Mode d’emploi
- Copiez la commande et examinez le paquet avant de l’installer dans votre environnement agent.
- Consultez le document original pour les outils, identifiants et environnements requis.
- Décrivez les entrées et le livrable souhaité ; comparez la sortie aux sources.
npx skills add https://github.com/oodle-ai/agent-skills --skill oodle-drop-rulesAvant de commencer
- L’installation n’inclut ni abonnement au modèle, ni quota API, ni accès aux fichiers.
- Étoiles et installations reflètent l’activité, pas un test indépendant de qualité ou de sécurité.
Rendu depuis le dépôt source en conservant titres, exemples, code, tableaux, liens et images.
Oodle Drop Rules — Cost Control
This skill teaches the agent to drop high-volume metrics safely: estimate the impact first, start with a narrow rule, and never silently delete a metric a dashboard depends on. There is no sampling mode: a rule drops every series that it matches, and dropped data cannot be recovered.
Prerequisites
brew install oodle-ai/oodle/oodle
oodle configureConfirm the drop-rules endpoint works:
oodle drop-rules list -o json | jq 'length'Command Execution Order
Before running any oodle command:
- Check whether the target metric prefix and matchers are already in context.
- If not, run
oodle metrics query --query 'count({__name__=~"<prefix>.*"})' -o jsonto estimate impact. - Confirm no critical dashboards or monitors depend on the affected series (see "Estimating impact before creating a rule").
- Start with a narrow rule (one metric, one non-production
env) on the first attempt; widen thefiltersonly after the dashboards and monitors of that env show no regression. - Run
oodle drop-rules create -f rule.json.
Quick Reference
| Task | Command |
|---|---|
| List rules | oodle drop-rules list -o json |
| Get rule | oodle drop-rules get <id> -o json |
| Create rule | oodle drop-rules create -f rule.json |
| Update rule | oodle drop-rules update <id> -f rule.json |
| Delete rule | oodle drop-rules delete <id> --force |
Common Operations
Rule schema
{
"rule_name": "drop-noisy-debug-metrics",
"metric_name": {"type": 2, "name": "__name__", "value": "debug_.*"},
"filters": [
{"type": 0, "name": "level", "value": "debug"},
{"type": 0, "name": "env", "value": "staging"}
],
"type": "series"
}| Field | Meaning |
|---|---|
rule_name | Required. Name of the rule |
metric_name | Matcher on the metric name ("name": "__name__") |
filters | Label matchers. All must match for the rule to fire (logical AND) |
type | Required. Must be "series" |
Matcher type: 0 equals, 1 not equals, 2 regex match, 3 regex not match.
Estimating impact before creating a rule
# ✅ CORRECT — count series that will be affected
oodle metrics query --query 'count({__name__=~"debug_.*"})' -o json
# ✅ CORRECT — confirm no dashboard panel queries the metric (list has no panels; get each dashboard)
for uid in $(oodle dashboards list -o json | jq -r '.[].uid'); do
oodle dashboards get "$uid" -o json | grep -q 'debug_traffic_total' && echo "$uid"
done
# ✅ CORRECT — confirm no monitor depends on it
oodle monitors list -o json | jq '.[] | select(.promql_query | contains("debug_traffic_total"))'
# ❌ WRONG — create the rule first, find out from on-call later
oodle drop-rules create -f rule.jsonCreating a narrow rule (safer first step)
{
"rule_name": "drop-debug-traffic-staging",
"metric_name": {"type": 0, "name": "__name__", "value": "debug_traffic_total"},
"filters": [
{"type": 0, "name": "env", "value": "staging"}
],
"type": "series"
}# ✅ CORRECT — one metric in staging; observe for a week before widening
oodle drop-rules create -f rule.jsonWidening a rule
# ✅ CORRECT — get → change filters → update
oodle drop-rules get dr_123 -o json > rule.json
jq '.filters = [{"type": 2, "name": "env", "value": "staging|dev"}]' rule.json > rule.new.json
oodle drop-rules update dr_123 -f rule.new.json
# ❌ WRONG — partial payload nulls metric_name and filters
oodle drop-rules update dr_123 -f <(echo '{"filters":[]}')Deleting (re-enabling ingestion)
# ✅ CORRECT
oodle drop-rules get dr_123 -o json > /dev/null
oodle drop-rules delete dr_123 --force
# ❌ WRONG — speculative delete by name match
oodle drop-rules delete "$(oodle drop-rules list | grep debug | awk '{print $1}')" --forceBest Practices
Estimate impact with oodle metrics query --query 'count({__name__=~"<prefix>.*"})' before creating a rule
A drop rule that matches more series than expected can hide real signal.
