@navikt/oracle-to-influx v0.10.0

node-oracle-to-influx

Node script/server that moves data to based on polling an oracle based on certain intervals.

Configuration

The server will take a config which will be something like this

export default {
    measurementName: 'someMeasurement',
    schedule: '* * * * *',
    queryString: `SELECT 
    CREATED as TIME,
       DUMMY_TAG,
       COUNT(ID) as COUNT_OF_FOO
FROM TEST_TABLE_NAME
WHERE CREATED > TO_TIMESTAMP_TZ(:UPDATED_TIME, 'YYYY-MM-DD"T"HH24:MI:SS.FF3TZR')
GROUP BY CREATED, DUMMY
    `,
    schema: 'TEST_NODE_ORACLE_TO_INFLUX',
    influx: {
      url: process.env.INFLUX_URL || 'http://localhost:8086',
      database: 'metrics',
      username: 'admin',
      password: '',
    },
    fields: {
      'COUNT_OF_FOO': 1,
    },
    tags: [
      'DUMMY_TAG',
    ],
    oraOptions: {
      connectString: process.env.ORA_CONNECT_STRING || 'localhost:1521/XE',
      user: process.env.ORA_USERNAME || 'username',
      password: process.env.ORA_PASSWORD || 'Pas$w0rd',
    }
  }

The config is as follows:

The query

Influx db takes a timestamp as primary key along with the TAGS. This makes a row idempotent, meaning that you can run the same query many times without getting duplicate data. Using the timestamp is a limitation, but after working with influx-db a smart one. So if something happens on the exact same time, with the same tags the field need to be grouped. You should read more about this here InfluxDB key concepts.

This script uses the time column to not having to update big tables, but be able to read just whats needed. For this to happen you need to add a line to your query where ora-to-influx can inject the time. The example query under explains this:

SELECT 
    CREATED as TIME,
       DUMMY_TAG,
       COUNT(ID) as COUNT_OF_FOO
FROM TEST_TABLE_NAME
WHERE CREATED > TO_TIMESTAMP_TZ(:UPDATED_TIME, 'YYYY-MM-DD"T"HH24:MI:SS.FF3TZR')
GROUP BY CREATED, DUMMY

For performance reasons, you should not set any ORDER BY in the query as its not needed. ora-to-influx uses oracle streaming and will split huge resultset into multiple queries towards influxdb. This way you could read millions of rows from oracle without thinking about performance. At least this should work in theory.

Creating your own service

To create your own ora-to-influx service you need to create a simple server script. Startup-scripts etc is custom as this might vary from network to network. The server could be created as simple as:

1. create a folder and go into it
2. Run npm init && npm install @navikt/oracle-to-influx
3. Add a server.js file
4. Add an dockerimage extending for instance FROM pipekung/node-oracle:latest or another dockerimage with instant client installed. See this repository for a setup.
5. Create a folder or a file called to your config. This should be included as a javascript file and passed to the bootstrap function. The config should be wrapped in an array, as you probably would like to run multiple scripts.

Example server.js

const rawConfig = require('./config')
const bootstrap = require('@navikt/oracle-to-influx').bootstrap
const port = process.env.PORT || 8081
const options = {}
bootstrap(rawConfig, options).listen(port)

Service endpoints

The service is stateless, meaning that there is no changes to the service after its deployed. All changes is done to your repo and deployed. This way all your changes will be version controlled. Anyway there are some usefull endpoints that you can use to debug the service:

For local package development

Not explained into detail, just here for personal reference

• yarn link "@navikt/oracle-to-influx" and
• yarn add @navikt/oracle-to-influx