Google Cloud Monitoring Dashboard Samples

These alert policies are designed for GKE nodes using Node Exporter to monitor three critical resource thresholds:

1. High CPU Usage alert: Triggers when CPU usage is high, indicating potential performance degradation due to excessive workloads.
2. Low available memory alert: Triggers when available memory is low, which may prevent GKE from assigning new pods to the node.
3. Low filesystem space alert: Triggers when filesystem space is low, which may prevent GKE from assigning new pods to the node.

This directory provides Cloud Monitoring dashboard JSON configurations for Azure Virtual Machines. These dashboards are specifically designed to work with metrics collected via BindPlane.

There are two available dashboard configurations:

1. Azure Virtual Machines - OS Metrics (os.json): Contains 13 charts focusing on OS-level performance, such as Disk Queue Depth, Disk Read/Write Data, Disk Read/Write Operations, CPU Utilization, and Network In/Out.
2. Azure Virtual Machines Overview (overview.json): Contains 7 widgets providing a high-level overview, including CPU Credits Consumed, CPU Credits Remaining, Network Received, Disk Read/Writes, CPU Utilization, and Network Sent.

This set of alerts is designed for Apache Flink workloads monitored via the Google Cloud Ops Agent. It covers job stability, checkpoint health, and JVM memory utilization.

Important: For any of these alerts to be useful, you must configure a Notification Channel in Cloud Monitoring, otherwise, the alerts will fire silently without notifying you.

This sample provides alert configurations for the NGINX Ingress Controller running on Google Kubernetes Engine (GKE). It covers three primary alert types based on connection and request metrics:

1. Connections dropped alert: Monitors the difference between connections accepted and connections handled. A rising value suggests resource saturation.
2. High request rate alert: Monitors spikes in traffic using a 5-minute request rate interval. Use the thresholdValue to define what constitutes a spike for your specific workload.
3. Low request rate alert: Monitors for drops in traffic using a 5-minute request rate interval. Use the thresholdValue to define the baseline for normal operation to detect environment issues limiting requests.

The Hadoop monitoring samples include three primary alert types designed for use with the Ops Agent:

1. Low available capacity alert: Triggers when total disk space usage across all HDFS clusters exceeds 80% of the capacity limit. This indicates a need to increase capacity.
2. Volume failure alert: Triggers when a hardware failure occurs on a volume. While replication typically prevents data loss, this alert signals that hardware needs replacement.
3. Dead data nodes alert: Triggers when a data node becomes unresponsive. This can cause network congestion as the NameNode attempts to re-replicate lost blocks. Multiple dead nodes may lead to data loss.