humioapi 0.6.0

An unofficial Python library for easy interaction with the Humio API

Humio API (unofficial lib)

This project requires Python>=3.6.1

This is an unofficial library for interacting with Humio's API. If you're looking for the official Python Humio library it can be found here: humiolib. This library mostly exists because the official library was very basic back in 2019 when I first needed this.

Installation

pip install humiocli

Main features

• Asyncronous and syncronous streaming queries supported by httpx.
• Queryjobs which can be polled once, or until completed.
• Chainable relative time modifiers (similar to Splunk e.g. -1d@h-30m).
• List repository details (NOTE: normal Humio users cannot see repos without read permission).
• Easy env-variable based configuration.
• Ingest data to Humio, although you probably want to use Filebeat for anything other than one-off things to your sandbox.
• CLI companion tool available at humiocli.
• Create and update parsers.
• (PoC) An updateable timeseries, which can follow a moving timewindow using relative modifiers, optionally querying only the changed timewindow since previous update.

Usage

For convenience your Humio URL and token should be set in the environment variables HUMIO_BASE_URL and HUMIO_TOKEN. These can be set in ~/.config/humio/.env and loaded by humioapi.loadenv().

Query repositories

Create an instance of HumioAPI to get started

import humioapi
import logging
humioapi.initialize_logging(level=logging.INFO, fmt="human")

api = humioapi.HumioAPI(**humioapi.loadenv())
repositories = api.repositories()

Iterate over syncronous streaming searches sequentially

import humioapi
import logging
humioapi.initialize_logging(level=logging.INFO, fmt="human")

api = humioapi.HumioAPI(**humioapi.loadenv())
stream = api.streaming_search(
    query="log_type=trace user=someone",
    repos=['frontend', 'backend', 'integration'],
    start="-1week@day",
    stop="now"
)
for event in stream:
    print(event)

Itreate over asyncronous streaming searches in parallell, from a syncronous context

import asyncio
import humioapi
import logging
humioapi.initialize_logging(level=logging.INFO, fmt="human")

api = humioapi.HumioAPI(**humioapi.loadenv())
loop = asyncio.new_event_loop()

try:
    asyncio.set_event_loop(loop)
    tasks = api.async_streaming_tasks(
        loop,
        query="log_type=trace user=someone",
        repos=['frontend', 'backend', 'integration'],
        start="-1week@day",
        stop="now",
        concurrent_limit=10,
    )

    for event in humioapi.consume_async(loop, tasks):
        print(event)
finally:
    try:
        loop.run_until_complete(loop.shutdown_asyncgens())
    finally:
        asyncio.set_event_loop(None)
        loop.close()

Jupyter Notebook

pew new --python=python36 humioapi
# run the following commands inside the virtualenv
pip install git+https://github.com/gwtwod/humio-api.git
pip install ipykernel seaborn matplotlib
python -m ipykernel install --user --name 'python36-humioapi' --display-name 'Python 3.6 (venv humioapi)'

Start the notebook by running jupyter-notebook and choose the newly created kernel when creating a new notebook.

Run this code to get started:

import humioapi
import logging
humioapi.initialize_logging(level=logging.INFO, fmt="human")

api = humioapi.HumioAPI(**humioapi.loadenv())
results = api.streaming_search(query='log_type=trace user=someone', repos=['frontend', 'backend'], start="@d", stop="now")
for i in results:
    print(i)

To get a list of all readable repositories with names starting with 'frontend':

repos = sorted([k for k,v in api.repositories().items() if v['read_permission'] and k.startswith('frontend')])

Making a timechart (lineplot):

%matplotlib inline
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd

sns.set(color_codes=True)
sns.set_style('darkgrid')

results = api.streaming_search(query='log_type=stats | timechart(series=metric)', repos=['frontend'], start=start, stop=stop)
df = pd.DataFrame(results)
df['_count'] = df['_count'].astype(float)

df['_bucket'] = pd.to_datetime(df['_bucket'], unit='ms', origin='unix', utc=True)
df.set_index('_bucket', inplace=True)

df.index = df.index.tz_convert('Europe/Oslo')
df = df.pivot(columns='metric', values='_count')

sns.lineplot(data=df)