#!/bin/bash
# oracle: rewrites changed files to the oracle state
set -euo pipefail
cd /workspace/monitoring
mkdir -p 'alertmanager'
cat > 'alertmanager/alertmanager.yml' <<'AF_EOF'
# Shared Alertmanager config (this repo, platform and data Prometheus servers).
# Receivers belong to the teams; ask before touching someone else's.
global:
  resolve_timeout: 5m
  slack_api_url_file: /etc/alertmanager/secrets/slack-url
templates:
- /etc/alertmanager/templates/*.tmpl
route:
  receiver: slack-alerts-default
  group_by: [alertname, service]
  group_wait: 30s
  group_interval: 5m
  repeat_interval: 4h
  routes:
  # rule evaluation problems. Nobody pages on these (yet).
  - matchers:
    - alertname=~"PrometheusRuleFailures|PrometheusRuleEvaluationSlow"
    receiver: slack-prometheus-meta
  # SRE sees every page as FYI; owners get paged by their own subtree below.
  - matchers:
    - severity=~"page|critical"
    receiver: slack-sre-fyi
    continue: true
  - matchers:
    - team="economy"
    receiver: slack-economy
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-economy
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-economy
      continue: true
    - receiver: slack-economy
  - matchers:
    - team="player-platform"
    receiver: slack-player-platform
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-player-platform
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-player-platform
      continue: true
    - receiver: slack-player-platform
  - matchers:
    - team="social"
    receiver: slack-social
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-social
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-social
      continue: true
    - receiver: slack-social
  # matchmaking: still on the old match syntax
  - match:
      team: matchmaking
    receiver: slack-matchmaking
    routes:
    - match_re:
        severity: page|critical
      receiver: pagerduty-matchmaking
      continue: true
    - match:
        severity: ticket
      receiver: jira-matchmaking
      continue: true
    - receiver: slack-matchmaking
  - matchers:
    - team="live-ops"
    receiver: slack-live-ops
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-live-ops
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-live-ops
      continue: true
    - receiver: slack-live-ops
  # live-ops services from before team labels were everywhere
  - matchers:
    - service=~"asset-indexer"
    receiver: slack-live-ops
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-live-ops
      continue: true
    - receiver: slack-live-ops
  - matchers:
    - team="telemetry"
    receiver: slack-telemetry
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-telemetry
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-telemetry
      continue: true
    - receiver: slack-telemetry
  # teams below route alerts from the platform and data Prometheus servers
  - matchers:
    - team="messaging"
    receiver: slack-messaging
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-messaging
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-messaging
      continue: true
    - receiver: slack-messaging
  # platform wanted tickets in Slack as well as Jira (OPS-2251)
  - matchers:
    - team="platform"
    receiver: slack-platform
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-platform
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-platform
      continue: true
    - receiver: slack-platform
  - matchers:
    - team="database-reliability"
    receiver: slack-database-reliability
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-database-reliability
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-database-reliability
      continue: true
    # tickets and warnings per pod in Slack, so the channel shows which pod; pages group per service above
    - receiver: slack-database-reliability
      group_by: [alertname, pod, service]
  - matchers:
    - team="security-ops"
    receiver: slack-security-ops
    group_wait: 10s
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-security-ops
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-security-ops
      continue: true
    - receiver: slack-security-ops
  - matchers:
    - team="edge"
    receiver: slack-edge
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-edge
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-edge
      continue: true
    - receiver: slack-edge
  - matchers:
    - team="data-platform"
    receiver: slack-data-platform
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-data-platform
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-data-platform
      continue: true
    - receiver: slack-data-platform
  # network asked for hourly repeats while something is still firing
  - matchers:
    - team="network"
    receiver: slack-network
    repeat_interval: 1h
    routes:
    - matchers:
      - severity=~"page|critical"
