feat(backend): Demonstrate metrics #1488
Conversation
Signed-off-by: Sylvain Leclerc <[email protected]>
Features
ContributorsCommit-Lint commandsYou can trigger Commit-Lint actions by commenting on this PR:
|
antarest/core/config.py
Outdated
| @staticmethod | ||
| def from_dict(data: JSON) -> "PrometheusConfig": | ||
| return PrometheusConfig(multiprocess=bool(data["multiprocess"])) |
There was a problem hiding this comment.
The pythonic way to implement multiple constructors is to use @classmethod. See this article about "Providing Multiple Constructors in Your Python Classes".
@classmethod
def from_dict(cls, data: JSON) -> "PrometheusConfig":
return cls(multiprocess=bool(data["multiprocess"]))
antarest/core/config.py
Outdated
| @staticmethod | ||
| def from_dict(data: JSON) -> "MetricsConfig": | ||
| return MetricsConfig( | ||
| prometheus=PrometheusConfig.from_dict(data["prometheus"]) | ||
| if "prometheus" in data | ||
| else None | ||
| ) |
There was a problem hiding this comment.
use @classmethod instead
antarest/core/exceptions.py
Outdated
| def __init__(self, message: str) -> None: | ||
| super().__init__(message) |
There was a problem hiding this comment.
This method is not really useful unless you add some information, for instance, you can use the message template here (I mean "Environment variable {_PROMETHEUS_MULTIPROCESS_ENV_VAR} must be defined for use of prometheus in a multiprocess environment")
| logger = logging.getLogger(__name__) | ||
|
|
||
|
|
||
| _PROMETHEUS_MULTIPROCESS_ENV_VAR = "PROMETHEUS_MULTIPROC_DIR" |
There was a problem hiding this comment.
Add a short comment to explain the purpose of this environment variable, for instance:
# The `PROMETHEUS_MULTIPROC_DIR` environment variable is used by
# the Python Prometheus client library to configure process-level metrics
# when running in a multi-process environment.
antarest/core/metrics.py
Outdated
| if _PROMETHEUS_MULTIPROCESS_ENV_VAR not in os.environ: | ||
| raise ConfigurationError( | ||
| f"Environment variable {_PROMETHEUS_MULTIPROCESS_ENV_VAR} must be defined for use of prometheus in a multiprocess environment" | ||
| ) |
There was a problem hiding this comment.
We check that this environment variable exists but it is not used directly in the code (but only in the client). So, a little explanation is needed.
Moreover, we should also check that it is a directory that exists (with the right permissions).
| @@ -8,7 +8,7 @@ checksumdir~=1.2.0 | |||
| click~=8.0.3 | |||
| contextvars~=2.4 | |||
| fastapi-jwt-auth~=0.5.0 | |||
| fastapi[all]~=0.73.0 | |||
| fastapi[all]~=0.74.0 | |||
There was a problem hiding this comment.
FYI: A version upgrade would be welcome to take advantage of the latest developments that simplify the update of the Swagger API documentation. But, we need to analyse the impacts...
There was a problem hiding this comment.
Ok! This was the minimum upgrade to get the "route" information from the request
| """ | ||
| Put the worker_id into an env variable for further use within the app. | ||
| """ | ||
| os.environ["APP_WORKER_ID"] = str(worker.age) |
There was a problem hiding this comment.
If I had a choice, I would prefer GUNICORN_WORKER_ID for that.
Why do you use worker.age instead of worker.id?
worker.age is a number that indicates how many requests the worker process has handled since it was started.
There was a problem hiding this comment.
From what I understood, worker.age is a counter incremented every time gunicorn creates a new worker (for ex. when one dies).
Using the age here allows to have an almost stable set of worker IDs between application runs (after a restart of the app, for example, we sill have IDs in the range [1-8] for 8 workers), to have mostly stable metrics labels.
I don't see any worker.id, we have the PID as an alternative, but it's not stable at all between runs.
Signed-off-by: Sylvain Leclerc <[email protected]>
Signed-off-by: Sylvain Leclerc <[email protected]>
Signed-off-by: Sylvain Leclerc <[email protected]>
|
@sylvlecl : do you know about that : https://autometrics.dev/ |
Description
This is a draft for defining metrics and exposing them to a prometheus instance, in order to improve the application observability.
The only metrics that are defined fro now are metrics about HTTP requests (count, duration).
Some technical details:
In the future, we could add metrics about various things: DB sessions, tasks durations, ...
In the long run, using opentelemetry could be an alternative.
As a very basic illustration of grah in grafana:
