Maken Elasticsearch Configs

This repo contains a few instructions on setting up Open Distro for Elasticsearch (ES) for the National Library of Norway AI-Lab similarity (maken) service.

1. Create a new VM. Make sure you have allowed access to its IP on ports 9200, 5600 and 5602.
2. Follow the instructions in OpenDistro to install Open Distro for Elasticsearch: https://opendistro.github.io/for-elasticsearch-docs/docs/install/deb/ Additionally, the Kibana dashboard can also be installed: https://opendistro.github.io/for-elasticsearch-docs/docs/kibana/#run-kibana-using-the-rpm-or-debian-package
3. Since in most cases ES will be secured by internal IT, disable SSL and HTTPS. If self-signed, you'll need to run a local proxy for connecting GUIs like Elasticvue, as in PORT=3000 TARGET=https://user:[email protected]:9200 node proxy.js, and then connect to http://localhost:3000 instead.

Data could be ingested using maken-es-data. Once Elasticsearch is running with some data, the API can be deployed using Docker as follows:

$ docker build -t maken-es-api .
$ docker run -d -p 80:80 -e ES_HOST=192.166.0.10 --name maken-es-api maken-es-api

There are a few optional parameters to connect to Elasticsearch that can be set using environment variables:

• ES_HOST, to set host IP or name. Defaults to localhost
• ES_PORT, to set port. Defaults to 9200
• ES_USER, to set user. Defaults to admin
• ES_PASS, to set password. Defaults to admin
• ES_TIMEOUT, to set a request timeout. Defaults to 60
• ES_MAX_RETRIES, to set the max number of retries. Defaults to 300

Once up, OpenAPI documentation can be accessed at /docs. A test site can be found at api.nb.no/maken.

For development, it can be useful to run it using uvicorn:

$ uvicorn app.main:app