STS:task_state | - |
| OS-EXT-STS:vm_state | stopped |
| OS-SRV-USG:launched_at | 2014-12-09T01:39:26.000000 |
| OS-SRV-USG:terminated_at | - |
| accessIPv4 | |
| accessIPv6 | |
| config_drive | |
| created | 2014-12-09T01:39:13Z |
| flavor | jbm-std (097059e7-d6ce-4c23-9fce-4504816771e4) |
| hostId | 155ee1eb84b05b3d1ecb5387856122bb664d2b83a74c8fea6c557d00 |
| id | 0cb1d9ab-573a-47a2-9a15-80aa91f274c5 |
| image ...

[tom murray (tmurray-a)]

one thing to add to this issue is that if the QEMU crashes under the VM, then even when the VM becomes ACTIVE again (using nova start on the VM) the DNS entry is not registered. This results in the VM in an ACTIVE state, but no vDNS records.