python-octaviaclient — Scenarios

Scenarios — python-octaviaclient — branch stable/2026.2

Branch stable/2026.2 @ b37975ec8f2d (tag 3.15.0 + 2 commits) — repository https://opendev.org/openstack/python-octaviaclient — analysed on 2026-11-01

The request sequences are [observed] against the local stub. Statuses and timing on a real service are not observed here. The asynchronous work on the service side belongs to reverse-spec-octavia/.

1. Create a load balancer and wait

openstack loadbalancer create --vip-subnet-id <subnet> --wait (octaviaclient/osc/v2/load_balancer.py:155-190).

sequenceDiagram
  autonumber
  actor U as User
  participant C as openstack
  participant K as Keystone
  participant N as Neutron
  participant O as Octavia
  U->>C: loadbalancer create --vip-subnet-id S --wait
  C->>K: GET /identity, POST /v3/auth/tokens
  K-->>C: token and catalog
  C->>N: GET /v2.0/subnets?id=S or ?name=S
  N-->>C: one subnet
  C->>O: POST /v2.0/lbaas/loadbalancers
  O-->>C: 201 loadbalancer with id
  loop every 3 s, no limit
    C->>O: GET /v2.0/lbaas/loadbalancers/ID
    O-->>C: provisioning_status
  end
  alt ACTIVE
    C->>O: GET /v2.0/lbaas/loadbalancers/ID
    C-->>U: table, exit 0
  else ERROR
    C-->>U: did not successfully reach ACTIVE status, exit 1
  else missing field
    C-->>U: NoneType has no attribute lower, exit 1
  end

Diagnostics: Service 'network' is disabled… means no network entry in the catalog. A hang means the load balancer stays in PENDING_*: look at the Octavia worker. With --debug, the request IDs are printed.

2. Show by name, and the fallback of a missing UUID

loadbalancer show <value> (octaviaclient/osc/v2/load_balancer.py:385-417, octaviaclient/osc/v2/utils.py:68-105).

sequenceDiagram
  autonumber
  participant C as openstack
  participant O as Octavia
  alt the value looks like a UUID
    C->>O: GET /lbaas/loadbalancers/UUID
    alt 200
      O-->>C: loadbalancer
    else 404 or 400
      O-->>C: fault
      C->>O: GET /lbaas/loadbalancers?id=UUID
      O-->>C: empty
      C->>O: GET /lbaas/loadbalancers?name=UUID
      O-->>C: empty
      C-->>C: Unable to locate UUID in loadbalancers, exit 1
    else 403 or other
      O-->>C: fault
      C-->>C: faultstring (HTTP code), exit 1
    end
  else a name
    C->>O: GET /lbaas/loadbalancers?name=N
    alt one match
      O-->>C: one row
      C->>O: GET /lbaas/loadbalancers/ID
    else several
      C-->>C: 2 loadbalancers found with name or ID of N, exit 1
    else none
      C-->>C: Unable to locate N in loadbalancers, exit 1
    end
  end

Diagnostics: Unable to locate also appears when the project in the token cannot see the resource. A KeyError-style message such as 'loadbalancers' means the service sent a list without its key.

3. Add a member and wait

loadbalancer member create <pool> --address A --protocol-port 80 --wait (octaviaclient/osc/v2/member.py:211-246). The member has its own status, but the command waits on the load balancer.

sequenceDiagram
  autonumber
  participant C as openstack
  participant O as Octavia
  C->>O: GET /lbaas/pools?name=P
  O-->>C: pool id
  C->>C: check port and weight
  C->>O: POST /lbaas/pools/ID/members
  O-->>C: 201 member
  C->>O: GET /lbaas/pools/ID
  O-->>C: loadbalancers list
  loop until ACTIVE or ERROR
    C->>O: GET /lbaas/loadbalancers/LB
  end
  C->>O: GET /lbaas/pools/ID/members/MEMBER
  O-->>C: member

Diagnostics: if the pool's loadbalancers is empty, the command ends with list index out of range although the member exists (02 § 5 item 11).

4. Delete with cascade and wait

loadbalancer delete <lb> --cascade --wait (octaviaclient/osc/v2/load_balancer.py:219-232, octaviaclient/osc/v2/utils.py:801-821).

sequenceDiagram
  autonumber
  participant C as openstack
  participant O as Octavia
  C->>O: GET /lbaas/loadbalancers?name=N
  O-->>C: id
  C->>O: DELETE /lbaas/loadbalancers/ID?cascade=True
  O-->>C: 204
  loop every 3 s, at most 300 s
    C->>O: GET /lbaas/loadbalancers/ID
    alt 404
      O-->>C: fault
      C-->>C: done, exit 0
    else status error
      O-->>C: ERROR
      C-->>C: could not be successfully deleted, exit 1
    else other status
      O-->>C: PENDING_DELETE
    end
  end
  C-->>C: could not be successfully deleted, exit 1

Diagnostics: a 409 on the DELETE usually means a parent is immutable; its faultstring is displayed. A 403 during polling stops the wait [observed].

5. A filtered list across pages

loadbalancer list --project P when the filter fills a page (octaviaclient/api/v2/octavia.py:82-102).

sequenceDiagram
  autonumber
  participant C as openstack
  participant O as Octavia
  C->>O: GET /lbaas/loadbalancers?project_id=P
  O-->>C: rows of page 1, next link with limit and marker only
  C->>O: GET /lbaas/loadbalancers?limit=L&marker=M
  Note over C,O: project_id is gone: the service filters only by its own rules
  O-->>C: next rows, not restricted to P
  C-->>C: prints both pages, exit 0

Diagnostics: more rows than expected on a large deployment, or rows of another project for an admin. Client side [observed]; Octavia's link was read (octavia:octavia/api/common/pagination.py:151-172). The two were not run together.

Source: reverse-spec-python-octaviaclient/04-scenarios_stable-2026.2.md. The _rewrite/ and _inventory/ files are not published on the site.