      receiver: pagerduty-network
      continue: true
    - matchers:
      - severity="ticket"
      receiver: jira-network
      continue: true
    - receiver: slack-network
receivers:
- name: slack-alerts-default
  slack_configs:
  - channel: "#alerts-default"
    send_resolved: true
- name: slack-sre-fyi
  slack_configs:
  - channel: "#sre-fyi"
    send_resolved: false
- name: slack-prometheus-meta
  slack_configs:
  - channel: "#prometheus-meta"
- name: pagerduty-sre
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/sre
- name: pagerduty-data-platform
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/data-platform
- name: slack-data-platform
  slack_configs:
  - channel: "#data-platform-alerts"
    send_resolved: true
- name: jira-data-platform
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/data_platform
- name: pagerduty-database-reliability
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/database-reliability
- name: slack-database-reliability
  slack_configs:
  - channel: "#database-reliability-alerts"
    send_resolved: true
- name: jira-database-reliability
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/database_reliability
- name: pagerduty-economy
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/economy
- name: slack-economy
  slack_configs:
  - channel: "#economy-alerts"
    send_resolved: true
- name: jira-economy
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/economy
- name: pagerduty-edge
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/edge
- name: slack-edge
  slack_configs:
  - channel: "#edge-alerts"
    send_resolved: true
- name: jira-edge
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/edge
- name: pagerduty-live-ops
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/live-ops
- name: slack-live-ops
  slack_configs:
  - channel: "#live-ops-alerts"
    send_resolved: true
- name: jira-live-ops
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/live_ops
- name: pagerduty-matchmaking
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/matchmaking
- name: slack-matchmaking
  slack_configs:
  - channel: "#matchmaking-alerts"
    send_resolved: true
- name: jira-matchmaking
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/matchmaking
- name: pagerduty-messaging
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/messaging
- name: slack-messaging
  slack_configs:
  - channel: "#messaging-alerts"
    send_resolved: true
- name: jira-messaging
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/messaging
- name: pagerduty-network
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/network
- name: slack-network
  slack_configs:
  - channel: "#network-alerts"
    send_resolved: true
- name: jira-network
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/network
- name: pagerduty-platform
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/platform
- name: slack-platform
  slack_configs:
  - channel: "#platform-alerts"
    send_resolved: true
- name: jira-platform
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/platform
- name: pagerduty-player-platform
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/player-platform
- name: slack-player-platform
  slack_configs:
  - channel: "#player-platform-alerts"
    send_resolved: true
- name: jira-player-platform
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/player_platform
- name: pagerduty-security-ops
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/security-ops
- name: slack-security-ops
  slack_configs:
  - channel: "#security-ops-alerts"
    send_resolved: true
- name: jira-security-ops
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/security_ops
- name: pagerduty-social
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/social
- name: slack-social
  slack_configs:
  - channel: "#social-alerts"
    send_resolved: true
- name: jira-social
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/social
- name: pagerduty-telemetry
  pagerduty_configs:
  - routing_key_file: /etc/alertmanager/secrets/pagerduty/telemetry
- name: slack-telemetry
  slack_configs:
  - channel: "#telemetry-alerts"
    send_resolved: true
- name: jira-telemetry
  webhook_configs:
  - url: https://jira-bridge.cobaltline.example/hooks/telemetry
# nothing routes here since the rota moved to PagerDuty; ask before deleting
- name: slack-old-oncall
  slack_configs:
  - channel: "#oncall-old"
    send_resolved: true
inhibit_rules:
# a critical copy holds back the warning copy of the same alert
- source_matchers:
  - severity="critical"
  target_matchers:
  - severity="warning"
  equal: [alertname, service]
- source_matchers:
  - alertname="ProfileSvcDown"
  target_matchers:
  - alertname=~"ProfileSvcErrorBudgetBurnFast|ProfileSvcErrorBudgetBurnSlow|ProfileSvcHighErrorRatio"
  equal: [service]
AF_EOF
mkdir -p 'rules'
cat > 'rules/capacity.yml' <<'AF_EOF'
