cloudkitty — Scenarios

sequenceDiagram
  participant P as cloudkitty-processor worker
  participant KS as Keystone
  participant TZ as tooz back end
  participant DB as MariaDB
  participant G as Gnocchi
  participant I as InfluxDB
  P->>KS: GET /v3/projects + role check per project (fetcher)
  KS-->>P: scopes with role "rating"
  loop each scope (shuffled)
    P->>TZ: acquire lock sha256(cloudkitty-<scope>-…) (non-blocking)
    alt lock busy
      Note over P: skip scope this loop
    else acquired
      P->>DB: SELECT cloudkitty_storage_states (last_processed_timestamp)
      Note over P: next = last + period, if next + wait_periods×period < now<br/>new scope: start of current month
      loop while a period is ready
        P->>DB: SELECT modules_state, hashmap_* / pyscripts_scripts active at period start
        par one thread per metric (max_threads)
          P->>G: POST /v1/aggregates (metric, scope, period, groupby)
          G-->>P: measures per group
          P->>G: POST /v1/search/resource/<type> (only if metadata)
        end
        alt every metric empty
          Note over P: WARNING "No usage data…"; nothing stored
        else data
          Note over P: hashmap/pyscripts set prices<br/>skip_datapoints_expression filters points
          P->>I: write_points measurement "dataframes" (chunks of 500)
        end
        P->>DB: UPDATE cloudkitty_storage_states SET last_processed_timestamp = period start
      end
      P->>TZ: release lock
    end
  end
  Note over P: sleep [collect]/period, then start again

sequenceDiagram
  participant C as Client (Skyline, CLI)
  participant A as cloudkitty-api
  participant MQ as RabbitMQ
  participant P as cloudkitty-processor
  participant DB as MariaDB
  C->>A: POST /v1/rating/quote {"resources":[{"service","desc","volume"}]}
  A->>A: policy rating:quote (open)
  A->>MQ: RPC call quote(res_data={"usage":{…}}) topic cloudkitty, ns rating
  MQ->>P: quote
  P->>DB: load enabled modules + all rules (tenant NULL only)
  Note over P: DataFrame [now, now+period)<br/>DataPoint(groupby=desc.groupby, metadata=desc.metadata)<br/>modules in priority order, sum of prices
  P-->>MQ: "0.042"
  MQ-->>A: reply
  alt no processor / timeout
    A-->>C: 500 (MessagingTimeout) [inferred]
  else
    A-->>C: 200 0.042
  end

sequenceDiagram
  participant Op as Administrator
  participant A as cloudkitty-api
  participant KS as Keystone
  participant DB as MariaDB
  participant P as cloudkitty-processor
  Op->>A: POST /v1/rating/module_config/hashmap/mappings {"field_id","value":"m1.small","type":"flat","cost":"0.05","start":"2026-12-01"}
  A->>A: policy rating:module_config (admin)
  A->>A: validate start/end (start in the past → 400 unless force=true)
  A->>KS: GET /v3/auth/tokens (X-Subject-Token = caller token, verify=False)
  KS-->>A: user id → created_by
  A->>DB: SELECT … FOR UPDATE (same name or overlapping window)
  alt conflict
    A-->>Op: 409 Mapping '…' already exists for field '…', tenant: '…'
  else
    A->>DB: INSERT hashmap_mappings
    A-->>Op: 201 Location: …/mappings/<uuid>
  end
  Note over P: at each scope cycle, rules active at the period start are reloaded:<br/>periods from 2026-12-01 use 0.05, earlier periods keep the old mapping

sequenceDiagram
  participant Op as Administrator
  participant A as cloudkitty-api
  participant DB as MariaDB
  participant R as cloudkitty-processor (reprocessor worker)
  participant TZ as tooz
  participant S as Storage back end
  participant Col as Collector back end
  Op->>A: POST /v2/task/reprocesses {"scope_ids":["p1"],"start_reprocess_time","end_reprocess_time","reason"}
  A->>DB: SELECT scope state, existing schedules
  alt misaligned / beyond last processed / overlapping
    A-->>Op: 400 …
  else
    A->>DB: INSERT storage_scope_reprocessing_schedule
    A-->>Op: 202
  end
  loop reprocessor loop (every period)
    R->>DB: SELECT unfinished schedules
    R->>TZ: acquire lock (schedule)
    R->>DB: next = start or current + period (≤ end)
    alt first period of the schedule
      R->>S: delete [start, end) for {scope_key: p1}
    end
    R->>Col: collect, rate with rules active at that period
    R->>S: push rated frame
    R->>DB: UPDATE current_reprocess_time
  end

sequenceDiagram
  participant Op as Administrator
  participant A as cloudkitty-api
  participant DB as MariaDB
  participant MQ as RabbitMQ
  participant P as cloudkitty-processor (any worker)
  participant TZ as tooz
  participant S as Storage back end
  Op->>A: PUT /v2/scope {"scope_id":"p1","last_processed_timestamp":"2026-10-01T00:00:00"}
  A->>DB: SELECT matching scopes
  A->>MQ: cast reset_state(res_data)
  A-->>Op: 202
  MQ->>P: reset_state
  loop each scope (shuffled)
    P->>TZ: acquire lock (blocking)
    P->>S: delete from last_processed_timestamp to now for {scope_key: id}
    P->>DB: UPDATE last_processed_timestamp
    P->>TZ: release
  end
  Note over P: the next rating loop re-rates from last_processed_timestamp + period