Changes in 3.0

OpenTSDB 3.0 is mostly a re-write of the original code with a focus on making it a time series data routing and analysis engine. The new framework goals include:

• Support for multiple types of data sources such as HBase, Bigtable, cloud providers and the latest time series DBs.

• Pluggable everything with support for custom data types, functions, etc.

• Streaming support for large queries that won’t fit in memory.

• Flexible querying by providing an executable graph of operations.

• Support for multiple query languages such as OpenTSDB, PromQL, SQL, etc.

• Provide backwards compatibility with previous versions to easy migration.

New Features

Noteable new features that have been implemented so far include:

• A new configuration system that supports properties, JSON or YAML (and eventually remote sources) that allows for configuration via: * Environment variables * Java system properties * Command line arguments * Java properties, YAML or JSON encoded files

• A new low-level DSL for querying that allows for flexible ordering of operations (e.g. group by then downsample or downsample then group by) and merging data from multiple sources.

• HA query support, querying two or more sources that should have the same data. E.g. when running HBase and dual-writing to two different clusters, we can query the two clusters and merge the results so that one can be in maintenance mode.

• Faster queries from HBase by re-writing the query pipeline.

• New aggregation functions and more advanced expressions.

Whats Missing

Stuff we have to get to include:

• The old Telnet style API, at least we’ll maintain the write method though we may drop the other commands.

• A GUI. Right nowe we only have the HTTP API.

• OS packages.

• Histograms and annotations.