Oracle Enterprise Manager or direct database pulls?

Grafana Configuration
,
1

  • What Grafana version and what operating system are you using? 12.2 Enterprise on-prem

  • What are you trying to achieve? Monitor Oracle databases across the enterprise without the Oracle-provided Grafana plugin

  • How are you trying to achieve it? Researching that.

  • What happened? Per Grafana Download page | Oracle, Oracle no longer has a compatible plug-in for Grafana 12.2 and we would like to avoid pulling directly from hundreds of individual databases. Would like to know what others are doing given this situation? We are recently starting to implement Grafana and had not yet configured Oracle integration. Is there an OEM plugin provided by others in lieu of Oracle’s or are most user pulling data from individual databases using Mimir?

2

I ran into essentially the same gap, so I built an open-source project around a different architecture.

Full disclosure: I am the author of Harry — Performance Scraper for Oracle Database.

With Harry, Grafana does not connect directly to the Oracle databases and no Oracle Grafana plugin is required. The data flow is:

Oracle databases → Harry collectors → PostgreSQL → Grafana

Harry collects operational and performance data from multiple Oracle databases on a schedule and stores it in a central PostgreSQL repository. Grafana then uses its native PostgreSQL data source.

For low-cardinality infrastructure metrics, Prometheus/Mimir can be a good fit. For historical analysis involving SQL IDs, sessions, waits, blocking, execution plans and other high-cardinality relational data, I preferred PostgreSQL rather than turning those fields into metric labels.

The architecture supports centralized multi-database collection and active/standby Harry instances with leader election. The largest deployment I currently have monitors approximately 25 Oracle databases, so I would not claim hundreds without proper sizing and testing.

Harry is open source and can be deployed independently:

It is not an OEM plugin or an OEM replacement. It is a separate collection and historical analysis pipeline designed specifically to avoid making Grafana query every Oracle database directly.

3

Hi @dodger_one thanks for sharing this information. The Oracle DBs → Harry collectors → PostgreSQL → Grafana architecture looks like a good approach for our use case, especially since we want to avoid connecting Grafana directly to hundreds of Oracle databases.

4

Thanks very much. I really enjoy discussing architecture; for me, it is essential.

For a deployment monitoring on the order of hundreds of Oracle databases, I would carefully assess collector sizing, retention, PostgreSQL capacity and high availability overall.

I’d be happy to exchange ideas about the architecture, sizing and performance in general.