Skip to main content

Overview

OpenSRE queries VictoriaLogs with LogsQL to retrieve structured log evidence — correlating application errors, request anomalies, and stream-level events with the alert under investigation.

Prerequisites

  • A reachable VictoriaLogs instance (for example http://vmlogs:9428)
  • LogsQL knowledge for any custom queries you want OpenSRE to run

Setup

There is no opensre integrations setup victoria_logs handler. Configure VictoriaLogs with environment variables or the persistent store only.

Option 1: Environment variables

Add to your .env:
Leave VICTORIA_LOGS_TENANT_ID unset for single-tenant clusters. OpenSRE does not send AccountID: 0 implicitly — that would target the default tenant on every request.

Option 2: Persistent store

Credentials

VictoriaLogs typically needs no API token for OpenSRE. Point VICTORIA_LOGS_URL at a reachable instance. On multi-tenant deployments, set VICTORIA_LOGS_TENANT_ID so OpenSRE sends the correct AccountID header.

Investigation tools

When VictoriaLogs is configured, the victoria_logs_query tool becomes available to the investigation agent. It runs LogsQL queries against /select/logsql/query and returns structured rows. The default is a wildcard match over the past hour; the agent narrows the query from alert context (service, level, trace ID, time window).

Verify

Expected output:

Troubleshooting

Security

Leave VICTORIA_LOGS_TENANT_ID unset for single-tenant clusters. OpenSRE does not send AccountID: 0 implicitly — that would target the default tenant on every request.

Extras

How it works in investigations

Typical agent uses:
  • Pull recent error-level logs for the affected service to surface stack traces
  • Filter by request ID or trace ID to follow a single failing transaction across services
  • Compare log volume in the incident window against a known-good baseline to spot regressions