Error when running the postgres exporter with postgres 17

[cipherself]

What's wrong?

postgres 17 removed the checkpoints_timed column from pg_stat_pgwriter causing the following error

alloy  | ts=2025-01-11T23:43:24.476671414Z level=error msg="collector failed" component_path=/ component_id=prometheus.exporter.postgres.dbname name=stat_bgwriter duration_seconds=0.005531275 err="pq: column \"checkpoints_timed\" does not exist

Apparently this has been fixed on the master branch of postgres-exporter prometheus-community/postgres_exporter#1072 but still hasn't been released.

Steps to reproduce

1. Run a postgres 17 database
2. Configure alloy to scrap metrics from it

System information

Linux 6.1.0-25-amd64

Software version

v1.5.1

Configuration

prometheus.exporter.postgres "testdb" {
  data_source_names = ["postgresql://monitoring:monitoring@db:5432/testdb?sslmode=disable"]
}

prometheus.scrape "testdb" {
  targets    = prometheus.exporter.postgres.testdb.targets
}

Logs

alloy  | ts=2025-01-11T23:43:24.476671414Z level=error msg="collector failed" component_path=/ component_id=prometheus.exporter.postgres.dbname name=stat_bgwriter duration_seconds=0.005531275 err="pq: column \"checkpoints_timed\" does not exist

[wildum]

hi, thanks for reporting this bug. We are currently running a fork of the postgres exporter. There is a todo to update our fork with upstream (https://github.com/grafana/alloy/blob/main/go.mod#L942). This bug is extra motivation to get it done :) @dehaansa you looked into this exporter recently, any pain points to get it updated?

[dehaansa]

Yeah, there were some relatively significant changes to get upstreamed, it might be more straightforward to cherry pick the fix for postgresl17 before continuing on the upstream work.