# Capacity tickets. Pages for crashes come from the platform Prometheus.
groups:
- name: capacity
  rules:
  - alert: ContainerMemoryNearLimit
    expr: |
      max by (service) (container_memory_working_set_bytes{service="telemetry-ingest",container="telemetry-ingest"} / container_spec_memory_limit_bytes{service="telemetry-ingest",container="telemetry-ingest"}) > 0.9
    for: 15m
    labels:
      severity: ticket
      team: telemetry
    annotations:
      summary: "{{ $labels.service }} keeps running close to its memory limit"
      runbook_url: https://runbooks.cobaltline.example/alerts/ContainerMemoryNearLimit
  - alert: ContainerMemoryNearLimit
    expr: |
      max by (service) (container_memory_working_set_bytes{service="asset-indexer",container="asset-indexer"} / container_spec_memory_limit_bytes{service="asset-indexer",container="asset-indexer"}) > 0.9
    for: 15m
    labels:
      severity: ticket
      team: live-ops
    annotations:
      summary: "{{ $labels.service }} keeps running close to its memory limit"
      runbook_url: https://runbooks.cobaltline.example/alerts/ContainerMemoryNearLimit
AF_EOF
mkdir -p 'rules'
cat > 'rules/legacy-economy.yml' <<'AF_EOF'
# economy alerts from before the conventions in README.md.
# Don't reformat in passing; move things out when you touch them for a reason.
groups:
# old dashboard records; the svc-overview board still reads these
- name: economy-records
  rules:
  # keep: finance reads this one for the monthly availability report
  - record: job:http_requests:rate5m
    expr: sum by (job, service) (rate(http_requests_total{service="store-api"}[5m]))
  - record: job:http_errors:rate5m
    expr: sum by (job, service) (rate(http_requests_total{service="store-api",code=~"5.."}[5m]))
- name: store-api
  rules:
  - alert: StoreApiHighErrorRatio
    expr: sum by (service) (rate(http_requests_total{service="store-api",code=~"5.."}[5m])) / sum by (service) (rate(http_requests_total{service="store-api"}[5m])) > 0.1
    for: "10m"
    labels: {severity: page, team: economy}
    annotations:
      summary: "{{ $labels.service }} is failing {{ $value | humanizePercentage }} of requests"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiHighErrorRatio
  - alert: StoreApiTrafficBelowFloor
    expr: sum by (service) (rate(http_requests_total{service="store-api"}[5m])) < 5 and on (service) sum by (service) (up{service="store-api"}) > 0
    for: "5m"
    labels: {severity: page, team: economy}
    annotations:
      summary: "{{ $labels.service }} is serving under its traffic floor with its pods up"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiTrafficBelowFloor
  # p99 across the whole service; the per-route view is on the team board
  - alert: StoreApiLatencyHigh
    expr: histogram_quantile(0.99, sum by (service, le) (rate(http_request_duration_seconds_bucket{service="store-api"}[5m]))) > 0.5
    for: "10m"
    labels: {severity: warning, team: economy}
    annotations:
      summary: "{{ $labels.service }} p99 above 500ms"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiLatencyHigh
  - alert: StoreApiLatencyHigh
    expr: histogram_quantile(0.99, sum by (service, le) (rate(http_request_duration_seconds_bucket{service="store-api"}[5m]))) > 1
    for: "10m"
    labels: {severity: critical, team: economy}
    annotations:
      summary: "{{ $labels.service }} p99 above 1s"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiLatencyHigh
  - alert: StoreApiDbConnectionsHigh
    expr: max by (service) (db_pool_connections_active{service="store-api"} / db_pool_connections_max{service="store-api"}) > 0.8
    for: "5m"
    labels: {severity: warning, team: economy}
    annotations:
      summary: "{{ $labels.service }} is using over 80% of its DB pool"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiDbConnectionsHigh
  - alert: StoreApiDbConnectionsHigh
    expr: max by (service) (db_pool_connections_active{service="store-api"} / db_pool_connections_max{service="store-api"}) > 0.95
    for: "5m"
    labels: {severity: critical, team: economy}
    annotations:
      summary: "{{ $labels.service }} DB pool nearly exhausted"
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiDbConnectionsHigh
AF_EOF
mkdir -p 'rules'
cat > 'rules/legacy-live-ops.yml' <<'AF_EOF'
