Observability

PerfStack Performance Issue: Slow or Not Loading Due to Long-Running AddToCortex Operations in SolarWinds Platform 2025.2 and below

This article provides information about an issue in SolarWinds Platform where the PerfStack widget fails to display metrics, often accompanied by delays or timeouts in real-time polling. The issue is related to the "AddToCortex" verb taking excessive time to execute, impacting performance.

First published date

10/25/2025 4:58 PM

Last published date

6/3/2026 1:32 PM

Overview

In versions prior to SolarWinds Platform 2025.4, Cortex handled real-time polling (RTP) for PerfStack.

In some environments, the AddToCortex verbs could take several minutes to complete, resulting in slow or unresponsive PerfStack dashboards.

Example SWISv3 WARN log entries:

LONG RUNNING QUERY: OperationContextId <GUID> - Query took <duration> ms: Verb Cortex.Orion.Node.Core.AddToCortex
LONG RUNNING QUERY: OperationContextId <GUID> - Query took <duration> ms: Verb Cortex.Orion.Interface.Core.AddToCortex

These long-running calls often caused web requests to exceed load time thresholds and contributed to higher CPU and memory utilization on systems using Cortex-based RTP.

Product section

Hybrid Cloud Observability

Cause

  • Legacy Cortex service handling real-time polling causes excessive resource consumption due to inefficient processing and occasional memory leaks.
  • AddToCortex verb operations can become slow due to processing delays, increased query times, and load on the Cortex service.

Resolution

This issue is resolved in SolarWinds Platform 2025.4 via Collector Real-Time Polling.

 

After upgrading to SolarWinds Platform 2025.4 or later:

  • PerfStack now uses Collector Real-Time Polling (RTP) instead of Cortex.
  • AddToCortex calls are no longer used.
  • The migration reduces CPU/memory utilization and eliminates long-running AddToCortex operations.

No manual configuration is required, the migration occurs automatically as part of the platform upgrade.