feat(alerts): raise one alert per open issue and pull request

Each open issue now has its own alert. An issue with recent activity is issue_fresh. An issue idle for issue_stale_after or longer is issue_stale. This replaces the single issues_stale line per repo.

A non-draft pull request younger than pr_warn_after now raises pr_fresh. Older pull requests keep pr_waiting.

All new alerts have warn severity, so a repo with open work is no longer green. The pr_open alert for auto actions is unchanged.

README.md

@@ -17,11 +17,11 @@ Release archives and the image checksums are signed with cosign.
 The dashboard lists only what needs attention, grouped by project. The project with the newest problem comes first:
 
 - Failing workflows, including scheduled and other background ones, with a link to the run.
-- Pull requests that wait too long or fail their checks, with links and age.
+- Pull requests, one line each: new ones (`pr_fresh`), ones that wait too long (`pr_waiting`) and ones that fail their checks.
-- Stale issues, as one line per project.
+- Open issues, one line each: new ones (`issue_fresh`) and ones without recent activity (`issue_stale`).
 - Projects that failed to refresh, or loaded only in part.
 
-A side panel lists every project with its latest release (or newest tag), a link to the release, and counts of workflows, open pull requests and issues. Healthy items do not appear anywhere else.
+A side panel lists every project with its latest release (or newest tag), a link to the release, and counts of workflows, open pull requests and issues. Every open non-draft pull request and every open issue raises a warning, so a project with open work is never green.
 
 ### Search and filters
 
@@ -48,7 +48,9 @@ A side panel lists every project with its latest release (or newest tag), a link
 | Non-draft PR open longer than `pr_warn_after`                                            | warn     |
 | PR checks failing                                                                        | warn     |
 | Workflow queued or running longer than `workflow_stuck_after`                            | warn     |
-| Open issues without activity for `issue_stale_after` (one alert per repo)                | warn     |
+| Non-draft PR open less than `pr_warn_after` (`pr_fresh`)                                 | warn     |
+| Open issue without activity for `issue_stale_after` (`issue_stale`, one alert per issue) | warn     |
+| Open issue with recent activity (`issue_fresh`, one alert per issue)                     | warn     |
 | A section failed to load, for example a missing App permission                           | warn     |
 | Data older than `data_stale_after`                                                       | warn     |
 
@@ -121,17 +123,17 @@ actions:
 
 All metrics use the `ghtracker_` prefix. Gauges are computed at scrape time.
 
-| Metric                                                                                                         | Labels                     |
+| Metric                                                                                                                        | Labels                     |
-| -------------------------------------------------------------------------------------------------------------- | -------------------------- |
+| ----------------------------------------------------------------------------------------------------------------------------- | -------------------------- |
-| `repo_up`, `repo_health` (0 ok, 1 warn, 2 crit)                                                                | repo                       |
+| `repo_up`, `repo_health` (0 ok, 1 warn, 2 crit)                                                                               | repo                       |
-| `repo_open_issues`, `repo_open_prs`, `repo_open_prs_failing_checks`, `repo_stale_issues`                       | repo                       |
+| `repo_open_issues`, `repo_open_prs`, `repo_open_prs_failing_checks`, `repo_stale_issues`                                      | repo                       |
-| `repo_oldest_pr_age_seconds`, `repo_last_refresh_timestamp_seconds`                                            | repo                       |
+| `repo_oldest_pr_age_seconds`, `repo_last_refresh_timestamp_seconds`                                                           | repo                       |
-| `pr_age_seconds`, `pr_checks_failing`                                                                          | repo, number               |
+| `pr_age_seconds`, `pr_checks_failing`                                                                                         | repo, number               |
-| `workflow_failing`, `workflow_running`, `workflow_last_run_timestamp_seconds`                                  | repo, workflow, background |
+| `workflow_failing`, `workflow_running`, `workflow_last_run_timestamp_seconds`                                                 | repo, workflow, background |
-| `release_info`, `release_published_timestamp_seconds`                                                          | repo, version, url         |
+| `release_info`, `release_published_timestamp_seconds`                                                                         | repo, version, url         |
-| `alerts`                                                                                                       | severity                   |
+| `alerts`                                                                                                                      | severity                   |
-| `refreshes_total`, `repo_refresh_errors_total`, `last_refresh_duration_seconds`, `github_rate_limit_remaining` | none                       |
+| `refreshes_total`, `repo_refresh_errors_total`, `last_refresh_duration_seconds`, `github_rate_limit_remaining`                | none                       |
-| `action_requests_total` (counter; result is `ok`, `rejected` or `error`, plus `auto_*` for auto actions; only with `actions`)                  | action, result             |
+| `action_requests_total` (counter; result is `ok`, `rejected` or `error`, plus `auto_*` for auto actions; only with `actions`) | action, result             |
 
