Time Series Database
Horizon stores performance data in a time series database (JRobin by default). For different scenarios you may want to use a different time series database. The following implementations are installed by default:
Time Series Database | Description |
---|---|
JRobin |
JRobin is a clone of RRDtool written in Java and is the default time series database when you install Horizon. It does not fully cover the latest RRDtool feature set. Data is stored on the local file system of the Horizon node. Depending on I/O capabilities, it works well for small- to medium-size installations. |
RRDtool |
RRDtool is actively maintained and the de-facto standard for time series data. Data is stored on the local file system of the Horizon node. Depending on I/O capabilities it works well for small- to medium-size installations. |
Newts |
Newts is an OpenNMS database schema for Cassandra. The time series is stored on a dedicated Cassandra cluster, which gives growth flexibility and lets time series data persist at a large scale. |
This chapter describes how to configure Horizon to use RRDtool and Newts.
Time series database plugins
You can use time series database integrations with plugins based on our OpenNMS Plugin API (formerly OIA). Several plugins are available as a replacement for the ones shipped with OpenNMS Horizon (see Time Series Integration Layer).
The way data is stored in different time series databases makes it extremely hard to migrate from one technology to another. You cannot prevent data loss when you switch from one database type to another. |
Run two time series writes in parallel
The default time series strategy (RRD, Newts, Integration Plugin) is used for both read and write operations. You can add another time series to write in parallel to the default strategy. This additional time series database may be used for validation purposes or for staging metrics before moving from one time series to another. If you plan to migrate to a new storage strategy, you can set the new strategy as a write destination while keeping reads on your current storage. Once you have written enough data to the new storage, you can then flip the reads over to the new system.
In order for the dual write plugin to work, you must install a time series feature (either Newts or an integration plugin) and leave org.opennms.timeseries.strategy
at your current value.
This will allow writing to both the strategy and the time series database.
Write to Newts in parallel
opennms-newts
feature:sudo vi etc/featuresBoot.d/newts.boot
opennms-newts
sudo systemctl restart opennms
Write to time series integration plugin in parallel
You can enable writing to any time series integration plugin in parallel to RRD or Newts by enabling the opennms-timeseries-api
feature and not setting the strategy to integration
.
This will direct writes to both your org.opennms.timeseries.strategy
destination and any installed timeseries integration plugin.
opennms-timeseries-api
feature:sudo vi etc/featuresBoot.d/timeseries.boot
opennms-timeseries-api
sudo systemctl restart opennms