Sourcegulldocs
/

Dashboards

Sourcegull's Grafana dashboards, one per service. They are generated from crates/monitoring into deploy/docker/grafana/dashboards/ (in the Compose bundle, to import into your Grafana) and the Helm chart's dashboard ConfigMaps (metrics.grafanaDashboards.enabled=true); README.md shows how to set them up. Panels with alerts link to alerts.md.

Sourcegull: web

The web role serves the UI, the GraphQL and HTTP APIs, streaming search, code host webhooks and the executor API.

Grafana uid: sourcegull-web.

web: HTTP requests

web_http_request_rate

Requests per second by route

Requests to the public listener by matched route pattern (fallback is the UI and raw files). Health checks and scrapes are left out.

promql
sum by (route) (rate(sourcegull_http_requests_total{role=~"web|all",route!~"/metrics|/healthz|/readyz"}[5m]))

Owner: platform

web_http_5xx_ratio

Share of requests answered with 5xx

promql
(sum(rate(sourcegull_http_requests_total{role=~"web|all",route!~"/metrics|/healthz|/readyz",status=~"5.."}[5m])) or 0 * sum(rate(sourcegull_http_requests_total{role=~"web|all",route!~"/metrics|/healthz|/readyz"}[5m])))
/ sum(rate(sourcegull_http_requests_total{role=~"web|all",route!~"/metrics|/healthz|/readyz"}[5m]))

Owner: platform

web_http_p90_latency

90th percentile request duration by route

promql
histogram_quantile(0.9, sum by (le, route) (rate(sourcegull_http_request_duration_seconds_bucket{role=~"web|all",route!~"/metrics|/healthz|/readyz|/.api/search/stream|/.api/compute/stream|/.api/blame/.*|/.api/scip/upload|/.api/files/.*|/.executors/.*"}[5m])))

Owner: platform

web: Search and GraphQL

web_search_stream_p90_duration

90th percentile streaming search duration

Time from the request until the search stream ends, as users see it. Histogram buckets end at 60 s.

promql
histogram_quantile(0.9, sum by (le) (rate(sourcegull_http_request_duration_seconds_bucket{role=~"web|all",route="/.api/search/stream"}[5m])))

Owner: search

web_search_stream_requests

Streaming searches per second by status

Search requests from the UI, the API and gull. Errors inside a stream are sent as events and still count as 200.

promql
sum by (status) (rate(sourcegull_http_requests_total{role=~"web|all",route="/.api/search/stream"}[5m]))

Owner: search

web_graphql_p90_duration

90th percentile GraphQL request duration

promql
histogram_quantile(0.9, sum by (le) (rate(sourcegull_http_request_duration_seconds_bucket{role=~"web|all",route="/.api/graphql"}[5m])))

Owner: platform

web: Code host webhooks

web_webhook_requests

Code host webhook deliveries per second by status

GitLab webhook deliveries (repository pushes, merge requests, pipelines) by response status.

promql
sum by (route, status) (rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.api/gitlab-webhooks|/.api/webhooks/.*"}[5m]))

Owner: source

web_webhook_rejection_ratio

Share of code host webhooks rejected

promql
(sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.api/gitlab-webhooks|/.api/webhooks/.*",status!~"2.."}[30m])) or 0 * sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.api/gitlab-webhooks|/.api/webhooks/.*"}[30m])))
/ sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.api/gitlab-webhooks|/.api/webhooks/.*"}[30m]))

Owner: source

web: Instances

web_instance_up

Instances answering Prometheus scrapes

1 while Prometheus can scrape the instance's internal listener (:6060), 0 when it cannot.

promql
up{role=~"web|all"}

Owner: platform

web_site_config_version

Site configuration version loaded by each instance

The site config version each process runs with. With the config in the database, all instances converge on the same version within seconds of a change; a lasting difference means an instance cannot reload it (its logs say why). With SOURCEGULL_SITE_CONFIG_FILE, versions count reloads per process and may differ.

promql
max by (role, instance) (sourcegull_site_config_version)

Owner: platform

Sourcegull: repo

Repo nodes clone and fetch repositories from the code hosts, build the text search index and serve git data and searches to the other roles over gRPC.

