Deprecate performance-analyzer-rca

[Gaganjuneja]

Deprecation of performance-analyzer-rca project

[Gaganjuneja]

Deprecation Notice: performance-analyzer-rca

We would like to inform the community that the performance-analyzer-rca will be deprecated starting with OpenSearch version 2.19 and discontinued from the version 3.0. The RCA Agent was initially designed to read the Performance Plugin’s metrics output, with the goal of executing root cause analysis (RCA) at both the node and cluster levels. It achieved this by reading metrics from the Performance Analyzer, publishing them to a metrics database, and running analysis on the collected data.

However, the performance-analyzer-rca has a few architectural and performance limitations that have impacted its scalability and resource efficiency:

• Scaling Constraints: The architecture of RCA Agent involves connecting every node to the master node over gRPC connection, which imposes scaling limitations on the cluster.
• Resource Consumption: Although the performance-analyzer-rca agent runs as a separate sidecar process, it shares system resources (CPU, memory, network, and IO) with OpenSearch, leading to performance issues. There have been instances where high resource consumption by the RCA Agent negatively affected OpenSearch processes.
• Integration Challenges: There has been no straightforward way to extract and integrate RCA Agent metrics (stored in the metrics DB, which rolls every 5 seconds) with external monitoring systems.

With the introduction of the Telemetry Plugin, we now have a more efficient and flexible way to handle performance analyzer metrics. The Telemetry Plugin, using the OpenTelemetry framework, allows for seamless integration with lightweight open-source agents to publish performance metrics to observability stores. This shift has already been discussed in detail in the RFC here (opensearch-project/performance-analyzer#585), and some of the existing collectors have already been modified to push data to the Telemetry Plugin's metric registry.

What this means

• RCA Agent will no longer receive new development from OpenSearch version 2.19 onward.
• Users are encouraged to transition to the Telemetry Plugin for better performance and scalability.

We appreciate your understanding and support as we continue to improve the OpenSearch ecosystem.

[peterzhuamazon]

Hi @Gaganjuneja thanks for the issue.

Just want to be clear PA + RCA plugin will stay just without update right? Means we are still planning to bundling them as part of the distribution?

Thanks.

[Gaganjuneja]

PA OpenSearch plugin will stay as part of the bundle but performance-analyzer-rca is a side car process to read PA plugin data will be deprecated and PA plugin would start publishing the metrics to telemetry now.

[krisfreedain]

Catch All Triage - 1