Cannot create loadbalancers under bionic train take 2

[Dan MacDonald]

We have re-installed the openstack-telemetry-bionic-train bundle with the octavia overlay but this time I was careful to set the openstack-origin settings in the overlay to match, which is where we went wrong last time. Unfortunately we are still unable to create loadbalancers.

The command we are trying to run (as the Openstack admin user) is:

$ openstack loadbalancer create --name lb1 --vip-subnet-id internal_subnet

We get told its being deployed, but shortly after if I run:

$ openstack loadbalancer show lb1

and we get a

provisioning_status | ERROR

The octavia health manager log makes it out to be a database issue:

no such table: amphora_health

[Dan MacDonald (allcoms)]

octavia-health-manager.log attached

[Dan MacDonald (allcoms)]

octavia-housekeeping.log attached

[Dan MacDonald (allcoms)]

I had forgot to tag the image I wanted to use for the amphora but after doing that I still cannot created a LB.

I tried creating my latest test LB about 40 minutes ago and now when I run:

openstack loadbalancer show lb3

I still see provisioning_status | PENDING_CREATE

and it eventually just gives up octavia-worker.log:

2020-11-26 17:24:54.618 148019 ERROR octavia.amphorae.drivers.haproxy.rest_api_driver [-] Connection retries (currently set to 120) exhausted. Th
e amphora is unavailable. Reason: HTTPSConnectionPool(host='fc00:b:374e:57fe:f816:3eff:fe91:4d71', port=9443): Max retries exceeded with url: // (
Caused by ConnectTimeoutError(<urllib3.connection.VerifiedHTTPSConnection object at 0x7ff2047b6828>, 'Connection to fc00:b:374e:57fe:f816:3eff:fe9
1:4d71 timed out. (connect timeout=10.0)'))
2020-11-26 17:24:54.618 148019 ERROR octavia.controller.worker.v1.tasks.amphora_driver_tasks [-] Amphora compute instance failed to become reachab
le. This either means the compute driver failed to fully boot the instance inside the timeout interval or the instance is not reachable via the lb
-mgmt-net.: octavia.amphorae.driver_exceptions.exceptions.TimeOutException: contacting the amphora timed out

Don't I need to switch Octavia to using a ipv4 management subnet if I want to use it with an IPv4 subnet?

[Dan MacDonald (allcoms)]

On https://github.com/openstack/charm-octavia it says I have the option of setting create-mgmt-network to false. I presume that is something that has to be set at deployment time in the juju config?

If I redeploy the bundle but set create-mgmt-network to false, what additional steps would be required to get Octavia running?