Oodle Alerting

Oodle Alerting — Notifiers, Policies, Muting

This skill teaches the agent to wire alert routing end-to-end: notifiers (where alerts go), notification policies (how they're matched and routed), and muting rules (when to silence them).

Prerequisites

brew install oodle-ai/oodle/oodle
oodle configure

Verify all three subsystems respond:

oodle notifiers list -o json | jq 'length'
oodle notification-policies list -o json | jq 'length'
oodle muting-rules list -o json | jq 'length'

Command Execution Order

Before running any oodle command:

1. Check whether the required resource ID or name is already in context.
2. If not, run the discovery command (e.g., oodle notifiers list -o json).
3. If the result is ambiguous, ask the user to confirm before proceeding.
4. Run the target command with the resolved ID.
5. Do not run speculative commands (e.g., do not delete a notifier without checking which policies reference it).

Quick Reference

Common Operations

Notifiers — channel definitions

A notifier wraps a destination (Slack, PagerDuty, email, webhook). The config block depends on type.

Slack:

{
  "name": "slack-ops",
  "type": "slack",
  "config": { "webhookUrl": "https://hooks.slack.com/services/T000/B000/XXXX" }
}

PagerDuty:

{
  "name": "pd-platform",
  "type": "pagerduty",
  "config": { "integrationKey": "abc123def456" }
}

Email:

{
  "name": "email-platform",
  "type": "email",
  "config": { "addresses": ["oncall@example.com", "alerts@example.com"] }
}

Webhook:

{
  "name": "webhook-incident-bot",
  "type": "webhook",
  "config": { "url": "https://incidents.example.com/oodle" }
}

# ✅ CORRECT
oodle notifiers create -f notifier.json

# ❌ WRONG — missing config block, server returns 400
oodle notifiers create -f <(echo '{"name":"x","type":"slack"}')

Notification policies — routing

A notification policy matches alerts by labels and routes them to a receiver (notifier name).

{
  "name": "platform-prod",
  "matchers": [
    {"name": "team", "value": "platform"},
    {"name": "env",  "value": "prod"}
  ],
  "receiver": "slack-ops",
  "routes": [
    {
      "matchers": [{"name": "severity", "value": "critical"}],
      "receiver": "pd-platform"
    }
  ]
}

# ✅ CORRECT
oodle notification-policies create -f policy.json

# ❌ WRONG — references a notifier that doesn't exist; server returns 400
# (no `receiver` set or `receiver: "non-existent"`)

Muting rules — scheduled silence

Muting rules silence alerts whose labels match the matchers, between startsAt and endsAt (RFC 3339).

Always scope a muting rule to a specific monitor by including the _oodle_monitor_id matcher. Without it, the rule can silence unrelated alerts that happen to share the other labels. Resolve the monitor ID first with oodle monitors list -o json (or oodle monitors get <id>).

Matcher type values: 0 = equals, 1 = not-equals, 2 = regex-match, 3 = regex-not-match.

{
  "name": "deploy-window-2024-01-15",
  "matchers": [
    {"type": 0, "name": "_oodle_monitor_id", "value": "01998476-6dd3-73e8-a4c0-d9137eef6bc9"},
    {"type": 0, "name": "service", "value": "api"},
    {"type": 0, "name": "env",     "value": "prod"}
  ],
  "startsAt": "2024-01-15T02:00:00Z",
  "endsAt":   "2024-01-15T06:00:00Z"
}

# ✅ CORRECT — bounded window, scoped to a specific monitor
oodle muting-rules create -f mute.json

# ❌ WRONG — endsAt before startsAt (rejected by server)
# {"startsAt": "2024-01-15T06:00:00Z", "endsAt": "2024-01-15T02:00:00Z"}

# ❌ WRONG — no _oodle_monitor_id matcher; silences every alert with these labels
# "matchers": [{"type":0,"name":"service","value":"api"},{"type":0,"name":"env","value":"prod"}]

Best Practices

Test a new notifier with a low-severity monitor before wiring it to a critical alert

A misconfigured Slack webhook fails silently — alerts disappear into the void.

# ✅ CORRECT — create notifier, attach to a `severity=info` test monitor, fire it once,
# confirm message arrives, then attach to the production policy.
oodle notifiers create -f notifier.json
oodle monitors create -f test-monitor.json   # severity=info, low threshold
# wait for fire, confirm receipt, then:
oodle notification-policies update pol_123 -f policy.json   # adds the new notifier

# ❌ WRONG — attach an untested notifier directly to a `severity=critical` policy
oodle notifiers create -f notifier.json
oodle notification-policies create -f policy-critical.json

Always get a notifier before deleting it — check which policies reference it

Deleting a notifier referenced by an active policy causes alerts to be dropped silently.

# ✅ CORRECT — find references first
oodle notifiers get notif_slack_ops -o json
oodle notification-policies list -o json | jq '.[] | select(.receiver=="slack-ops" or (.routes[]?.receiver=="slack-ops")) | .id'
# update those policies to a different receiver, then:
oodle notifiers delete notif_slack_ops --force

# ❌ WRONG — delete without checking; live policies suddenly route to a missing receiver
oodle notifiers delete notif_slack_ops --force

Always include _oodle_monitor_id in muting rule matchers

A mute without _oodle_monitor_id silences every alert whose labels happen to match — frequently far more than intended. Pin it to the monitor being silenced.

# ✅ CORRECT — scoped to one monitor
"matchers": [
  {"type":0,"name":"_oodle_monitor_id","value":"01998476-6dd3-73e8-a4c0-d9137eef6bc9"},
  {"type":0,"name":"service","value":"api"}
]

# ❌ WRONG — any monitor with service=api gets silenced
"matchers": [{"type":0,"name":"service","value":"api"}]

Resolve the monitor ID with oodle monitors list -o json | jq '.[] | {id,name}' before authoring the rule.

Always set endsAt on muting rules — never create open-ended mutes

An open-ended mute that nobody remembers becomes a permanent silence on a critical alert.

# ✅ CORRECT — bounded mute window
"startsAt": "2024-01-15T02:00:00Z", "endsAt": "2024-01-15T06:00:00Z"

# ❌ WRONG — open-ended; alerts stay silenced forever
"startsAt": "2024-01-15T02:00:00Z", "endsAt": null

Scope policy matchers to the smallest set of labels that uniquely identify the team

Overly-broad matchers ({env: prod} only) route every prod alert to one notifier.

# ✅ CORRECT
"matchers": [{"name":"team","value":"platform"},{"name":"env","value":"prod"}]

# ❌ WRONG — every prod alert in the org routes here
"matchers": [{"name":"env","value":"prod"}]

Use nested routes for severity escalation, not separate top-level policies

A child route inherits the parent matchers; defining a separate top-level policy for severity=critical will double-fire.

# ✅ CORRECT
"matchers": [{"name":"team","value":"platform"}],
"receiver": "slack-ops",
"routes": [{"matchers":[{"name":"severity","value":"critical"}],"receiver":"pd-platform"}]

# ❌ WRONG — two top-level policies, both match, both fire

Failure Handling

References

• Oodle CLI repo
• Oodle docs