Oracle DB Exporter in Kubernetes: Handling HA and Duplicate Queries

[ankitakhot92]

I'm currently using Oracle DB Exporter to expose database metrics to Prometheus and visualize them in Grafana. I’d like to understand the best-practice approach for setting up High Availability (HA) for the exporter. Currently, I have deployed the Oracle DB Exporter as a Kubernetes Deployment. When I increase the number of replicas (say 3 replicas), each exporter replica independently runs the same SQL queries against the database, which effectively triples the query load. This behavior isn’t ideal, so I’m looking for a recommended solution to achieve HA without adding unnecessary load to the database.

[anders-swanson]

Hi @ankitakhot92, this sounds like a great feature request. Let me think about how this could be implemented - from what you describe, it sounds like you're looking for a scenario where 1 exporter runs the queries, and any replicas are effectively in "standby" mode. Is that correct?

[anders-swanson]

One way this could be implemented is with DBMS_LOCK.

Per scrape, exporters request the lock. It they acquire it, they are the leader for a given lease. The delegates "standby" exporters to periodically checking the lock. It also has the benefit of not being tied to something like Kubernetes client leader election, which would require Kube-API permissions and additional dependencies.

In practice, you could select one of the connected databases as the leader election host for the exporter (using the config file). If this is set, leader election implemented with DMBS_LOCK would occur - otherwise the exporter would effectively run in "standalone" mode.