 Example alert rules:
 

internal/collector/collector_test.go

@@ -67,7 +67,7 @@ func TestEvaluate(t *testing.T) {
 		}, map[string]model.Severity{KindWorkflowStuck: model.Warn}, model.Warn},
 		"pr fresh": {func(r *model.Repo) {
 			r.PRs = []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0.Add(-47 * time.Hour)}}
-		}, map[string]model.Severity{}, model.OK},
+		}, map[string]model.Severity{KindPRFresh: model.Warn}, model.Warn},
 		"pr waiting warn": {func(r *model.Repo) {
 			r.PRs = []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0.Add(-49 * time.Hour)}}
 		}, map[string]model.Severity{KindPRWaiting: model.Warn}, model.Warn},
@@ -79,11 +79,13 @@ func TestEvaluate(t *testing.T) {
 		}, map[string]model.Severity{}, model.OK},
 		"pr checks failing": {func(r *model.Repo) {
 			r.PRs = []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0, Checks: model.CheckSummary{State: model.ChecksFail, Failing: []string{"build"}}}}
-		}, map[string]model.Severity{KindPRChecks: model.Warn}, model.Warn},
+		}, map[string]model.Severity{KindPRFresh: model.Warn, KindPRChecks: model.Warn}, model.Warn},
-		"issues stale aggregated": {func(r *model.Repo) {
+		"issue stale": {func(r *model.Repo) {
-			old := t0.Add(-40 * 24 * time.Hour)
+			r.Issues = []model.Issue{{Number: 1, CreatedAt: t0.Add(-60 * 24 * time.Hour), UpdatedAt: t0.Add(-40 * 24 * time.Hour)}}
-			r.Issues = []model.Issue{{Number: 1, UpdatedAt: old}, {Number: 2, UpdatedAt: old}, {Number: 3, UpdatedAt: t0}}
+		}, map[string]model.Severity{KindIssueStale: model.Warn}, model.Warn},
-		}, map[string]model.Severity{KindIssuesStale: model.Warn}, model.Warn},
+		"issue fresh": {func(r *model.Repo) {
+			r.Issues = []model.Issue{{Number: 1, CreatedAt: t0.Add(-time.Hour), UpdatedAt: t0}}
+		}, map[string]model.Severity{KindIssueFresh: model.Warn}, model.Warn},
 		"archived repo skips work alerts": {func(r *model.Repo) {
 			r.Archived = true
 			r.PRs = []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0.Add(-30 * 24 * time.Hour)}}
@@ -125,8 +127,10 @@ func TestEvaluateDisabledThresholds(t *testing.T) {
 	r := model.Repo{FullName: "o/r", RefreshedAt: t0, PRs: []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0.Add(-100 * 24 * time.Hour)}},
 		Issues: []model.Issue{{UpdatedAt: t0.Add(-100 * 24 * time.Hour)}}}
 	off := config.Thresholds{PRWarnAfter: -1, PRCritAfter: -1, IssueStaleAfter: -1, WorkflowStuck: -1, DataStaleAfter: -1}
-	if a := withoutPROpen(Evaluate(&r, t0, off)); len(a) != 0 {
+	for _, a := range withoutPROpen(Evaluate(&r, t0, off)) {
-		t.Fatalf("disabled checks produced %+v", a)
+		if a.Kind != KindPRFresh && a.Kind != KindIssueFresh {
+			t.Fatalf("disabled checks produced %+v", a)
+		}
 	}
 }
 
@@ -134,11 +138,11 @@ func TestAgeAdvancesWithoutRefresh(t *testing.T) {
 	r := model.Repo{FullName: "o/r", RefreshedAt: t0, PRs: []model.PullRequest{{Number: 1, Fork: true, CreatedAt: t0.Add(-40 * time.Hour)}}}
 	cfg := th()
 	cfg.DataStaleAfter = -1
-	if a := withoutPROpen(Evaluate(&r, t0, cfg)); len(a) != 0 {
+	if a := withoutPROpen(Evaluate(&r, t0, cfg)); len(a) != 1 || a[0].Kind != KindPRFresh {
-		t.Fatalf("unexpected %+v", a)
+		t.Fatalf("a PR below the warn threshold must be fresh: %+v", a)
 	}
-	if a := withoutPROpen(Evaluate(&r, t0.Add(10*time.Hour), cfg)); len(a) != 1 || a[0].Severity != model.Warn {
+	if a := withoutPROpen(Evaluate(&r, t0.Add(10*time.Hour), cfg)); len(a) != 1 || a[0].Kind != KindPRWaiting {
-		t.Fatalf("PR must cross the warn threshold as time passes: %+v", a)
+		t.Fatalf("PR must turn from fresh to waiting as time passes: %+v", a)
 	}
 }
 