Grafana uid: sourcegull-repo.

repo: Disk

repo_disk_free_ratio

Free disk space on the data volume

Measured by the repo node janitor (2 minutes after start, then hourly). The node pauses clones when free space is below both SOURCEGULL_REPO_PAUSE_CLONES_FREE_PERCENT (default 10%) of the volume and SOURCEGULL_REPO_PAUSE_CLONES_FREE_GIB (default 20 GiB); Site admin > Repo nodes shows the threshold.

promql
min by (instance) (sourcegull_repo_disk_free_bytes{role=~"repo|all"} / sourcegull_repo_disk_total_bytes{role=~"repo|all"})

Owner: source

repo_disk_free_bytes

Free bytes on the data volume

Free space for clones, the search index and caches, measured hourly. A repo node needs about 1.5 times the size of its repositories.

promql
min by (instance) (sourcegull_repo_disk_free_bytes{role=~"repo|all"})

Owner: source

repo: Clones and fetches

repo_repo_update_rate

Clones and fetches per second by kind and outcome

Repository updates finished by the repo nodes: clone (first copy), fetch (scheduled or webhook-triggered update) and reclone.

promql
sum by (kind, outcome) (rate(sourcegull_repo_updates_total{role=~"repo|all"}[5m]))

Owner: source

repo_repo_update_error_ratio

Share of clones and fetches failing

promql
(sum by (kind) (rate(sourcegull_repo_updates_total{role=~"repo|all",outcome="error"}[30m])) or 0 * sum by (kind) (rate(sourcegull_repo_updates_total{role=~"repo|all"}[30m])))
/ sum by (kind) (rate(sourcegull_repo_updates_total{role=~"repo|all"}[30m]))

Owner: source

repo_repo_update_p90_duration

90th percentile clone and fetch duration by kind

Histogram buckets end at 60 s, so long clones show as 60 s.

promql
histogram_quantile(0.9, sum by (le, kind) (rate(sourcegull_repo_update_duration_seconds_bucket{role=~"repo|all"}[15m])))

Owner: source

repo_update_queue_oldest

Age of the oldest queued clone or fetch

promql
max(sourcegull_queue_oldest_queued_seconds{queue="repo_update_jobs"})

Owner: source

repo_update_queue_depth

Queued and running clones and fetches

Jobs in the repo_update_jobs queue. Scheduled fetches and webhook pushes add jobs; one queued job per repository at most.

promql
max by (state) (sourcegull_queue_jobs{queue="repo_update_jobs",state=~"queued|processing"})

Owner: source

repo: Repositories

repo_repos_by_clone_status

Repositories by clone status

Repositories known to Sourcegull (not deleted). not_cloned falls as the repo nodes clone new repositories.

promql
max by (clone_status) (sourcegull_repos)

Owner: source

repo_repos_fetch_errors

Repositories whose last update failed

Repositories with a last error recorded by their repo node; Site admin > Repositories lists them (filter: Failed fetch).

promql
max(sourcegull_repos_fetch_errors)

Owner: source

repo_repos_corrupted

Repositories marked corrupt

promql
max(sourcegull_repos_corrupted)

Owner: source

repo: Search index

repo_index_queue_length

Repositories waiting to be indexed

promql
max by (instance) (sourcegull_index_queue_length{role=~"repo|all"})

Owner: search

repo_index_update_rate

Index updates per second by outcome

built: new shards written; unchanged: nothing to do; removed: index dropped; error: the build failed.

promql
sum by (outcome) (rate(sourcegull_index_updates_total{role=~"repo|all"}[5m]))

Owner: search

repo_index_error_ratio

Share of index builds failing

promql
(sum(rate(sourcegull_index_updates_total{role=~"repo|all",outcome="error"}[30m])) or 0 * sum(rate(sourcegull_index_updates_total{role=~"repo|all"}[30m])))
/ sum(rate(sourcegull_index_updates_total{role=~"repo|all"}[30m]))

Owner: search

repo_index_build_p90_duration

90th percentile index build duration

Time to build one repository's shards. Histogram buckets end at 60 s.

promql
histogram_quantile(0.9, sum by (le) (rate(sourcegull_index_build_duration_seconds_bucket{role=~"repo|all"}[30m])))

Owner: search

repo_index_repos

Repositories by index status

Repositories with a text index (indexed) and without one yet (not_indexed).

promql
max by (status) (sourcegull_index_repos)

Owner: search

repo_index_repos_errored

Repositories whose last index build failed

Cleared by the next successful build. Site admin > Repositories shows the error per repository.

