Monitoring with Prometheus

Dawarich exports application and Sidekiq metrics in the Prometheus exposition format. Since 1.7.7, the metrics backend is Yabeda (replacing discourse/prometheus_exporter). Metrics are now served in-process by each application — no separate exporter container or sidecar is required.

If you're upgrading from a version before 1.7.7, see the 1.7.7 entry in the Updating guides for the metric-name rename table and the env-var changes you need to apply.

Configuration

Enable the exporter

Set the same flag on both the dawarich_app and dawarich_sidekiq services in your docker-compose.yml:

Variable	Default	Description
PROMETHEUS_EXPORTER_ENABLED	false	Master on/off switch. Must be true on both services.
METRICS_USERNAME	(unset)	HTTP basic auth username Prometheus uses to scrape /metrics. Required.
METRICS_PASSWORD	(unset)	HTTP basic auth password. Required.
PROMETHEUS_EXPORTER_PORT	9394	Port on dawarich_sidekiq that serves the in-process Sidekiq metrics endpoint.
SIDEKIQ_METRICS_URL	http://dawarich_sidekiq:9394/metrics	Internal URL the web container uses to fetch Sidekiq metrics. Override on Dokku/Kubernetes or any deployment where the worker hostname differs from the compose default.

Behaviour notes:

• Single scrape target. Prometheus scrapes dawarich_app:3000/metrics. The web service fetches Sidekiq metrics over the internal network and concatenates them into the same response.
• Graceful Sidekiq fallback. If Sidekiq is unreachable during a scrape, the web container returns its own metrics only and logs a warning. Prometheus sees a momentary gap in sidekiq_* series instead of a failed scrape.
• Basic auth is required. Requests without valid METRICS_USERNAME / METRICS_PASSWORD credentials return 401.
• No more PROMETHEUS_EXPORTER_HOST / PROMETHEUS_EXPORTER_HOST_SIDEKIQ. These environment variables were removed in 1.7.7 — metrics are no longer hosted by a separate process, so there is nothing to bind to.

Prometheus scrape config

scrape_configs:
  - job_name: dawarich
    metrics_path: /metrics
    basic_auth:
      username: prometheus     # value of METRICS_USERNAME
      password: prometheus     # value of METRICS_PASSWORD
    static_configs:
      - targets: ['dawarich_app:3000']

If you're scraping from outside the Docker network, replace dawarich_app:3000 with the host and port you exposed for the dawarich_app service.

Metrics

HTTP / Rails

Type	Name	Description
Counter	rails_requests_total	Total HTTP requests handled by the web app.
Histogram	rails_request_duration	End-to-end request duration in seconds.

Sidekiq

Type	Name	Description
Counter	sidekiq_jobs_executed_total	Jobs executed (per worker class / queue).
Counter	sidekiq_jobs_failed_total	Jobs that raised an exception.
Histogram	sidekiq_job_runtime_seconds	Per-job runtime in seconds.
Gauge	sidekiq_queue_latency	Oldest enqueued job's wait time per queue, in seconds.
Gauge	sidekiq_jobs_waiting_count	Jobs sitting in each queue (backlog).
Gauge	sidekiq_active_processes	Sidekiq processes currently up.

Puma

Type	Name	Description
Gauge	puma_workers	Active Puma worker processes.
Gauge	puma_backlog	Requests waiting on the listen socket.
Gauge	puma_pool_capacity	Remaining thread-pool capacity.

ActiveRecord

Type	Name	Description
Gauge	activerecord_connection_pool_size	Maximum allowed pool size.

Dawarich-specific

dawarich_archive_* metric names are unchanged from the pre-Yabeda era — existing dashboards and alerts built on them continue to work.

Additional examples:

# HELP ruby_dawarich_map_tiles_usage
# TYPE ruby_dawarich_map_tiles_usage counter
ruby_dawarich_map_tiles_usage 99

The ruby_dawarich_map_tiles_usage counter tracks total map tiles loaded across users. You can find a chart of this in the user map settings.

Process / GC metrics

Process and GC metrics that prometheus_exporter used to emit by default (ruby_rss, ruby_heap_live_slots, etc.) are not emitted by default under Yabeda. If you need them, register a custom Yabeda group in your fork or wait for them to be added upstream.

Migrating from prometheus_exporter (pre-1.7.7)

If you have dashboards or alerts that reference the old metric names, update them per this table:

Category	Before (prometheus_exporter)	After (Yabeda)
HTTP requests total	ruby_http_requests_total	rails_requests_total
HTTP request duration	ruby_http_request_duration_seconds	rails_request_duration
Sidekiq jobs total	ruby_sidekiq_jobs_total	sidekiq_jobs_executed_total
Sidekiq failed jobs	ruby_sidekiq_failed_jobs_total	sidekiq_jobs_failed_total
Sidekiq job duration	ruby_sidekiq_job_duration_seconds	sidekiq_job_runtime_seconds
Sidekiq queue latency	ruby_sidekiq_queue_latency_seconds	sidekiq_queue_latency
Sidekiq queue backlog	ruby_sidekiq_queue_backlog_total	sidekiq_jobs_waiting_count
Sidekiq process count	ruby_sidekiq_process_count	sidekiq_active_processes
Puma workers	ruby_puma_workers	puma_workers
Puma backlog	ruby_puma_request_backlog	puma_backlog
Puma thread-pool capacity	ruby_puma_thread_pool_capacity	puma_pool_capacity
ActiveRecord pool size	active_record_connection_pool_connections	activerecord_connection_pool_size
Process / GC	various ruby_* series	not emitted by default; register a custom Yabeda group

For deployment-side changes (env-var renames, removed sidecar), see the 1.7.7 Updating guide entry.