{{ message }}
feat(alerting): per-rule/per-host alert routing (plan 30) - #2269
Merged
Conversation
Lets an operator route health-sweep findings to different channel webhook
URLs by rule id/type and/or host (glob or *), fanning out to every matched
route and falling back to the legacy global HEALTH_ALERT_WEBHOOK_URL when
nothing matches (today's exact behavior for deployments with no routes
configured). Dedup still runs once per finding, not once per channel.
- alert_routes D1 migration (0014) + lib/health/alert-routing.ts: pure
matchRoutes/resolveTargets core plus best-effort D1 CRUD, mirroring
alert-history-store.ts's fail-open pattern.
- server-sweep.ts: loads routes once per sweep, fans out postWebhook +
per-channel audit rows over resolveTargets, commits the dedup decision
exactly once per finding.
- /api/v1/health/routes (GET/POST/DELETE): owner-scoped, fails open to the
OSS single-tenant owner ('') when Clerk isn't configured, gates writes
behind sign-in only in cloud mode (lib/health/alert-routing-auth.ts).
- Routing tab in HealthSettingsDialog (alert-routing-dialog.tsx) to
list/create/delete routes and send a test alert per route.
Co-Authored-By: duyetbot <bot@duyet.net>
Reconcile plan 30 (per-rule/per-host alert routing) with the sibling
alerting work already merged to main. In server-sweep.ts the only real
conflict was plan 30 (routing fan-out) x plan 31 (compound rules, which
had refactored the inline dispatch into a shared `dispatchFinding`
helper).
Resolution:
- Fold plan 30's route fan-out INTO main's `dispatchFinding` helper so
both base rules and compound rules route identically. The helper now
takes a `ruleType` param and, inside the `decision.notify` branch,
calls `resolveTargets(routes, {ruleId, ruleType, hostId, hostName})`
and posts to every matched channel, falling back to the legacy global
webhook when nothing matches. Per-channel audit rows use
`detectAdapter(url).id` (composes with plan 26's Opsgenie adapter).
- Dedup (`evaluateAlert`) still runs exactly ONCE per finding, above the
fan-out, so cooldown state is never multiplied; the notify decision
gates the whole fan-out (a suppressed finding reaches no channel).
- Deleted the leftover inline routing block that git could not place
(plan 30 predated the helper extraction).
- Pass `ruleType: rule.type` at the base-rule call site and
`ruleType: 'compound'` at the compound-rule call site.
- Renamed migration 0014_alert_routes.sql -> 0015_alert_routes.sql to
clear the collision with main's 0014_weekly_reports.sql (wrangler
applies migrations in filename order).
Verified: apps/dashboard build + tsc clean; `bun test src/lib/health/`
196/196 pass (fan-out single/double-route/dedup-once, no-match fallback,
compound throwing/dedup/compound-on-compound); biome lint clean.
Co-Authored-By: duyetbot <bot@duyet.net>
github-actions
Bot
added
the
app: dashboard
Contributor
This was referenced Jul 3, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Plan 30 — per-rule / per-host alert routing (reconciled onto latest main)
Implements per-rule/per-host alert routing (
plans/30-per-rule-alert-routing.md) and merges it cleanly onto the currentmain, which had already absorbed the sibling alerting work (compound rules, Opsgenie adapter, custom rules, remediation links, etc.).What routing adds
Operators can define routes (D1 table
alert_routes) that match a rule id/type and/or host (glob or*) to one or more channel webhook URLs. Each notifying finding fans out to every matched channel; when nothing matches it falls back to the legacy globalHEALTH_ALERT_WEBHOOK_URL— so deployments that never configure a route behave exactly as before. Routing is best-effort/fail-open (D1 unavailable →[]→ legacy fallback).How the merge was reconciled (
server-sweep.ts)The only real conflict in this file was plan 30 (routing fan-out) × plan 31 (compound rules). Plan 31 had refactored the inline dispatch into a shared
dispatchFindinghelper; plan 30 (which predated that extraction) added fan-out inline at the old call site.Resolution:
dispatchFindingso base rules and compound rules route identically. The helper gained aruleTypeparam and, inside thedecision.notifybranch, callsresolveTargets(routes, {ruleId, ruleType, hostId, hostName}, settings.webhookUrl)and posts to every returned channel.evaluateAlert) runs exactly once per finding, above the fan-out — so cooldown state is never multiplied, only the destination set is. Thedecision.notifygate wraps the entire fan-out, so a finding that is deduped/suppressed reaches zero channels (routed or not). Per-channel audit rows usedetectAdapter(url).id, so a fan-out across mixed Slack/Discord/Opsgenie destinations is labelled per its own adapter (composes with plan 26).commit()fires when there was nothing to deliver or at least one channel succeeded; a total-failure delivery leaves no record so the next sweep retries.ruleType: rule.type(base rules) andruleType: 'compound'(compound rules).0014_alert_routes.sql→0015_alert_routes.sqlto clear the collision with main's0014_weekly_reports.sql(wrangler applies migrations in filename order).Note on the other sibling plans
Plans 26/28/29/32/33 did not touch
server-sweep.tsonmain(verified viagit log); they live in adapters / rule-registry / UI / API and merged cleanly. The suppression concepts (maintenance windows, ACKs) do not gate this session-less cron dispatch path in the current tree, so nothing there needed reconciling here.Verification
apps/dashboardbuild +tsc --noEmit: cleanbun test src/lib/health/ --isolate: 196/196 pass — fan-out (single route replaces legacy, two routes fan out with dedup-once, no-match fallback) and compound dispatch (throwing rule, dedup under own key, compound-on-compound) both exercisedbiome lint: cleanCloses plan 30.