promql
max(sourcegull_index_repos_errored)

Owner: search

repo: Rebalancing

repo_repo_moves

Repository moves between repo nodes by state

After repo nodes are added or removed, the repositories whose node changes are copied to their new node (queued, copying), switched over (flipped, the old copy is kept for a few minutes) and then removed from the old node. Nothing moves until every repo node runs with the same node list.

promql
max by (state) (sourcegull_repo_moves)

Owner: source

repo_repo_move_rate

Repository moves finished per second by outcome

flipped: copied and switched over; reassigned: the old node was gone, cloned from the code host instead; retry: an attempt failed and is retried; failed: gave up; cancelled: no longer needed.

promql
sum by (outcome) (rate(sourcegull_repo_moves_total{role=~"repo|all"}[5m]))

Owner: source

repo_repo_move_throughput

Bytes copied between repo nodes per second

Pack data received by the repositories' new nodes.

promql
sum by (instance) (rate(sourcegull_repo_move_bytes_total{role=~"repo|all"}[5m]))

Owner: source

repo_repo_moves_failed

Repository moves that gave up

promql
max(sourcegull_repo_moves{state="failed"})

Owner: source

repo: Instances

repo_instance_up

Instances answering Prometheus scrapes

1 while Prometheus can scrape the instance's internal listener (:6060), 0 when it cannot.

promql
up{role=~"repo|all"}

Owner: platform

Sourcegull: worker

Workers run the job queues (code host and permission syncs, SCIP uploads, bulk changes, search jobs, watches, trends, outbound webhooks) and the singleton background jobs.

Grafana uid: sourcegull-worker.

worker: Job queues

worker_queue_depth

Queued jobs by queue

Jobs waiting for a worker, per queue table. Site admin > Background jobs shows the same counts with the newest failures.

promql
max by (queue) (sourcegull_queue_jobs{state="queued",queue!~"codeintel_index_jobs|batch_spec_workspace_execution_jobs|repo_update_jobs"})

Owner: platform

worker_queue_processing

Running jobs by queue

Jobs a worker holds right now. Stalled jobs (no heartbeat) are returned to the queue by the resetters.

promql
max by (queue) (sourcegull_queue_jobs{state="processing",queue!~"codeintel_index_jobs|batch_spec_workspace_execution_jobs|repo_update_jobs"})

Owner: platform

worker_queue_oldest_queued

Age of the oldest queued job by queue

promql
max by (queue) (sourcegull_queue_oldest_queued_seconds{queue!~"codeintel_index_jobs|batch_spec_workspace_execution_jobs|repo_update_jobs"})

Owner: platform

worker_queue_failed

Failed and errored jobs kept by queue

errored jobs are retried, failed ones are not. Growth means a job type keeps failing; Site admin > Background jobs lists the messages and can retry them.

promql
max by (queue) (sourcegull_queue_jobs{state=~"failed|errored",queue!~"codeintel_index_jobs|batch_spec_workspace_execution_jobs|repo_update_jobs"})

Owner: platform

worker: Code intelligence

worker_upload_processing_rate

SCIP uploads processed per second by outcome

completed, errored (retried) and failed (given up) upload processing runs.

promql
sum by (outcome) (rate(sourcegull_codeintel_upload_processing_seconds_count[5m]))

Owner: code-intel

worker_upload_processing_failure_ratio

Share of SCIP upload processing runs failing

promql
(sum(rate(sourcegull_codeintel_upload_processing_seconds_count{outcome!="completed"}[1h])) or 0 * sum(rate(sourcegull_codeintel_upload_processing_seconds_count[1h])))
/ sum(rate(sourcegull_codeintel_upload_processing_seconds_count[1h]))

Owner: code-intel

worker_upload_processing_p90_duration

90th percentile SCIP upload processing duration

Histogram buckets end at 60 s, so larger uploads show as 60 s.

promql
histogram_quantile(0.9, sum by (le) (rate(sourcegull_codeintel_upload_processing_seconds_bucket[30m])))

Owner: code-intel

worker_upload_processed_bytes

SCIP upload bytes processed per second

Size of the SCIP indexes turned into code graph data.

promql
sum(rate(sourcegull_codeintel_upload_processed_bytes_total[5m]))

Owner: code-intel

worker_autoindex_commits_considered

Commits considered for auto-indexing per second

Commits the auto-indexing scheduler looked at, on ref updates (event) and in its sweep.

