fix: Route shell session alerts to compute

config/components/alerts/shell-sessions.yaml

@@ -32,9 +32,9 @@ spec:
           for: 10m
           labels:
             severity: critical
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: Shell sessions in {{ $labels.location }} are ending with {{ $labels.reason }}
             description: |
@@ -59,9 +59,9 @@ spec:
           for: 5m
           labels:
             severity: warning
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: Shell agent {{ $labels.agent }} on cell {{ $labels.cell }} cannot reach its endpoint
             description: |
@@ -77,9 +77,9 @@ spec:
           for: 5m
           labels:
             severity: critical
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: No shell agent on cell {{ $labels.cell }} can reach its endpoint
             description: |
@@ -102,9 +102,9 @@ spec:
           for: 15m
           labels:
             severity: warning
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: Shell sessions in {{ $labels.location }} take {{ $value | humanizeDuration }} to connect (p95)
             description: |
@@ -125,9 +125,9 @@ spec:
             sum by (cell) (increase(compute_shell_agent_cleanup_unconfirmed_total[1h])) > 0
           labels:
             severity: warning
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: Cell {{ $labels.cell }} did not confirm cleanup for {{ $value | humanize }} shell sessions
             description: |
@@ -155,9 +155,9 @@ spec:
           for: 10m
           labels:
             severity: warning
-            service: compute-shell-sessions
+            service: compute
             namespace: compute-system
-            team: compute
+            team: build
           annotations:
             summary: Shell session creation in {{ $labels.location }} is far above its weekly baseline
             description: |

test/alerts/shell-sessions.test.yaml

@@ -0,0 +1,16 @@
+# Run with test/alerts/test-shell-sessions.sh (requires yq and promtool).
+rule_files:
+- shell-sessions.rules.yaml
+evaluation_interval: 1m
+tests:
+- name: A session ending Unavailable pages compute and its team, build, as the platform's catalog names them
+  interval: 1m
+  input_series:
+  - series: compute_shell_sessions_ended_total{location="us-east-1",reason="Unavailable"}
+    values: 0+1x30
+  promql_expr_test:
+  - expr: ALERTS{alertname="ComputeShellSessionsUnavailable",alertstate="firing"}
+    eval_time: 25m
+    exp_samples:
+    - labels: ALERTS{alertname="ComputeShellSessionsUnavailable",alertstate="firing",location="us-east-1",reason="Unavailable",severity="critical",service="compute",namespace="compute-system",team="build"}
+      value: 1

test/alerts/test-shell-sessions.sh

@@ -0,0 +1,23 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Extract the actual VMRule rather than maintaining a second copy of its expr.
+root="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+rules="$root/config/components/alerts/shell-sessions.yaml"
+work="$(mktemp -d)"
+trap 'rm -rf "$work"' EXIT
+
+# Every alert names the service and team the platform's catalog gives compute.
+# An alert whose service the catalog does not list reaches no team, and a team
+# the catalog does not list is counted against nobody.
+owners="$(yq '[.spec.groups[].rules[] | select(.alert) | .labels.service + "/" + .labels.team] | unique | join(",")' "$rules")"
+if [ "$owners" != "compute/build" ]; then
+  echo "shell-session alerts name service/team '$owners'; want 'compute/build', the catalog's row" >&2
+  exit 1
+fi
+
+yq '.spec' "$rules" > "$work/shell-sessions.rules.yaml"
+cp "$root/test/alerts/shell-sessions.test.yaml" "$work/test.yaml"
+cd "$work"
+promtool check rules shell-sessions.rules.yaml
+promtool test rules test.yaml