Changelog

note

Please see the upgrade guide for guidance on upgrading from the preview version.

Version 3.5 includes major improvements to observability, through the introduction of a new message system between system components.

Agent Changes

• Agents gain an additional API for communication with Pipes, which by default listens on 127.0.0.1:4041.
  • This setting can be controlled via:
    • The command line option --agent-api-bind-address, or
    • The environment variable HOTROD_AGENT_API_BIND_ADDRESS
  • The default bind address is 127.0.0.1:4041 on Agents, and 127.0.0.1:3041 the Server Agent

Server Changes

• Server Agent changes
  • The Server now always includes the Server Agent, where previously it was an optional Server feature
  • The Server Agent has been renamed:
    • previously, Hotrod Server Agent (with ID hotrod-server-agent)
    • now Built-In Agent (with ID built-in-agent)
    • When upgrading a Server, the previous name and id will be retained for backwards compatibility
  • The built-in Agent's default setting for it's existing Pipe API changed from 127.0.0.1:4040 to 127.0.0.1:3040
    • This change avoids port conflicts with default settings when an Agent is co-located with the Server process on the same host

Pipe Language & Runtime Changes

Inputs and Outputs

• The new internal-messages input introduces a powerful capability to subscribe to various telemetry events generated by the Server, Agents and Pipes.

• The new message output allows for generating custom events, which can be subscribed to with the new internal-messages input

• The s3 and azure-blob inputs:

  • support a new preprocessors option, which enables:
    • handling gzip-compressed input objects
    • handling parquet-encoded input objects
    • automatically applying either of the above, depending on file extensions (.gz or .parquet)
  • supports a new fingerprinting-db-path option which allows customizing the path of the YAML state file

• The s3 and azure-blob outputs:

  • support a new preprocessors option, which supports compression of output objects with gzip
  • support a new input-field option

• The file output supports a new append-to-existing option, for appending data to existing files

• The http-poll and scuba inputs support a new timeout parameter, for controlling the timeout on HTTP requests

• New: initial MQTT client and virtual broker support (authentication or TLS not yet available)

• The kafka input and output support a new config option for specifying rdkafka configuration parameters

• The azure-monitor output supports a new time_generated_field option

• The scuba input now supports specifying either a query response id or BQL query

• The azure-monitor input now has improved feedback from errors

• The splunk-hec output supports a new remove option that can remove certain default fields from the output payload

• The splunk-hec output supports new index and index-field metadata options

Actions

• A new LUA method, batch_number(), is available in the script action and other actions that can evaluate LUA expressions

  • This method returns the last batch number of a Pipe's output, or 1 if the Pipe's output doesn't use batch mode
  • This feature should not be used for generating output file or object names; timing differences in processing between Pipe actions and outputs may cause unpredictable behaviour, due to timing differences in the processing of actions and output components of Pipes.

• The add action supports a new jsonpath-fields option, which allows for extracting data using JSONPath expressions

• The transaction action supports a new combined-payload option, which combines collected events into a single event and concatenate payloads

• The transaction action now supports looking for transaction start and end markers

• The filter action gets a new option to drop empty JSON (nil, arrays, strings, etc)

• The collapse action now has improved handling for the creation of CSV data