StationA/gke-logging

Utilities for interacting with logging facilities in GKE workloads

gke-logging

Utilities for interacting with logging facilities in GKE workloads

Installation

Requirements

• Python 3.7+
• Poetry (for development only)

Install from PyPI (recommended)

pip install gke-logging

Installing from Github

pip install git+https://github.com/StationA/gke-logging.git#egg=gke-logging

Installing from source

git clone https://github.com/StationA/gke-logging.git
cd gke-logging
poetry install

Usage

gke_logging.GKELoggingFormatter

One of the core components is the GKELoggingFormatter, which is an implementation of the built-in logging.Formatter protocol that translates a logging.LogRecord into a JSON format that GKE's logging infrastructure can understand. At a minimum, this enables any software running on GKE to integrate structured logging simply by applying this formatter for your loggers, e.g.:

import logging

from gke_logging import GKELoggingFormatter


LOGGER = logging.getLogger(__name__)
h = logging.StreamHandler()
h.setFormatter(GKELoggingFormatter())
LOGGER.addHandler(h)
LOGGER.setLevel(logging.INFO)


# ...

LOGGER.info("Look at me! I can haz GKE structured logging!")
# Prints out: {"time": "2022-01-13T23:22:26.336686+00:00", "severity": "INFO", "message": "Look at me! I can haz GKE structured logging!", "logging.googleapis.com/sourceLocation": {"file": "test_log.py", "line": "14", "function": "<module>"}, "logging.googleapis.com/labels": {}}

Furthermore, this formatter allows you to set app-level metadata to be sent along with each log message, which is useful in order to better organize collected log data:

# ...
h.setFormatter(
    GKELoggingFormatter(default_labels=dict(app_id="my-cool-app", version="0.1.0"))
)
# ...

Also the formatter also allows you to add HTTP metadata to any logs that occur during the course of a request. This enhances logs that are emitted during request-handling logic in APIs with additional data. This functionality is primarily utilized in the included GKELoggingMiddleware in order to provide basic access logs.

gke_logging.asgi.GKELoggingMiddleware

gke_logging.asgi.GKELoggingMiddleware is an ASGI middleware that emits basic access logs in "common log format", with a default behavior that integrates with the GKELoggingFormatter to write the access logs in a format that GKE's logging infrastructure better understands. By implementing per the ASGI spec, this means it can work with any ASGI-compatible server, including FastAPI, starlette, and ASGI implementations:

from fastapi import FastAPI
from gke_logging.asgi import GKELoggingMiddleware

app = FastAPI()
app.add_middleware(GKELoggingMiddleware)

@app.get("/")
def get_it() -> str:
    return "OK"

Additionally, because this middleware integrates with gke_logging.context bindings, it enables any logger used during the course of handling a request to emit logs that also contain request-time data, e.g. request URL, user-agent, response latency, etc.

import logging

from fastapi import FastAPI
from gke_logging import GKELoggingFormatter
from gke_logging.asgi import GKELoggingMiddleware

app = FastAPI()
app.add_middleware(GKELoggingMiddleware)


root_logger = logging.getLogger()
h = logging.StreamHandler()
h.setFormatter(GKELoggingFormatter())
root_logger.setLevel(logging.INFO)
root_logger.addHandler(h)


@app.get("/")
def get_it() -> str:
    # Any log records created during request-handling will be enriched with other HTTP request data
    root_logger.info("TEST")
    return "OK"

gke_logging.context

In order to control additional metadata labels for log records that correspond to one logical operation, e.g. an HTTP request, a batch job operation, etc., you should use the helper functions exported in gke_logging.context:

import logging

from contextvars import copy_context

from gke_logging import GKELoggingFormatter
from gke_logging.context import set_labels


LOGGER = logging.getLogger(__name__)
h = logging.StreamHandler()
h.setFormatter(GKELoggingFormatter())
LOGGER.addHandler(h)
LOGGER.setLevel(logging.INFO)

# ...


def run_job(job_id: str):
    set_labels(job_id=job_id)
    LOGGER.info("TEST")


ctx = copy_context()
for i in range(10):
    ctx.run(run_job, f"{i + 1}")

Because ContextVars bind natively to Python's asyncio, you can use these same helper functions within asynchronous tasks in a similar fashion.

Additional examples

Additional usage examples can be found in examples/

Contributing

When contributing to this repository, please follow the steps below:

1. Fork the repository
2. Submit your patch in one commit, or a series of well-defined commits
3. Submit your pull request and make sure you reference the issue you are addressing