Pmk Troubleshooting Guide

Issues by Component

Cluster

• Cluster A PMK cluster needs to be created by using onboarded (authorized) nodes. Refer the table below for issues related to a cluster: auto$ Cluster

	Cluster
Component/Topic	Symptoms/Error Messages	Link to KB Article
BareOS Cluster Creation	Cluster creation fails, UI may show the failing step.	auto$
Cluster Creation using Public Cloud Provider (e.g. AWS)	Cluster creation fails, UI may show the failing step.	auto$
Etcd Configuration	- Heartbeat/Election Timeout Interval - Database Size Exceeded	auto$

Nodes

• Nodes: Linux servers are configured by PMK before they can be used to create a cluster. The configuration process includes installing PMK-specific packages and verifying other prerequisites.

	Nodes
Component/Topic	Symptoms/Error Messages	Link to KB Article
VIP association on Master nodes	VIP Not Routable from Other Masters (Misconfigured)	auto$
Node Preparation / Onboarding / Node Not Converged	Incompatible Package Version(s)	auto$
Clock Skew	- PF9 Host agent fails to generate certificates. - Error message in hostagent.log: “Unable to vouch URL …”	auto$

Pods

• Pods While deploying workloads to Kubernetes (PMK), you may encounter issues around starting pods for deployments. If the dashboard (UI) reports unhealthy workload, refer the table below:

	Pods
Component/Topic	Symptoms/Error Messages	Link to KB Article
Pods / Deployments	Error: ImagePullBackOff	auto$
Node Preparation / Onboarding / Node Not Converged	Error: CrashLoopBackOff	auto$

Networking

• Network: Various issues seemingly related to Kubernetes workloads may be caused by underlying network issues. Refer the table below for known networking issues:

	Network
Component/Topic	Symptoms/Error Messages	Link to KB Article
DNS	Errors due to domain name/host name resolution failure	auto$
Calico CNI	Pod Networking broken / Kernel IP Forwarding not enabled on host	auto$

Applications

• Applications: Applications can be deployed using the Apps Catalog tab of Apps Dashboard. Applications can only be deployed to clusters that have been registered with a repository. The table below outlines some common issues when deploying or managing apps.

	Applications
Component/Topic	Symptoms/Error Messages	Link to KB Article
MetalLB	MetalLB is configured but doesn’t work.	auto$

CLI

• ** CLI
• *: Troubleshooting steps for known issues of Command Line Interface clients.

	CLI
Component/Topic	Symptoms/Error Messages	Link to KB Article
Kubectl	- API Server Unreachable - Invalid Token in Kubeconfig	auto$
Etcdctl	- Incorrect Endpoint(s) - Certificates Not Specified	auto$

AWS EC2 Clusters

• Troubleshooting guidance for known issue on clusters using AWS Cloud Provider:

	AWS
Component/Topic	Symptoms/Error Messages	Link to KB Article
Instance registration on Elastic Load Balancer	Elastic Load Balancer (ELB) Shows No Active Instances	auto$
Node Preparation / Onboarding / Node Not Converged	NodePort Service Isn't Externally Reachable	auto$