Toggling admin_state_up has no effect

[Paul Glass]

This was my setup - a single loadbalancer, one listener, one pool, and two members. `admin_state_up = False` on all of these resources.

  ===== Loadbalancer pglbutt-lb =====
  {
    "name": "pglbutt-lb",
    "operating_status": "ONLINE",
    "id": "782ede0c-251e-4bbb-bde0-c9cdc4564368",
    "pools": "{\"id\": \"bef3be76-0b57-4e1c-b35d-687949d5f3ef\"}",
    "provider": "octavia",
    "description": "",
    "admin_state_up": false,
    "tenant_id": "98a68f01476c4a3899ee709906d1cab3",
    "provisioning_status": "ACTIVE",
    "vip_subnet_id": "70fc37ca-aebb-4f2c-b4c1-0c9923aa7ef2",
    "listeners": "{\"id\": \"0786e713-331a-464f-9541-de1bf7c4fda3\"}",
    "vip_address": "10.0.0.10",
    "vip_port_id": "83168479-4ef6-4e0e-a2ff-0c99bb3a4a64"
  }
  ==== Listener pglbutt-listener =====
  {
    "name": "pglbutt-listener",
    "id": "0786e713-331a-464f-9541-de1bf7c4fda3",
    "default_pool_id": "bef3be76-0b57-4e1c-b35d-687949d5f3ef",
    "connection_limit": -1,
    "protocol_port": 80,
    "protocol": "HTTP",
    "description": "",
    "default_tls_container_ref": "",
    "admin_state_up": false,
    "loadbalancers": "{\"id\": \"782ede0c-251e-4bbb-bde0-c9cdc4564368\"}",
    "tenant_id": "98a68f01476c4a3899ee709906d1cab3",
    "sni_container_refs": ""
  }
  ===== Pool pglbutt-pool =====
  {
    "name": "pglbutt-pool",
    "id": "bef3be76-0b57-4e1c-b35d-687949d5f3ef",
    "members": "6b24a190-19f4-4d3c-bba2-b117d5f5b95e\nba471f88-c2b9-4573-9858-4fc4101ebc9d",
    "listeners": "{\"id\": \"0786e713-331a-464f-9541-de1bf7c4fda3\"}",
    "lb_algorithm": "ROUND_ROBIN",
    "protocol": "HTTP",
    "description": "",
    "admin_state_up": false,
    "loadbalancers": "{\"id\": \"782ede0c-251e-4bbb-bde0-c9cdc4564368\"}",
    "tenant_id": "98a68f01476c4a3899ee709906d1cab3",
    "session_persistence": "",
    "healthmonitor_id": ""
  }
  ==== Member pglbutt-2 =====
  {
    "id": "6b24a190-19f4-4d3c-bba2-b117d5f5b95e",
    "protocol_port": 80,
    "address": "10.0.0.13",
    "tenant_id": "98a68f01476c4a3899ee709906d1cab3",
    "subnet_id": "70fc37ca-aebb-4f2c-b4c1-0c9923aa7ef2",
    "admin_state_up": false,
    "weight": 1,
    "name": "pglbutt-2"
  }
  ==== Member pglbutt-1 =====
  {
    "id": "ba471f88-c2b9-4573-9858-4fc4101ebc9d",
    "protocol_port": 80,
    "address": "10.0.0.8",
    "tenant_id": "98a68f01476c4a3899ee709906d1cab3",
    "subnet_id": "70fc37ca-aebb-4f2c-b4c1-0c9923aa7ef2",
    "admin_state_up": false,
    "weight": 1,
    "name": "pglbutt-1"
  }

In this case admin_state_up is false everywhere, but curling the loadbalancer vip still allows traffic through:

    stack@octavia-devstack:~$ curl 10.0.0.10
    Hello, world (10.0.0.8)!
    stack@octavia-devstack:~$ curl 10.0.0.10
    Hello, world (10.0.0.13)!
    stack@octavia-devstack:~$ curl 10.0.0.10
    Hello, world (10.0.0.8)!
    stack@octavia-devstack:~$ curl 10.0.0.10
    Hello, world (10.0.0.13)!

[Michael Johnson (johnsom)]

I am seeing that admin-state-up=False does not work for loadbalancer and listener via the neutron client.

Pool and members do work.

[Michael Johnson (johnsom)]

Hmm, interesting. So, with listener admin-state-up=False, the haproxy configuration file is empty as expected, but traffic is still flowing.

[Michael Johnson (johnsom)]

The problem is here: https://github.com/openstack/octavia/blob/master/octavia/amphorae/drivers/haproxy/rest_api_driver.py#L77

In the case of admin-state-up=False for the listener and load balancer, this check is returning the incorrect information which causes the controller to send a start instead of the reload it should.

[Paul Glass (pnglass)]

Just confirming Michael's findings. I tested this again with fresh api resources. The pools and members respect respect admin-state-up = False, while the listener and loadbalancer do not.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/364707

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/364707
Committed: https://git.openstack.org/cgit/openstack/octavia/commit/?id=83731fd9a4b1f7748bb6af277b291f57756431df
Submitter: Jenkins
Branch: master

commit 83731fd9a4b1f7748bb6af277b291f57756431df
Author: Michael Johnson <email address hidden>
Date: Fri Sep 2 07:05:49 2016 +0000

    Fix admin_state_up for loadbalancer and listener

    The admin-state-up=False action for loadbalancer and listener
    failed to affect the appropriate change. This patch corrects that
    as well as removes an un-necessary call to the amphora-agent.

    Change-Id: I698f964f584d150f162f6c8cb41c65f5c5556b52
    Closes-Bug: #1619449

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/octavia 0.9.0 release.