Soru

Zorluk: Çok zorIncident Management and Automated Alerting

A global gaming platform deployed on Google Cloud experiences periodic database latency spikes. These spikes cause temporary downstream HTTP 500 error bursts across Compute Engine backend instance groups. During these incidents, Cloud Monitoring generates hundreds of redundant notification emails per minute, causing severe alert fatigue. Furthermore, automated remediation scripts triggered directly by raw error thresholds attempt to reboot instance group VMs, leading to prolonged service degradation because load balancer health checks are configured to execute deep database queries. Which incident management and automated alerting architecture should you implement to eliminate alert storms and prevent cascading service outages during transient database latency?

  1. Transition alerting policies to multi-window burn rate conditions based on latency and error rate Service Level Indicators (SLIs), route alert notifications through Cloud Pub/Sub to trigger a Cloud Run remediation service that dynamically applies alerting silences during active incidents, and update backend load balancer health checks to query shallow HTTP health endpoints.Cevap
  2. B
    Configure metric threshold alerting policies directly on backend HTTP 500 counts to invoke Cloud Functions via Webhooks for service restarts, and modify load balancer health checks to run synthetic SQL queries against the primary database to guarantee end-to-end backend availability.
  3. C
    Set static metric threshold alert policies on Compute Engine instance group CPU utilization, publish notifications to a central broadcast email list, and grant the Owner primitive IAM role to automated remediation service accounts so they can reconfigure firewall rules during operational incidents.
  4. D
    Establish static metric alerts based on raw latency SLIs configured with zero-duration alignment windows, and restrict remediation Cloud Functions access using IAM roles while allowing external webhook calls through VPC Service Controls without defining ingress/egress rules.

Cevap

Implement multi-window error budget burn rate alerting policies paired with Cloud Pub/Sub and Cloud Run for automated incident muting, and switch load balancer health checks to shallow local endpoints.
The solution addresses both root causes: using multi-window SLO burn rate alerting prevents alert storms caused by transient spikes, while decoupling backend health checks (using shallow local endpoints) prevents load balancers from tearing down healthy compute nodes when downstream databases experience transient latency. Automated event handling via Pub/Sub and Cloud Run enables controlled alert suppression and remediation.

Adım Adım Çözüm

1
Redesign alerting logic using Service Level Objectives (SLOs) and multi-window burn rate alerts in Cloud Monitoring.
Alerts trigger only when error budget consumption threatens the monthly target, eliminating alert storms from short transient spikes.
Multi-window burn rate alerting correlates short-term and long-term consumption windows to prevent false positives while remaining responsive to major outages.
2
Decouple health check probes from downstream backend dependencies.
Load balancers inspect shallow HTTP endpoints (e.g., local application process status) instead of executing deep database queries.
Deep dependency health checks cause load balancers to falsely mark instances as dead during transient database slowdowns, triggering cascading instance failures.
3
Automate incident notification routing and silence management via Cloud Pub/Sub and Cloud Run.
Notifications publish to Pub/Sub topics, triggering Cloud Run services that suppress duplicate alerts during active incidents.
Programmatic incident handling via Cloud Pub/Sub avoids uncoordinated script execution and prevents alert fatigue.

Anahtar Kavram

Multi-window SLO burn rate alerting and decoupled shallow health checking in enterprise GCP incident response.
Bu soruyu puanla