@@ -490,9 +494,6 @@ func TestEvaluatePROpen(t *testing.T) {
 		t.Run(name, func(t *testing.T) {
 			r := model.Repo{FullName: "o/r", Up: true, RefreshedAt: t0, Archived: tc.archived, PRs: []model.PullRequest{tc.pr}}
 			alerts := Evaluate(&r, t0, th())
-			if r.Health != model.OK {
-				t.Fatalf("health = %v, want ok", r.Health)
-			}
 			if !tc.want {
 				if len(alerts) != 0 {
 					t.Fatalf("alerts = %+v, want none", alerts)
@@ -501,21 +502,31 @@ func TestEvaluatePROpen(t *testing.T) {
 			}
 			want := model.Alert{Severity: model.OK, Repo: "o/r", Kind: KindPROpen, Subject: "#7 fix", Detail: "open for 5h",
 				URL: tc.pr.URL, Ref: "abc123", Author: "alice", Since: tc.pr.CreatedAt}
-			if len(alerts) != 1 || alerts[0] != want {
+			if alerts = onlyPROpen(alerts); len(alerts) != 1 || alerts[0] != want {
 				t.Fatalf("alerts = %+v, want [%+v]", alerts, want)
 			}
 		})
 	}
 }
 
-func TestPROpenDoesNotChangeCounts(t *testing.T) {
+func onlyPROpen(in []model.Alert) []model.Alert {
+	var out []model.Alert
+	for _, a := range in {
+		if a.Kind == KindPROpen {
+			out = append(out, a)
+		}
+	}
+	return out
+}
+
+func TestPROpenHasNoSeverityOfItsOwn(t *testing.T) {
 	f := &fakeFetcher{fn: func(name string) (model.Repo, error) {
 		return model.Repo{FullName: name, PRs: []model.PullRequest{{Number: 1, HeadSHA: "s", CreatedAt: t0}}}, nil
 	}}
 	c := newCollector(f, "o/a")
 	c.Refresh(context.Background())
 	v := c.View()
-	if len(v.Alerts) != 1 || v.Alerts[0].Kind != KindPROpen || v.Crit != 0 || v.Warn != 0 || v.Repos[0].Health != model.OK {
+	if len(onlyPROpen(v.Alerts)) != 1 || v.Alerts[0].Severity != model.Warn || v.Crit != 0 || v.Warn != 1 {
 		t.Fatalf("view = %+v", v)
 	}
 }

internal/collector/health.go

@@ -19,7 +19,9 @@ const (
 	KindWorkflowStuck = "workflow_stuck"
 	KindPRWaiting     = "pr_waiting"
 	KindPRChecks      = "pr_checks_failing"
-	KindIssuesStale   = "issues_stale"
+	KindPRFresh       = "pr_fresh"
+	KindIssueStale    = "issue_stale"
+	KindIssueFresh    = "issue_fresh"
 	// KindPROpen is informational plumbing for automatic actions; it has OK severity so it never moves health or counts.
 	KindPROpen = "pr_open"
 )
@@ -71,27 +73,20 @@ func Evaluate(r *model.Repo, now time.Time, th config.Thresholds) []model.Alert
 			add(model.Crit, KindPRWaiting, subject, "open "+Age(age), pr.URL, pr.CreatedAt)
 		case th.PRWarnAfter > 0 && age >= th.PRWarnAfter:
 			add(model.Warn, KindPRWaiting, subject, "open "+Age(age), pr.URL, pr.CreatedAt)
+		default:
+			add(model.Warn, KindPRFresh, subject, "new, open "+Age(age), pr.URL, pr.CreatedAt)
 		}
 		if pr.Checks.State == model.ChecksFail {
 			add(model.Warn, KindPRChecks, subject, summarizeChecks(pr.Checks), pr.URL, pr.UpdatedAt)
 		}
 	}
 
