Alert Reference
ArgoCD Alerts
ArgoCD Sync Status
Whether the desired state in Git matches what's actually deployed in the cluster. A failed sync usually means Git access issues, resource conflicts, or a misconfiguration.
ArgoCD Health Status
The overall health of an ArgoCD-managed application. A degraded or failed status points to a service disruption, connectivity issue, or configuration error.
Application Alerts
Daemonset Replicas Unavailable
Fewer DaemonSet replicas are running than desired. Usually caused by node failures, resource exhaustion, scheduling constraints, or a misconfiguration.
Deployment Replicas Unavailable
Fewer Deployment replicas are running than desired. Usually caused by insufficient resources, container failures, or a rollout conflict.
StatefulSet Replicas Unavailable
Fewer StatefulSet replicas are running than desired. Usually caused by node failures, resource constraints, or a misconfiguration.
Pod CPU Throttling Error
A pod is being CPU-throttled because it's exceeding its CPU limit, degrading its performance.
Pod Errors
One or more pods hit an error during execution. Causes range from misconfiguration and resource limits to network issues or application-specific failures. Check the pod's events for the specific cause.
Pod Memory Failures
A pod hit a memory-related failure: an allocation error, an OOM kill, or a leak in one of its containers.
Pod Memory Limits
A pod's memory usage exceeded its configured limit, triggering resource contention. Adjust the limit or the workload's actual memory usage.
Pod Memory Requests
A pod's memory usage is running well above its configured request. Requests set too low risk resource starvation; requests set too high waste cluster capacity.
Pod OOM Kill
A pod was killed because the node ran out of memory. Kubernetes terminates the pod to free memory and keep the node stable.
Pod Readiness
One or more pods aren't reaching a ready state, so they can't serve traffic. Common causes: application startup failures, unmet dependencies, or failing readiness probes.
Pod Restarts
A pod restarted more than 5 times within an hour, indicating instability or resource pressure. Frequent restarts disrupt availability and need investigation.
Ingress Alerts
Ingress Certificate Expiry
The TLS certificate on an Ingress is approaching its expiration date. An expired certificate causes service interruptions and browser warnings for users.
Ingress 5xx Error Rate
The rate of HTTP 5xx errors on an Ingress is elevated. Usually points to an issue in application logic, a backend service, or infrastructure.
Ingress Request Count
An Ingress is receiving an unusually high number of requests over a given period.
Ingress Request Latency
Requests through an Ingress are taking longer than expected to process. High latency degrades user experience and often points to network, backend, or resource contention issues.
Node Alerts
Node CPU Usage
A node's CPU usage is high. Often means resource contention, inefficient allocation, or increased demand from running workloads.
Node Load15
A node's 15-minute load average is high, indicating sustained resource saturation.
Node Load5
A node's 5-minute load average is high, a more immediate signal of resource pressure than Load15.
Node Disk Usage
A node's disk usage is high. Left unaddressed, this risks running out of storage capacity.
Node Network Throughput In
A node's inbound network throughput is high, useful for spotting traffic bottlenecks and bandwidth pressure.
Node Network Throughput Out
A node's outbound network throughput is high, useful for spotting bottlenecks in outbound traffic.
Node Unschedulable
A node has been marked unschedulable and won't receive new pod assignments. Causes include resource exhaustion, network issues, or a manual cordon.
Node Host OOM Kill
A node ran out of memory at the host level and killed processes to reclaim it. Points to insufficient node memory, a leak, or workloads using more than expected.
Node Memory Free
A node's free memory is low, which risks resource contention or a host-level OOM kill if it keeps dropping.
Node Memory Pressure
A node is under memory pressure. Left unaddressed, this leads to performance degradation, resource contention, or OOM kills.
Volume Alerts
Volume Stats Missing
Statistics for one or more volumes aren't available. Usually a storage-provider connectivity problem, misconfiguration, or missing permissions.
Volume Usage Critical
A volume has reached 95% usage. At this level, applications relying on it risk running out of space.
Volume Usage Warning
A volume has passed 90% usage. Plan for cleanup or expansion before it becomes critical.
Other Alerts
Velero Backups Failed
A Velero backup failed. Usually caused by object-storage connectivity issues, missing permissions, or resource constraints, jeopardizing your recovery capability.
Velero Backup Partially Failed
Part of a Velero backup succeeded, but some resources failed to back up. Often a transient error, resource conflict, or data-consistency issue.
Verify Image Cosign
Cosign signature verification failed for a container image. Points to an invalid signature, missing key, or a compromised image.