# live-ops alerts from before the conventions in README.md.
# Don't reformat in passing; move things out when you touch them for a reason.
groups:
# old dashboard records; the svc-overview board still reads these
- name: live-ops-records
  rules:
  - record: job:http_requests:rate5m
    expr: sum by (job, service) (rate(http_requests_total{service="telemetry-ingest"}[5m]))
  - record: job:http_errors:rate5m
    expr: sum by (job, service) (rate(http_requests_total{service="telemetry-ingest",code=~"5.."}[5m]))
- name: telemetry-ingest
  rules:
  - alert: TelemetryIngestHighErrorRatio
    expr: sum by (service) (rate(http_requests_total{service="telemetry-ingest",code=~"5.."}[5m])) / sum by (service) (rate(http_requests_total{service="telemetry-ingest"}[5m])) > 0.02
    for: "10m"
    labels: {severity: page, team: telemetry}
    annotations:
      summary: "{{ $labels.service }} is failing {{ $value | humanizePercentage }} of requests"
      runbook_url: https://runbooks.cobaltline.example/alerts/TelemetryIngestHighErrorRatio
  - alert: TelemetryIngestLatencyHigh
    expr: histogram_quantile(0.99, sum by (service, le) (rate(http_request_duration_seconds_bucket{service="telemetry-ingest"}[5m]))) > 0.5
    for: "10m"
    labels: {severity: warning, team: telemetry}
    annotations:
      summary: "{{ $labels.service }} p99 above 500ms"
      runbook_url: https://runbooks.cobaltline.example/alerts/TelemetryIngestLatencyHigh
  - alert: TelemetryIngestLatencyHigh
    expr: histogram_quantile(0.99, sum by (service, le) (rate(http_request_duration_seconds_bucket{service="telemetry-ingest"}[5m]))) > 1
    for: "10m"
    labels: {severity: critical, team: telemetry}
    annotations:
      summary: "{{ $labels.service }} p99 above 1s"
      runbook_url: https://runbooks.cobaltline.example/alerts/TelemetryIngestLatencyHigh
  - alert: ReplicationLagHigh
    expr: max by (service) (pg_replication_lag_seconds{service="telemetry-ingest"}) > 30
    for: "10m"
    labels: {severity: page, team: telemetry}
    annotations:
      summary: "{{ $labels.service }} replica is behind the primary"
      description: Replication lag is {{ $value | humanize }}s.
      runbook_url: https://runbooks.cobaltline.example/alerts/ReplicationLagHigh
AF_EOF
mkdir -p 'rules'
cat > 'rules/service-health.yml' <<'AF_EOF'
