During instance creation tenant network should be given always

[Slawek Kaplonski]

Example of such failure: http://logs.openstack.org/74/659274/7/check/networking-ovn-tempest-dsvm-ovs-release/c567cd1/testr_results.html.gz

Logstash query to find similar problems: http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22Multiple%20possible%20networks%20found%2C%20use%20a%20Network%20ID%20to%20be%20more%20specific%5C%22%20AND%20build_status%3A%5C%22FAILURE%5C%22

From log in failed test we can note that None tenant network was found:

2019-07-02 01:50:18,568 13517 INFO [tempest.lib.common.fixed_network] (ServersTestJSON:test_update_access_server_address) Found network None available for tenant

Looking at fixed_network.get_tenant_network() function above log means that there was no network given in net_creds (https://github.com/openstack/tempest/blob/eac094a8cf834d035316a900107f601adcc42ff5/tempest/lib/common/fixed_network.py#L95)

As "compute.fixed_network_name" option is also not configured, it didn't even tried to get any tenant network and returned None (https://github.com/openstack/tempest/blob/eac094a8cf834d035316a900107f601adcc42ff5/tempest/lib/common/fixed_network.py#L97).

This usually should works if tenant has got only one network available to use.
But sometimes it looks that there is more than one network and it fails.

Maybe as a first step to investigate it it would be good to add in nova debug log to log what networks were found for the tenant. It can be added in https://github.com/openstack/nova/blob/46a3bcd80b41e99ec4923c7cf3d0f8dd8505e97c/nova/network/neutronv2/api.py#L874

[Paras Babbar (pbabbar)]

Can you shared the detailed logs, the links shared are not working. I think from the logs it shows there were no network found but yes from the nova compute point of view if there are two network found and we run the create instance method it will failed this will happen the same way I tested manually too like if we use openstack server create without specifying network id(--net) the instance deployment will fail as we need to specify explicitly which network to use that is based on design.

So If I understand correctly your suggestion would be to explicitly specify one network and use that as part of instance creation in whole tempest suite?

However my suggestion is to cover this as a part of deployment task like create pre-tempest.sh script where you add images and all and in same script create only one network and use that for running tempest suite? as this is based on environment.

Let me know if it make sense and we can close this.

[Martin Kopec (mkopec)]

Per the previous comment moving to Incomplete as we're waiting for further input from the reporter's side.
Also we have merged several patches related to network creation in compute tests, see https://bugs.launchpad.net/tempest/+bug/1844568 and we have fixed a bug in scenario tests which were failing when more than 1 network was present, see https://bugs.launchpad.net/tempest/+bug/1856671.

Based on the above I advise to rerun the tests with newer tempest and check if the problem still exists.

[Launchpad Janitor (janitor)]

[Expired for tempest because there has been no activity for 60 days.]