Bånderapp:qol:status
Show the stream governor's status in every worker of the running web server: its learning state, algorithm profile, active streams and learning model. The command shows what GET /api/admin/qol/status returns in the admin API. Both read the workers through the server's control socket.
Quick start
make exec cmd="php bin/console app:qol:status"
Options
| Option | Description |
|---|---|
--json |
Print the data as the admin API returns it, in JSON |
Details
Run the command in the web container: it talks to the web server running there. The table has one row per HTTP worker and a total row:
| Column | Content |
|---|---|
| Worker | The Swoole worker ID |
| State | learning while the model gathers samples, active once it predicts stream cost |
| Profile | The algorithm profile: conservative, balanced or aggressive |
| Active streams | Streams the worker has admitted and not yet released; the total row adds them up |
| Samples | Utilization samples in the worker's learning model |
| Model ready | yes when the model has enough samples to predict cost |
| Budget cap | The share of CPU the profile lets streams use |
Each worker admits streams against its own active streams, so the counts differ between workers. The profile is shared: every worker reports the same one.
A worker that does not answer within two seconds is named on stderr, as is a worker whose status could not be read. The other workers' rows are still printed, and the command exits with code 1.
With --json, stdout holds only the data object of the API response: workers, total, missing_workers and worker_errors.
When no web server runs in the container, the command prints no web server is running in this container and exits with code 1.
Exit codes
| Code | Meaning |
|---|---|
| 0 | Every worker answered and the status was printed |
| 1 | No web server is running, a worker did not answer or failed, or the server could not be reached; the message says why |