# One Down alert per service. See README, Service health.
groups:
- name: service-health
  rules:
  - alert: StoreApiDown
    expr: |
      sum by (service) (up{service="store-api"}) == 0
      or absent(up{service="store-api"})
    for: 3m
    labels:
      severity: page
      team: economy
    annotations:
      summary: store-api has no healthy scrape targets
      runbook_url: https://runbooks.cobaltline.example/alerts/StoreApiDown
  - alert: ProfileSvcDown
    expr: |
      sum by (service) (up{service="profile-svc"}) == 0
      or absent(up{service="profile-svc"})
    for: 3m
    labels:
      severity: page
      team: player-platform
    annotations:
      summary: profile-svc has no healthy scrape targets
      runbook_url: https://runbooks.cobaltline.example/alerts/ProfileSvcDown
  - alert: TelemetryIngestDown
    expr: |
      sum by (service) (up{service="telemetry-ingest"}) == 0
      or absent(up{service="telemetry-ingest"})
    for: 3m
    labels:
      severity: page
      team: telemetry
    annotations:
      summary: telemetry-ingest has no healthy scrape targets
      runbook_url: https://runbooks.cobaltline.example/alerts/TelemetryIngestDown
  - alert: ReplayExportDown
    expr: |
      sum by (service) (up{service="replay-export"}) == 0
      or absent(up{service="replay-export"})
    for: 3m
    labels:
      severity: page
      team: telemetry
    annotations:
      summary: replay-export has no healthy scrape targets
      runbook_url: https://runbooks.cobaltline.example/alerts/ReplayExportDown
  - alert: AssetIndexerDown
    expr: |
      sum by (service) (up{service="asset-indexer"}) == 0
      or absent(up{service="asset-indexer"})
    for: 3m
    labels:
      severity: page
      team: live-ops
    annotations:
      summary: asset-indexer has no healthy scrape targets
      runbook_url: https://runbooks.cobaltline.example/alerts/AssetIndexerDown
AF_EOF
mkdir -p 'rules'
cat > 'rules/slo-profile-svc.yml' <<'AF_EOF'
groups:
- name: slo-profile-svc
  rules:
  - alert: ProfileSvcErrorBudgetBurnFast
    expr: |
      slo:sli_error:ratio_rate1h{service="profile-svc"} > (3.36 * 0.001)
      and slo:sli_error:ratio_rate5m{service="profile-svc"} > (3.36 * 0.001)
    for: 2m
    labels:
      severity: page
      team: player-platform
      slo: profile-svc-availability
    annotations:
      summary: "{{ $labels.service }} is spending its error budget fast"
      runbook_url: https://runbooks.cobaltline.example/alerts/ProfileSvcErrorBudgetBurnFast
  - alert: ProfileSvcErrorBudgetBurnSlow
    expr: |
      slo:sli_error:ratio_rate6h{service="profile-svc"} > (1.4 * 0.001)
      and slo:sli_error:ratio_rate30m{service="profile-svc"} > (1.4 * 0.001)
    for: 15m
    labels:
      severity: ticket
      team: player-platform
      slo: profile-svc-availability
    annotations:
      summary: "{{ $labels.service }} is spending its error budget steadily"
      runbook_url: https://runbooks.cobaltline.example/alerts/ProfileSvcErrorBudgetBurnSlow
AF_EOF
mkdir -p 'rules'
cat > 'rules/social-rules.yaml' <<'AF_EOF'
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: social-rules
  namespace: monitoring
  labels:
    role: alert-rules
spec:
  groups:
  - name: leaderboard-svc
    rules:
    - alert: LeaderboardSvcDown
      expr: |
        sum by (service) (up{service="leaderboard-svc"}) == 0
        or absent(up{service="leaderboard-svc"})
      for: 3m
      labels:
        severity: page
        team: social
      annotations:
        summary: leaderboard-svc has no healthy scrape targets
        runbook_url: https://runbooks.cobaltline.example/alerts/LeaderboardSvcDown
    - alert: LeaderboardSvcHighErrorRatio
      expr: |
        sum by (service) (rate(http_requests_total{service="leaderboard-svc",code=~"5.."}[5m]))
        / sum by (service) (rate(http_requests_total{service="leaderboard-svc"}[5m])) > 0.05
