thinknet-observer 0.4.14

lib in development

thinknet-observer-python

A library use for collect metrics, log and tracing.

Installation

pip install thinknet-observer

Get started

Metrics

How to collect metrics with this lib:

from thinknet_observer import PrometheusMiddleware


# Instantiate a app framework
app = Flask(__name__) # or app = FastAPI()

# use middleware to collect default metrics
middleware = PrometheusMiddleware(app) # or PrometheusMiddleware(app,is_multiprocessing = True) when us multiprocess
middleware.register_metrics()

# add exclude (optional)
list_exclude = [{"route": "/healthz"}]
middleware.add_exclude(list_exclude)

to create custom metrics can be done with following code

from  thinknet_observer  import  MetricCollector

# NOTE: metrics name(1st param) must be unique for each metrics

# Custom gauge 
# (for gauge metrics if using multiprocess add param 'multiprocess_mode="livesum"')
CUSTOM_GAUGE = MetricCollector.gauge(
    "CUSTOM_GAUGE", "desc of CUSTOM_GAUGE", ["something"]
)
CUSTOM_GAUGE_NOLABEL = MetricCollector.gauge(
    "CUSTOM_GAUGE_NOLABEL", "desc of CUSTOM_GAUGE_NOLABEL"
)


# Custom summary
CUSTOM_SUMMARY = MetricCollector.summary(
    "CUSTOM_SUMMARY", "desc of CUSTOM_SUMMARY", ["something"]
)
CUSTOM_SUMMARY_NOLABEL = MetricCollector.summary(
    "CUSTOM_SUMMARY_NOLABEL",
    "desc of CUSTOM_SUMMARY_NOLABEL",
)


# Custom histogram
CUSTOM_HISTOGRAM = MetricCollector.histogram(
    "CUSTOM_HISTOGRAM", "desc of CUSTOM_HISTOGRAM", ["something"]
)
CUSTOM_HISTOGRAM_NOLABEL = MetricCollector.histogram(
    "CUSTOM_HISTOGRAM_NOLABEL", "desc of CUSTOM_HISTOGRAM_NOLABEL"
)
CUSTOM_HISTOGRAM_NOLABEL_CUSTOMBUCKET = MetricCollector.histogram(
    "CUSTOM_HISTOGRAM_NOLABEL_CUSTOMBUCKET",
    "desc of CUSTOM_HISTOGRAM_NOLABEL_CUSTOMBUCKET",
    buckets=[0.5, 0.75, 1],
)


# Custom counter
CUSTOM_COUNTER = MetricCollector.counter(
    "CUSTOM_COUNTER", "desc of CUSTOM_COUNTER", ["something"]
)
CUSTOM_COUNTER_NOLABEL = MetricCollector.counter(
    "CUSTOM_COUNTER_NOLABEL", "desc of CUSTOM_COUNTER_NOLABEL"
)

Example of custom metrics usage

@app_flask.route("/inc_gauge/<number>", methods=["POST"])
def inc_gauge(number):
    CUSTOM_GAUGE.labels("something'value").inc(float(number))
    CUSTOM_GAUGE_NOLABEL.inc(float(number))
    return {"msg": f"inc {number}"}

@app_flask.route("/dec_gauge/<number>", methods=["POST"])
def dec_gauge(number):
    CUSTOM_GAUGE.labels("something'value").dec(float(number))
    CUSTOM_GAUGE_NOLABEL.dec(float(number))
    return {"msg": f"dec {number}"}

@app_flask.route("/summary_observe/<number>", methods=["POST"])
def summary_observe(number):
    CUSTOM_SUMMARY.labels("something'value").observe(float(number))
    CUSTOM_SUMMARY_NOLABEL.observe(float(number))
    return {"msg": f"summary_observe {number}"}

@app_flask.route("/histogram_observe/<number>", methods=["POST"])
def histogram_observe(number):
    CUSTOM_HISTOGRAM.labels("something'value").observe(float(number))
    CUSTOM_HISTOGRAM_NOLABEL.observe(float(number))
    return {"msg": f"histogram_observe {number}"}

@app_flask.route("/histogram_observe2/<number>", methods=["POST"])
def histogram_observe2(number):
    CUSTOM_HISTOGRAM_NOLABEL_CUSTOMBUCKET.observe(float(number))
    return {"msg": f"histogram_observe {number}"}

@app_flask.route("/count/<number>", methods=["POST"])
def count_metric(number):
    CUSTOM_COUNTER.labels("something'value").inc(float(number))
    CUSTOM_COUNTER_NOLABEL.inc(float(number))
    return {"msg": f"count {number}"}

Additionally must include those environment variable when using multiprocess mode

PROMETHEUS_MULTIPROC_DIR=./prometheus_multiproc/ (folder for multiprocess metrics)

Note: only include variable PROMETHEUS_MULTIPROC_DIR when use multiprocess otherwise don't

In case of using gunicorn, gunicorn.config.py file must include

in order to make multiprocess work

import os

from pathlib import Path
from multiprocessing import cpu_count
from os import environ
from prometheus_client import multiprocess
from thinknet_observer import clear_multiproc_dir

def child_exit(server, worker):
    multiprocess.mark_process_dead(worker.pid)

def max_workers():    
    return cpu_count()

clear_multiproc_dir()