promql
sum by (trigger) (rate(sourcegull_autoindex_commits_considered_total[5m]))

Owner: code-intel

worker_autoindex_jobs_queued

Auto-indexing jobs queued per second

Index jobs added to the codeintel_index_jobs executor queue (see the executor dashboard).

promql
sum(rate(sourcegull_autoindex_jobs_queued_total[5m]))

Owner: code-intel

worker_autoindex_rate_limited

Auto-indexing passes skipped by the rate limit per second

Repositories skipped because they were indexed too recently.

promql
sum(rate(sourcegull_autoindex_rate_limited_total[5m]))

Owner: code-intel

worker_codeintel_coverage_repos

Repositories by precise index coverage

up_to_date: a completed index serves the default branch; stale: the newest index is not at the default branch tip and older than 7 days; failing: the latest index job or upload failed; none: an indexing policy applies but nothing is indexed; not_targeted: no indexing policy applies. Site admin > Index coverage lists the repositories. Refreshed every 5 minutes.

promql
max by (state) (sourcegull_codeintel_coverage_repos)

Owner: code-intel

worker_codeintel_failing_repos_growth

Growth in repositories whose precise indexing fails, over 6 hours

promql
max(sourcegull_codeintel_coverage_repos{state="failing"}) - max(sourcegull_codeintel_coverage_repos{state="failing"} offset 6h)

Owner: code-intel

worker_codeintel_failing_repos_by_source

Failing repositories by where the latest failure happened

index_job: an auto-indexing job failed on an executor; upload: processing an uploaded index failed.

promql
max by (source) (sourcegull_codeintel_coverage_failing_repos)

Owner: code-intel

worker: Code intelligence data retention (collapsed)

worker_codeintel_uploads_deleted

SCIP uploads deleted per hour by reason

Uploads removed by the janitor: expired by data retention policies, abandoned (never finished uploading) and repository deleted.

promql
sum by (reason) (increase(sourcegull_codeintel_janitor_uploads_deleted_total[1h]))

Owner: code-intel

worker_codeintel_orphaned_blobs_deleted

Orphaned upload files deleted per hour

Blob store files without an upload, removed by the janitor.

promql
sum(increase(sourcegull_codeintel_janitor_orphaned_blobs_deleted_total[1h]))

Owner: code-intel

worker_codeintel_retention_expired

Uploads expired by data retention policies per hour

Uploads the expirer marked expired; the janitor deletes them later. Uploads on commits the repo node does not know are skipped until it does.

promql
sum(increase(sourcegull_codeintel_retention_uploads_expired_total[1h]))

Owner: code-intel

worker_codeintel_ranking_duration

Duration of the last ranking run

How long the last PageRank run over the code graph took.

promql
max(sourcegull_codeintel_ranking_duration_seconds)

Owner: code-intel

worker: Bulk changes

worker_reconciler_outcomes

Changesets reconciled per second by outcome

ok: the changeset matches its spec on the code host; scheduled: waiting for the rollout window; retry/failed: an error.

promql
sum by (outcome) (rate(sourcegull_batches_changesets_reconciled_total[5m]))

Owner: batches

worker_reconciler_failure_ratio

Share of changeset reconciliations failing

promql
(sum(rate(sourcegull_batches_changesets_reconciled_total{outcome="failed"}[30m])) or 0 * sum(rate(sourcegull_batches_changesets_reconciled_total[30m])))
/ sum(rate(sourcegull_batches_changesets_reconciled_total[30m]))

Owner: batches

worker_changeset_sync_error_ratio

Share of changeset syncs failing

promql
(sum(rate(sourcegull_batches_changeset_syncs_total{outcome="error"}[1h])) or 0 * sum(rate(sourcegull_batches_changeset_syncs_total[1h])))
/ sum(rate(sourcegull_batches_changeset_syncs_total[1h]))

Owner: batches

worker_bulk_operations

Bulk operations per second by type and outcome

Bulk actions on changesets: comment, detach, reenqueue, merge, close, publish.

promql
sum by (type, outcome) (rate(sourcegull_batches_bulk_jobs_total[5m]))

Owner: batches

worker_code_monitor_triggers

Watch searches queued per second

Trigger searches queued on pushes and by the evaluator's sweep.