      for: 10m
      labels:
        severity: page
        team: social
      annotations:
        summary: "{{ $labels.service }} is failing {{ $value | humanizePercentage }} of requests"
        runbook_url: https://runbooks.cobaltline.example/alerts/LeaderboardSvcHighErrorRatio
    - alert: LeaderboardSvcErrorBudgetBurnFast
      expr: |
        slo:sli_error:ratio_rate1h{service="leaderboard-svc"} > (14.4 * 0.0005)
        and slo:sli_error:ratio_rate5m{service="leaderboard-svc"} > (14.4 * 0.0005)
      for: 2m
      labels:
        severity: page
        team: social
        slo: leaderboard-svc-availability
      annotations:
        summary: "{{ $labels.service }} is spending its error budget fast"
        runbook_url: https://runbooks.cobaltline.example/alerts/LeaderboardSvcErrorBudgetBurnFast
    - alert: LeaderboardSvcErrorBudgetBurnSlow
      expr: |
        slo:sli_error:ratio_rate6h{service="leaderboard-svc"} > (6 * 0.0005)
        and slo:sli_error:ratio_rate30m{service="leaderboard-svc"} > (6 * 0.0005)
      for: 15m
      labels:
        severity: ticket
        team: social
        slo: leaderboard-svc-availability
      annotations:
        summary: "{{ $labels.service }} is spending its error budget steadily"
        runbook_url: https://runbooks.cobaltline.example/alerts/LeaderboardSvcErrorBudgetBurnSlow
  - name: chat-gw
    rules:
    - alert: ChatGwDown
      expr: |
        sum by (service) (up{service="chat-gw"}) == 0
        or absent(up{service="chat-gw"})
      for: 3m
      labels:
        severity: page
        team: social
      annotations:
        summary: chat-gw has no healthy scrape targets
        runbook_url: https://runbooks.cobaltline.example/alerts/ChatGwDown
    - alert: ChatGwHighErrorRatio
      expr: |
        sum by (service) (rate(http_requests_total{service="chat-gw",code=~"5.."}[5m]))
        / sum by (service) (rate(http_requests_total{service="chat-gw"}[5m])) > 0.1
      for: 10m
      labels:
        severity: page
        team: social
      annotations:
        summary: "{{ $labels.service }} is failing {{ $value | humanizePercentage }} of requests"
        runbook_url: https://runbooks.cobaltline.example/alerts/ChatGwHighErrorRatio
    # factors from the SRE workbook's multiwindow table, don't tune per service
    - alert: ChatGwErrorBudgetBurnFast
      expr: |
        slo:sli_error:ratio_rate1h{service="chat-gw"} > (14.4 * 0.001)
        and slo:sli_error:ratio_rate5m{service="chat-gw"} > (14.4 * 0.001)
      for: 2m
      labels:
        severity: page
        team: social
        slo: chat-gw-availability
      annotations:
        summary: "{{ $labels.service }} is spending its error budget fast"
        runbook_url: https://runbooks.cobaltline.example/alerts/ChatGwErrorBudgetBurnFast
    - alert: ChatGwErrorBudgetBurnSlow
      expr: |
        slo:sli_error:ratio_rate6h{service="chat-gw"} > (6 * 0.001)
        and slo:sli_error:ratio_rate30m{service="chat-gw"} > (6 * 0.001)
      for: 15m
      labels:
        severity: ticket
        team: social
        slo: chat-gw-availability
      annotations:
        summary: "{{ $labels.service }} is spending its error budget steadily"
        runbook_url: https://runbooks.cobaltline.example/alerts/ChatGwErrorBudgetBurnSlow
    # average the pods: they all run with the same limit, so the mean is what we size it by
    - alert: ChatGwCpuThrottled
      expr: |
        avg by (service) (rate(container_cpu_cfs_throttled_periods_total{service="chat-gw",container="chat-gw"}[5m]) / rate(container_cpu_cfs_periods_total{service="chat-gw",container="chat-gw"}[5m])) > 0.25
      for: 30m
      labels:
        severity: ticket
        team: social
      annotations:
        summary: "{{ $labels.service }} pods spend over a quarter of their CPU periods throttled"
        runbook_url: https://runbooks.cobaltline.example/alerts/ChatGwCpuThrottled
AF_EOF