cluster-logger 0.5.0

Library for wrapping cluster logging for Altamira projects.

Cluster Logging library for python 3.6

This is a library for python 3.6 that will help fulfil a requirement for cluster services and logging to elastic search.

Requirements

• Python 3.6+
• pipenv
• Elasticsearch
• Fluentd
• Kibana

Pipenv & Pipfile

Pipenv combines package management and virtualenv into one tool.
pipenv: https://docs.pipenv.org/basics

To install:

$ pip install pipenv

Pipfile replaces requirements.txt and it will specify both dependencies and dev dependencies in one file.

Development Installation

# install dependencies
pipenv install --dev

# activate virtual environment
pipenv shell

# alternativly you can start your script in a virtual environment context
pipenv run python main.py message that you want to send

ULM Configuration

The default configuration uses Environment variables to configure the logger.
These values were grabbed from version 2.0 of ULM-logging-common Java implementation. Not every value is implemented yet and changes are upto date in table.
Repo

Variable	Default Value	Description	Implemented
ULM_FLUENTD_HOST	localhost	host of the FluentD service	✔️
ULM_FLUENTD_PORT	24224	port of FluentD tcp "in_forward"	✔️
ULM_FLUENTD_LABEL_ENVS		comma separated list(with or without space) of other ENV vars to add as fields to FluentD logs	✔️
ULM_FLUENTD_TAG	(project).(app)	tag to use when sending logs to fluentd, e.g. the MLA_IDENT string such as "Myapp.logs"	✔️
ULM_FLUENTD_BUFFER	1048576	an integer of the number of bytes to allow in the buffer. Excess will be dropped (default is 1MiB)	❌
ULM_FLUENTD_TIMEOUT	3.0	an float of the number of ms to allow for communication with FluentD.	✔️
ULM_FLUENTD_CLIENT	async	Whether to send async messages or not. This will send async as long as the value has async somewhere in the string.	✔️
ULM_FLUENTD_ENABLE_TIMEMS_FIELD	False	Enables an additional field to contain the millisecond timestamp	❌
ULM_FLUENTD_TIMEMS_FIELD	timems	the name of the field to contain the millisecond timestamp	❌

Standalone Usage (Development)

It may be required to run the following command to get elastic search to run.

sudo sysctl -w vm.max_map_count=262144

Local

$ pipenv run python main.py -h
usage: main.py [-h] [-n HOST] [-p PORT] [-e [ENV [ENV ...]]] [-o PROJECT]
               [-a APP] [-c COUNT]

Console Application to test Cluster Logging using fluentd.

positional arguments:
  message               Message to be logged.

optional arguments:
  -h, --help            show this help message and exit
  -n HOST, --host HOST  Hostname to listen on. (default=localhost)
  -p PORT, --port PORT  Port number to bind to. (default=24224)
  -e [ENV [ENV ...]], --env [ENV [ENV ...]]
                        Environment Keys to add to message.
  -o PROJECT, --project PROJECT
                        Tag for the project
  -a APP, --app APP     Tag for the application
  -c COUNT, --count COUNT
                        The number of times to send message.

Docker

$ docker build -t TAG_FOR_THE_BUILD:VERSION .

# all env values can be set with -e

$ docker run -e host=fluentd \
  -e port=24224 -e msg="message you want to send" TAG_FOR_THE_BUILD:VERSION

# If you want to run the whole environment you can use docker-compose up

$ docker-compose up

Api

Follows the python logger implementation with logging levels. python.org

Level	Numeric value
CRITICAL	50
ERROR	40
WARNING	30
INFO	20
DEBUG	10
NOTSET	0

Logging level can be set for logging using:

import logging
logging.basicConfig(level=logging.WARNING)

Setting the logging level will filter levels below the set logging level. For logging levels error and above the stack trace will be added to the log message.

Standard Configuration

These environment variables can be set either system-wide or added to the .env file:

There are 3 Environment variables that will be extracted by default:

• HOST - host that the container is running on.
• APP_VERSION - Version of the app.
• LOGGER_VERBOSE - Whether the fluent logger library is verbose or not.
• MARATHON_APP_ID - The ID that is assigned by DCOS.
• MARATHON_APP_DOCKER_IMAGE - The image that the container was built from.

There are three ways to get application properties into log messages.

• pass a dictionary of key value pairs
• pass a list of keys that are on the container environment
• or pass a JSON file with key value pairs.

NOTE:

• The dictionary will overwrite values in environment and JSON file.
• The Environment will overwrite values in the JSON file.

Usage

Activate your venv

# Linux / MacOS
source {NAME_OF_VENV}/bin/activate

REM Windows
{NAME_OF_VENV}\Scripts\activate

Install cluster logger into your venv. Cluster Logger has been installed into pypi.

$ pip install cluster_logger

To Use

import cluster_logger
import logging
# Then init the ClusterLogger class

env_keys = ['MARATHON_APP_LABELS', 'MARATHON_APP_RESOURCE_CPUS']
props = {'HOST', '192.168.0.12'}

logging.basicConfig(level=logging.DEBUG)             # set logging level
cluster_logger.initLogger('Rasters',                 # Project name
                          'intent_client',           # Application name
                          'fluentd',                 # Fluentd Host name (optional) will default to fluentd if nothing is passed in and HOST isn't set on the environment
                          24224,                     # Fluentd Port Number(optional)
                          '/path/to/settings.json',  # JSON settings file (optional)
                          env_keys,                  # Environment Keys (optional)
                          props)                     # Specific Properties (optional)

# in Module where you will log
import cluster_logger

# General Logger
logger = cluster_logger.getLogger(__name__)
# Metric Logger
metrics = cluster_logger.getLogger(__name__ + '.metrics', is_metric=True)

logger.exception('Custom Message to send typically an error')
logger.log(logging.INFO, {'Duration', 58})

Contributing

# bump the version that is stored in setup.py and cluster_logger/__init__.py
# assuming pipenv install --dev was ran first
$ pipenv run bumpversion minor # possible: major / minor / patch
$ git push