promql
sum by (trigger) (rate(sourcegull_code_monitor_triggers_total[5m]))

Owner: code-insights

worker_code_monitor_delivery_error_ratio

Share of watch notifications failing by kind

promql
(sum by (kind) (rate(sourcegull_code_monitor_deliveries_total{outcome="error"}[1h])) or 0 * sum by (kind) (rate(sourcegull_code_monitor_deliveries_total[1h])))
/ sum by (kind) (rate(sourcegull_code_monitor_deliveries_total[1h]))

Owner: code-insights

worker_insights_search_calls

Trend searches per second by job

Searches run by trend backfills and recordings; limited by insights.query.worker.rateLimit.

promql
sum by (job) (rate(sourcegull_insights_count_calls_total[5m]))

Owner: code-insights

worker_outbound_webhook_error_ratio

Share of outbound webhook deliveries failing

promql
(sum(rate(sourcegull_outbound_webhook_deliveries_total{outcome="error"}[1h])) or 0 * sum(rate(sourcegull_outbound_webhook_deliveries_total[1h])))
/ sum(rate(sourcegull_outbound_webhook_deliveries_total[1h]))

Owner: platform

worker: Out-of-band migrations

worker_oob_migration_progress

Out-of-band migration progress

Progress of background data migrations, e.g. re-encryption after a key rotation. Site admin > Migrations shows the same.

promql
max by (id) (sourcegull_oob_migration_progress)

Owner: platform

worker_oob_migration_errors

Out-of-band migration errors in the last hour

promql
sum by (id) (increase(sourcegull_oob_migration_errors_total[1h]))

Owner: platform

worker: Housekeeping (collapsed)

worker_sessions_expired_deleted

Expired sessions deleted per hour

Removed by the session janitor every 10 minutes.

promql
sum(increase(sourcegull_sessions_expired_deleted_total[1h]))

Owner: platform

worker_changesets_purged

Changesets deleted by retention per hour

Detached changesets older than batchChanges.changesetsRetention.

promql
sum(increase(sourcegull_batches_changesets_purged_total[1h]))

Owner: batches

worker_changesets_released

Changesets released by the rollout window scheduler per hour

Changesets published as batchChanges.rolloutWindows allows.

promql
sum(increase(sourcegull_batches_changesets_released_total[1h]))

Owner: batches

worker_outbound_webhook_logs_deleted

Outbound webhook logs deleted per hour

Delivery logs past their retention, removed by the janitor.

promql
sum(increase(sourcegull_outbound_webhook_logs_deleted_total[1h]))

Owner: platform

worker: Instances

worker_instance_up

Instances answering Prometheus scrapes

1 while Prometheus can scrape the instance's internal listener (:6060), 0 when it cannot.

promql
up{role=~"worker|all"}

Owner: platform

Sourcegull: executor

Executors take auto-indexing jobs (codeintel_index_jobs) and server-side bulk change spec executions (batch_spec_workspace_execution_jobs) from the web role's executor API and run them in containers or Kubernetes jobs. Executor processes expose no metrics; this dashboard shows their queues, heartbeats and API traffic.

Grafana uid: sourcegull-executor.

executor: Executors

executor_executors_active

Executors sending heartbeats

Executors that sent a heartbeat in the last minute (they send one every 5 s). Site admin > Executors lists them.

promql
max(sourcegull_executors_active)

Owner: platform

executor_running_jobs

Jobs running on executors

Jobs the active executors report running, across both queues.

promql
max(sourcegull_executor_running_jobs)

Owner: platform

executor_jobs_without_executors

Queued executor jobs while no executor is active

The number of queued executor jobs when no executor has sent a heartbeat for a minute, else 0.

promql
sum(max by (queue) (sourcegull_queue_jobs{queue=~"codeintel_index_jobs|batch_spec_workspace_execution_jobs",state="queued"}))
* on() (max(sourcegull_executors_active) == bool 0)

Owner: platform

executor: Executor queues

executor_queue_depth

Queued and running executor jobs by queue

Jobs waiting for and held by executors.

promql
max by (queue, state) (sourcegull_queue_jobs{queue=~"codeintel_index_jobs|batch_spec_workspace_execution_jobs",state=~"queued|processing"})

Owner: platform

executor_queue_oldest_queued

Age of the oldest queued executor job by queue

promql
max by (queue) (sourcegull_queue_oldest_queued_seconds{queue=~"codeintel_index_jobs|batch_spec_workspace_execution_jobs"})

