search

K0nkere / MLOps_Car-prices-project

Covers the most important stages of ML system / Auction car prices prediction model for MLOps-zoomcamp
0 stars 0 forks source link

4 Monitoring #5

Open K0nkere opened 2 years ago

K0nkere commented 2 years ago

Lets create two services: prediction service and evidently service.

K0nkere commented 2 years ago

Prediction service for monitoring

Created on the base of prediction deployment of part 3 Its need to add import of environment variables to the Dockerfile

Creating test-image from Dockerfile located in /project/4-monitoring/prediction_service/ for prediction service docker build -t prediction-service-tst ./4-monitoring/prediction_service/ docker run -it --rm -p 9696:9696 prediction-service-tst We can test it with test.py and send-data.ipynb

K0nkere commented 2 years ago

Evidently service for monitoring

All needed files located in the /project/4-monitoring/evidently_service The only modifications is to update requirements.txt and section of config.yaml containing type of variables and test dataset.

From the folder /project/4-monitoring run docker-compose up --build (--build option is used for first creation of images or recreation after changings) it will create and run containers.

Maybe it will need to kill previous launched docker containers to release the ports docker ps docker kill <id>

Graphana will be accesible on 127.0.0.1:3000 (admin/admin) via local browser Maybe it will need to tunnel the port in VSCode or PuttySSH

If login/pass for dockerized Grafana is forgotten it is possible to reset it by running docker ps > grafana_id (like 01fc8622d337) docker exec -it 01fc8622d337 grafana-cli admin reset-admin-password admin