-	if th.IssueStaleAfter > 0 {
+	for _, is := range r.Issues {
-		var stale int
+		subject := fmt.Sprintf("#%d %s", is.Number, is.Title)
-		var oldest time.Time
+		if idle := now.Sub(is.UpdatedAt); th.IssueStaleAfter > 0 && idle >= th.IssueStaleAfter {
-		for _, is := range r.Issues {
+			add(model.Warn, KindIssueStale, subject, "no activity for "+Age(idle), is.URL, is.UpdatedAt)
-			if now.Sub(is.UpdatedAt) >= th.IssueStaleAfter {
+		} else {
-				stale++
+			add(model.Warn, KindIssueFresh, subject, "new, open "+Age(now.Sub(is.CreatedAt)), is.URL, is.CreatedAt)
-				if oldest.IsZero() || is.UpdatedAt.Before(oldest) {
-					oldest = is.UpdatedAt
-				}
-			}
-		}
-		if stale > 0 {
-			add(model.Warn, KindIssuesStale, fmt.Sprintf("%d stale issues", stale),
-				"no activity for "+Age(th.IssueStaleAfter)+" or more",
-				r.URL+"/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-asc", oldest)
 		}
 	}
 	return alerts

internal/web/auto_test.go

@@ -138,10 +138,10 @@ func TestNewAutoRunnerNilWithoutAutoActions(t *testing.T) {
 
 func TestStateCarriesPROpenRef(t *testing.T) {
 	st := buildState(autoView(autoPR(1, "abc", "x", nil)), true)
-	if len(st.Alerts) != 1 || st.Alerts[0].Kind != collector.KindPROpen || st.Alerts[0].Ref != "abc" || st.Alerts[0].Severity != "ok" {
+	if len(st.Alerts) != 2 || st.Alerts[0].Kind != collector.KindPROpen || st.Alerts[0].Ref != "abc" || st.Alerts[0].Severity != "ok" || st.Alerts[1].Kind != collector.KindPRFresh {
 		t.Fatalf("alerts = %+v", st.Alerts)
 	}
-	if st.Counts.Crit != 0 || st.Counts.Warn != 0 || st.Repos[0].Health != "ok" {
+	if st.Repos[0].Health != "warn" {
-		t.Fatalf("counts/health changed: %+v %s", st.Counts, st.Repos[0].Health)
+		t.Fatalf("health = %s, want warn", st.Repos[0].Health)
 	}
 }

src/components/components.test.ts

@@ -156,7 +156,7 @@ describe('AttentionRow', () => {
   it('omits the age for an unknown timestamp', () => {
     const w = mount(AttentionRow, {
       props: {
-        row: row({ since: '0001-01-01T00:00:00Z', kind: 'issues_stale' }),
+        row: row({ since: '0001-01-01T00:00:00Z', kind: 'issue_stale' }),
         nowMs,
         critSeconds: CRIT_SECONDS,
       },

src/lib/filters.test.ts

@@ -38,7 +38,7 @@ const fleet = () =>
         }),
         alert({
           repo: 'b/two',
-          kind: 'issues_stale',
+          kind: 'issue_stale',
           severity: 'warn',
           subject: '4 stale issues',
           url: '3',
@@ -62,7 +62,7 @@ describe('kindGroup', () => {
     ['workflow_stuck', 'workflow'],
     ['pr_waiting', 'pr'],
     ['pr_checks_failing', 'pr'],
-    ['issues_stale', 'issue'],
+    ['issue_stale', 'issue'],
     ['refresh_failed', 'data'],
     ['data_stale', 'data'],
     ['partial_data', 'data'],

src/lib/filters.ts

@@ -24,7 +24,7 @@ export const emptyFilters = (): Filters => ({
 export function kindGroup(kind: string): KindGroup {
   if (kind.startsWith('workflow_')) return 'workflow'
   if (kind.startsWith('pr_')) return 'pr'
-  if (kind === 'issues_stale') return 'issue'
+  if (kind.startsWith('issue_')) return 'issue'
   return 'data'
 }
 

src/lib/kinds.ts

@@ -14,7 +14,9 @@ export const KIND_ICON: Record<string, IconDefinition> = {
   workflow_stuck: faGears,
   pr_waiting: faCodePullRequest,
   pr_checks_failing: faCodePullRequest,
-  issues_stale: faCircleDot,
+  pr_fresh: faCodePullRequest,
+  issue_stale: faCircleDot,
+  issue_fresh: faCircleDot,
   refresh_failed: faPlugCircleXmark,
   data_stale: faClockRotateLeft,
   partial_data: faTriangleExclamation,