Owner: platform

executor_queue_failed

Failed and errored executor jobs kept by queue

Failed auto-indexing jobs are shown per repository under Code graph > Index jobs; failed bulk change spec workspaces on the bulk change spec's execution page.

promql
max by (queue, state) (sourcegull_queue_jobs{queue=~"codeintel_index_jobs|batch_spec_workspace_execution_jobs",state=~"failed|errored"})

Owner: platform

executor: Executor API

executor_api_requests

Executor API requests per second by route

Heartbeats, dequeues, log updates, git and file downloads from executors.

promql
sum by (route) (rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.executors/.*"}[5m]))

Owner: platform

executor_api_5xx_ratio

Share of executor API requests answered with 5xx

promql
(sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.executors/.*",status=~"5.."}[15m])) or 0 * sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.executors/.*"}[15m])))
/ sum(rate(sourcegull_http_requests_total{role=~"web|all",route=~"/.executors/.*"}[15m]))

Owner: platform

Sourcegull: blob

The blob role stores SCIP uploads, bulk change workspace files and search job results on its volume and serves them to the other roles. With an S3-compatible store (blobstore.mode=s3) there is no blob role and this dashboard stays empty.

Grafana uid: sourcegull-blob.

blob: Requests

blob_request_rate

Blob requests per second by method and status

PUT writes, GET/HEAD reads, DELETE removals. 404 on reads is normal for missing keys.

promql
sum by (method, status) (rate(sourcegull_blob_requests_total{role=~"blob|all"}[5m]))

Owner: platform

blob_error_ratio

Share of blob requests answered with 5xx

promql
(sum(rate(sourcegull_blob_requests_total{role=~"blob|all",status=~"5.."}[10m])) or 0 * sum(rate(sourcegull_blob_requests_total{role=~"blob|all"}[10m])))
/ sum(rate(sourcegull_blob_requests_total{role=~"blob|all"}[10m]))

Owner: platform

blob_p90_latency

90th percentile blob request duration by method

Until the response starts: reads stream their body afterwards, writes include receiving the upload.

promql
histogram_quantile(0.9, sum by (le, method) (rate(sourcegull_blob_request_duration_seconds_bucket{role=~"blob|all"}[5m])))

Owner: platform

blob: Instances

blob_instance_up

Instances answering Prometheus scrapes

1 while Prometheus can scrape the instance's internal listener (:6060), 0 when it cannot.

promql
up{role=~"blob|all"}

Owner: platform

Sourcegull: postgres

Sourcegull's database, measured by the workers (pg_stat_activity, database size) and by every process (its connection pool). For server-level metrics (replication, I/O, vacuum), run a Postgres exporter next to it.

Grafana uid: sourcegull-postgres.

postgres: Database

postgres_postgres_up

Workers reaching Postgres

1 when the last metrics collection by a worker (every 30 s) reached Postgres, else 0.

promql
max(sourcegull_postgres_up)

Owner: platform

postgres_connections

Connections to Sourcegull's database by state

From pg_stat_activity. Many idle in transaction connections point to transactions left open by a client.

promql
max by (state) (sourcegull_postgres_connections)

Owner: platform

postgres_connection_saturation

Share of max_connections in use by Sourcegull's database

promql
sum(max by (state) (sourcegull_postgres_connections))
/ max(sourcegull_postgres_max_connections)

Owner: platform

postgres_oldest_transaction

Age of the oldest open transaction

promql
max(sourcegull_postgres_oldest_transaction_seconds)

Owner: platform

postgres_database_size

Size of Sourcegull's database

Grows with repositories, code graph data and job history. Watch the free space of the Postgres volume.

promql
max(sourcegull_postgres_database_size_bytes)

Owner: platform

postgres: Connection pools

postgres_pool_in_use

Pool connections in use by process

Connections each process has checked out of its pool. Workers next to other roles get a separate worker pool.

promql
max by (role, instance, pool) (sourcegull_db_pool_connections{state="in_use"})

Owner: platform

postgres_pool_saturation

Share of the connection pool in use by process

promql
max by (role, instance, pool) (sourcegull_db_pool_connections{state="in_use"} / ignoring(state) sourcegull_db_pool_connections{state="max"})

Owner: platform

Edit this page on GitHub