# ✅ CORRECT
oodle metrics query --query 'count({__name__=~"kube_pod_.*"})' -o json | jq -r '.data.result[0].value[1]'
# 1742 series — confirm with the team that all 1742 are safe to drop before creating a rule
# ❌ WRONG — create rule based on a guess; later discover a critical metric was matched
oodle drop-rules create -f rule.jsonStart with a narrow rule in a non-production env on the first iteration
There is no sampling mode. A narrow first rule shows that the metric truly is low-value before you drop it everywhere.
# ✅ CORRECT — week 1: drop in staging only, observe dashboards
"filters": [{"type": 0, "name": "env", "value": "staging"}]
# week 2: if no dashboards or monitors regressed, widen the filters
# ❌ WRONG — drop everywhere on first attempt; can break a dashboard nobody remembered
"filters": []Always include both env and service (or an exact metric_name) in matchers
Broad matchers like {level: debug} alone can match production metrics that happen to share a label.
# ✅ CORRECT — scoped to one env
"filters": [{"type":0,"name":"level","value":"debug"},{"type":0,"name":"env","value":"staging"}]
# ❌ WRONG — also drops debug metrics in prod
"filters": [{"type":0,"name":"level","value":"debug"}]Always get before update to preserve fields
Update is a full-document replace.
# ✅ CORRECT
oodle drop-rules get dr_123 -o json > rule.json
jq '.filters += [{"type": 0, "name": "service", "value": "api"}]' rule.json > rule.new.json
oodle drop-rules update dr_123 -f rule.new.json
# ❌ WRONG — drops metric_name and type
oodle drop-rules update dr_123 -f <(echo '{"filters":[{"type":0,"name":"service","value":"api"}]}')Name rules with drop-<metric-or-domain>-<scope>
Predictable names make it easy to find and revert a rule when a dashboard breaks.
# ✅ CORRECT
"rule_name": "drop-debug-metrics-staging"
"rule_name": "drop-kube-pod-info-prod"
# ❌ WRONG
"rule_name": "rule1"Failure Handling
| Error | Cause | Fix |
|---|---|---|
| 401 Unauthorized | Invalid or missing API key | Run oodle configure or set OODLE_API_KEY |
| 404 Not Found | Drop rule ID does not exist | Verify with oodle drop-rules list -o json |
| connection refused | Wrong OODLE_DEPLOYMENT URL | Check OODLE_DEPLOYMENT env var |
rule_name cannot be empty | Missing rule_name | Add rule_name |
type cannot be empty / invalid type "...", must be "series" | type is missing or wrong | Set "type": "series" |
| Dashboard panel suddenly empty | Drop rule matched a metric the panel queries | Run oodle drop-rules list -o json to find the rule; delete it (oodle drop-rules delete <id> --force) or narrow its matchers |
Monitor went into no data | Drop rule matched the monitor's metric | Same fix as above |
| Cost did not decrease | Matchers don't actually match the high-volume series | Re-run the count(...) query with the rule's matchers and compare label sets |
| 429 Too Many Requests | Bulk drop-rule sync | Add --retries 3, throttle to <10 creates per second |