Quickstart

09/06/2022, 30 minutes to complete

Get started with the Logsight SDK for Python and client library to verify deployments. Follow these steps to install the package and start using the algorithms provided by logsight.ai service. The verification library enables you to evaluate the risk of a deployment of a new version of an application by automatically using deep learning models trained on millions lines of code, regardless of the underlying IT system, failure scenario, or data volume.

Use the Logsight SDK for Python to:

Send data logs to your logsight.ai account
Verify deployments using application data logs
Retrieve and display the verification risk of your new deployment

Prerequisites

logsight.ai subscription (create one for free)
You will need the login and password to paste into the code below, later

Setting up

Create a directory

Create a directory to store your exercise:

$ mkdir quick_guide
$ cd quick_guide

Create a virtual env

Create a Python virtual environment to decouple and isolate the packages we will install from you environment.

$ python3 -m venv venv
$ source venv/bin/activate

Prepare code file

You can start with an empty Python file:

$ touch quick_guide.py

Alternatively, you can download the Python file directly from git:

$ curl https://raw.githubusercontent.com/aiops/logsight-sdk-py/main/docs/source/quick_guide/quick_guide.py --output quick_guide.py

Install the client library

Install the Incident Detector client library for python with pip:

$ pip install logsight-sdk-py

or directly from the sources:

$ git clone https://github.com/aiops/logsight-sdk-py.git
$ cd logsight-sdk-py
$ python setup.py install

Create environment variables

Using the password from your subscription, create one environment variables for authentication:

LOGSIGHT_PASSWORD - Password for authenticating your requests
LOGSIGHT_EMAIL - Email associated with your subscription

Copy the following text to your bash file:

$ export LOGSIGHT_PASSWORD=<replace-with-your-password>
$ export LOGSIGHT_EMAIL=<replace-with-your-email>

After you add the environment variables, you may want to add them to ~/.bashrc.

For the impatient

mkdir quick_guide
cd quick_guide
python3 -m venv venv
source venv/bin/activate
curl https://raw.githubusercontent.com/aiops/logsight-sdk-py/main/docs/source/quick_guide/quick_guide.py --output quick_guide.py
pip install logsight-sdk-py
unset LOGSIGHT_PASSWORD LOGSIGHT_EMAIL
export LOGSIGHT_PASSWORD=mowfU5-fyfden-fefzib
export LOGSIGHT_EMAIL=logsight.testing.001@gmail.com
python quick_guide.py

Code example

The following code snippets show what can be achieved with the Logsight SDK client library for Python:

Authenticate the client
Set tags
Attach the logger
Log logging statements
Verify the new deployment
Show the results of the verification

Load packages

Load the various packages used in this guide.

import sys
import time
import logging

from logsight.config import set_host
from logsight.exceptions import InternalServerError
from logsight.authentication import LogsightAuthentication
from logsight.logger.logger import LogsightLogger
from logsight.compare import LogsightCompare

Authenticate the client

To enable client authentication, set your LOGSIGHT_PASSWORD and LOGSIGHT_EMAIL. If you use an on-prem deployment, setup the endpoint of your logsight system using function set_host.

EMAIL = os.getenv('LOGSIGHT_EMAIL') or 'logsight.testing.001@gmail.com'
PASSWORD = os.getenv('LOGSIGHT_PASSWORD') or 'mowfU5-fyfden-fefzib'
set_host("https://demo.logsight.ai/api/v1/")

auth = LogsightAuthentication(email=EMAIL, password=PASSWORD)

Attach the logger

Add logsight.ai logging handler to your logging system:

handler = LogsightLogger(auth.token)
handler.setLevel(logging.DEBUG)

logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)
logger.addHandler(handler)

Set tags

You can set tags using a dictionary

tags = {'service': 'redis', 'version': 'v1'}

or by using environment variables. To set tags as environment variables you need to use the prefix LOGSIGHT_TAG.

$ export LOGSIGHT_TAG_SERVICE=redis

from config import get_tags_from_env

env_tags = get_tags_from_env()  # {"service": "redis"}
handler.set_tags(tags=env_tags)

Execute Redis Version v1.1.1

We assume you are a core developer of Redis in-memory data structure store.

Run v1.1.1 of your Redis application
Logs are tagged with: service=redis and version=v1.1.1
Logs generated are transparently sent to logsight.ai

print('Redis running (v1.1.1)')
tags_1 = {'service': 'redis', 'version': 'v1.1.1'}
handler.set_tags(tags=tags_1)
for i in range(10):
    logger.info(f'Connecting to database (instance ID: {i % 4})')
    logger.info(f'Reading {i * 100} KBytes')
    logger.info(f'Closing connection (instance ID: {i % 4})')
handler.flush()

Your Redis deployment runs for several months without any problems. It is deemed reliable.

Execute Redis version v2.1.1

You implement a few new features for Redis. Your new version is v2.1.1.

Now, you run v2.1.1 of your Redis application in pre-production
Logs are tagged with: service=redis and version=v2.1.1
Logs generated are transparently sent to logsight.ai

print('Redis running (v2.1.1)')
tags_2 = {'service': 'redis', 'version': 'v2.1.1'}
handler.set_tags(tags=tags_2)
for i in range(15):
    logger.info(f'Connecting to database (instance ID: {i % 4})')
    logger.info(f'Unable to read {i * 100} KBytes')
    logger.error(f'Underlying storage is corrupted')
    logger.info(f'Closing connection (instance ID: {i % 4})')
handler.flush()

Verify New Release of Redis

You call the Compare function of Logsight to verify the new deployment

print('Calculate new deployment risk')
comp = LogsightCompare(auth.token)
result = {}
retry = 5
while retry:
    try:
        result = comp.compare(baseline_tags=tags_1, candidate_tags=tags_2)
        break
    except InternalServerError as e:
        print(f'Trying in 5s (#{retry})')
        time.sleep(5)
        retry -= 1

Show verification results

Display the deployment risk and access the webpage with the verification results

print(f'Deployment risk: {result["risk"]}')
print(f'Report webpage: {result["link"]}')

You can copy the url to your browser to see the results of the evaluation.

Run the application

Run the Python code from your quick_guide directory.

$ python quick_guide.py