Skip to main content

Harry - Performance Scraper for Oracle Database

Performance Scraper for Oracle Database collects performance data on a schedule and stores it in PostgreSQL for Grafana dashboards and troubleshooting.

warning

Oracle Diagnostics Pack
The Oracle ASH collector is DISABLED by default.
Enabling it requires that YOU verify your Oracle Diagnostics Pack licensing.

This is not a Prometheus exporter. It does not expose Oracle metrics on /metrics. Instead, it:

  • connects to one or more Oracle databases,
  • collects native operational, SQL, session, blocking-session, and session-sampled database activity by default,
  • optionally collects additional SQL-derived metrics from TOML or YAML definitions,
  • writes those samples to PostgreSQL,
  • exposes a small health endpoint on /healthz,
  • lets Grafana read PostgreSQL directly.

Native performance data uses dedicated PostgreSQL tables. Details such as SQL_ID, child cursor, plan hash, SID, serial number, module, machine, wait event, and blocking session remain relational columns that can be indexed and queried by SQL-backed Grafana dashboards.

Main Features​

  • Scheduled Oracle scraping controlled by metrics.scrapeInterval.
  • PostgreSQL storage using batched inserts.
  • Daily range-partitioned PostgreSQL sample tables created on demand.
  • Optional PostgreSQL retention that drops old daily partitions.
  • Native operational collection for database state, instance load, resource limits, storage capacity, system counters, wait classes, and collection health.
  • Optional additional metrics loaded from ordered TOML or YAML definition files.
  • Direct performance collection from Oracle dynamic performance views:
    • GV$SQLSTATS
    • GV$SQL
    • GV$SQL_PLAN
    • GV$SESSION
    • optional GV$ACTIVE_SESSION_HISTORY, only when explicitly enabled
  • Grafana dashboards backed by PostgreSQL:
    • Database Activity History (DAH)
    • Oracle SQL Performance
    • Oracle SQL Top Consumers
    • Current Sessions and Blocking
    • Oracle Operational Overview
  • Provisioned starter Grafana alerts backed directly by PostgreSQL.
  • Oracle alert log export to JSON files.
  • Oracle wallet, external authentication, OCI Vault, Azure Vault, and HashiCorp Vault credential integrations inherited from the upstream codebase.
  • Builds with either godror and Oracle Instant Client, or the no-CGO go-ora driver using -tags goora.

PostgreSQL Tables​

The scraper writes to these primary PostgreSQL tables:

  • oracle_metric_samples
  • oracle_sql_samples
  • oracle_sql_texts
  • oracle_sql_plans
  • oracle_session_samples
  • oracle_blocking_session_samples
  • oracle_database_activity_samples
  • oracle_database_status_samples
  • oracle_instance_samples
  • oracle_resource_limit_samples
  • oracle_tablespace_samples
  • oracle_asm_diskgroup_samples
  • oracle_system_counter_samples
  • oracle_wait_class_samples
  • oracle_system_metric_samples
  • oracle_scrape_status

When output.postgresql.autoMigrate: true is configured, the scraper creates the parent partitioned tables, the SQL text and execution-plan lookup tables, and indexes automatically. Daily child partitions are created just before data is written.

Supported Oracle Versions​

The scraper is intended for Oracle Database 19c and newer. The default activity collector uses GV$SESSION. GV$ACTIVE_SESSION_HISTORY is accessed only when the operator explicitly configures performance.activity.source: ash.

Origins and acknowledgements​

Harry - Performance Scraper for Oracle Database is developed and maintained by Jorge Holgado and Ciberterminal S.L.

Harry originated as a fork of Oracle's database application observability project, which incorporated earlier work from Seth Miller's Oracle DB Exporter.

The inherited code provided the original Oracle connectivity and metrics collection foundation. Harry substantially changes that architecture by replacing the Prometheus exposition model with scheduled collection, historical persistence in PostgreSQL, and a dedicated database-performance analysis and visualization layer.

Copyright and licensing information for the original projects and Harry's fork-specific development is available in the project's LICENSE, LICENSES/, and THIRD_PARTY_NOTICES.md files.