This Cast AI feature informs you via UI, Slack, or webhook about key issues affecting the cluster. This guide outlines all notification types and examples you may see in Cast AI with relevant action points.

Once new items are ready for you to view, the bell icon in the top menu will show a count. You can view all items in the Notifications page.

Due to the dynamic nature of Kubernetes clusters, notifications are set to expire automatically in 24 hours.

Notification severity types

Cast AI uses several notification severity types to categorize the severity and importance of messages.

Severity	Description
Critical	Indicates a severe issue that requires immediate attention and may significantly impact cluster operations.
Error	Signifies a problem that is causing a malfunction or preventing expected behavior.
Warning	Alerts about potential issues or situations that could lead to problems if not addressed.
Info	Provides general information about cluster operations, updates, or status changes.
Success	Confirms that an operation or process has completed successfully.

Notification categories

Cast AI organizes notifications into the following categories:

Workload Autoscaler - Workload autoscaling event notifications
Node Autoscaler - Node autoscaling and scheduling event notifications
Reporting anomalies - Cost and performance anomaly detection notifications
Inventory - New cloud provider instance availability notifications
Security - Image security and best practice notifications
Other - General cluster operations, system connectivity, and configuration issues

Complete notifications reference

This section provides a comprehensive list of all notifications that Cast AI can generate, organized by category and severity level.

Critical notifications

Other category

Notification	Description
The Cast AI agent is unable to connect to the API	The Cast AI agent cannot communicate with Cast AI API endpoints and is unable to send data. Check the status of agent pods, network connectivity, firewall restrictions, or authentication.
Cluster controller not responding	The cluster controller component is unresponsive. This can prevent cluster management operations and autoscaling from functioning properly.
IP Address quota exceeded	Your cloud provider's IP address quota has been exceeded, preventing new nodes from being created with proper network connectivity.
Node Configuration Validation Failed	Node configuration is invalid. Check the notification details for specific configuration errors.
Node deletion failed	Cast AI was unable to delete a node from the cluster. This may indicate permission issues or cloud provider API problems.
Operation failed	A critical cluster operation has failed. Check notification details for which operation failed and why.
Spot Instance quota exceeded	Your cloud provider's Spot Instance quota has been exceeded, preventing Cast AI from launching cost-effective Spot Instances for your workloads. Additionally, since Spot Fallback is not enabled, the Autoscaler might not be able to add any capacity.

Error notifications

Other category

Notification	Description
Missing permission when adding a node to a target group	Cast AI lacks the necessary IAM permissions to add nodes to target groups in your load balancer.
Missing permission when adding a node to load balancer(s)	Cast AI lacks the necessary IAM permissions to add nodes to load balancers.
Missing permission when adding a node to target groups	Cast AI lacks the necessary IAM permissions to add nodes to target groups.
Missing permission when adding VMSS IP address to a backend pool	Cast AI lacks the necessary Azure permissions to add Virtual Machine Scale Set IP addresses to backend pools.
Missing permission when deleting a node from target groups	Cast AI lacks the necessary IAM permissions to remove nodes from target groups.
Missing permission when removing a node from load balancer(s)	Cast AI lacks the necessary IAM permissions to remove nodes from load balancers.
SSO Connection problem	There is an issue with your Single Sign-On configuration preventing proper user authentication.

Warning notifications

Workload Autoscaler category

Notification	Description
Continuous OOMKilled Events Detected	A continuous stream of out-of-memory (OOMKilled) events was detected for a workload. Workload automation is disabled and re-enabled automatically after a cooldown period. Check if the application is healthy.

Node Autoscaler category

Notification	Description
Pending pod detected	A pod has been stuck in a pending (unscheduled) state. Cast AI is monitoring the situation; this notification resolves automatically once the pod is scheduled.

Reporting anomalies category

Notification	Description
Cost anomaly detected on the {metric} metric / Cost anomalies detected ...	Cast AI monitors cost and efficiency metrics across your cluster (e.g. compute cost, CPU/RAM overprovisioning, network egress). The notification name and message identify which metrics triggered detection.

Other category

Notification	Description
Failed Helm Test of castai-workload-autoscaler	The Helm test for the Cast AI workload autoscaler component has failed, indicating potential deployment or configuration issues.
Failed to reconcile cluster	A non-critical reconciliation issue occurred. While not immediately severe, this should be monitored and addressed.
GPU quota exceeded	Your cloud provider's GPU quota has been exceeded, preventing allocation of GPU resources for workloads that require them.
Outdated cluster-controller	The cluster controller component is running an outdated version and should be updated for optimal functionality and security.
Unable to update pool	Cast AI was unable to update a node pool configuration, which may prevent scaling operations or configuration changes.
Spot Instance quota exceeded	Your cloud provider's Spot Instance quota has been exceeded, preventing Cast AI from launching cost-effective Spot Instances for your workloads. Fires as a warning when Spot Fallback is enabled, since the Autoscaler can fall back to on-demand instances.

Info notifications

Inventory category

Notification	Description
New machines available in AWS	New AWS instance types have become available and can now be used by Cast AI for your clusters.
New machines available in Azure	New Azure virtual machine sizes have become available and can now be used by Cast AI for your clusters.
New machines available in GCP	New Google Cloud Platform machine types have become available and can now be used by Cast AI for your clusters.

Other category

Notification	Description
Read-Only access activated	Cast AI has been activated in read-only mode, which means it can monitor your cluster but cannot make changes to it.
Trial expires soon	Your Cast AI trial period is approaching its expiration date. Consider upgrading to a paid plan to continue using Cast AI features.
Trial has expired	Your Cast AI trial period has expired. Upgrade to a paid plan to restore full functionality.

Success notifications

Other category

Notification	Description
Cluster reconciled	The cluster has been successfully reconciled, indicating that Cast AI has successfully synchronized the desired cluster state with the actual state.

Taking action on notifications

When you receive notifications, consider the following general action steps:

Critical notifications - Address immediately as they can severely impact cluster operations
Error notifications - Investigate and resolve permission issues or configuration problems
Warning notifications - Review for potential cost savings or performance improvements
Info notifications - Stay informed about new capabilities and system status
Success notifications - Confirm that operations completed as expected

For specific troubleshooting steps related to individual notifications, consult the relevant Cast AI documentation or contact support if the issue persists.