Troubleshoot Elastic Node Pools

If an Elastic Node Pool's nodes are not becoming ready or are otherwise stuck, consider the following steps.

Operations

The Operations tab shows recent or running operations accompanied by detailed logs. Check them first to see if something is going on or went wrong.

Inspect Nodes via kubectl

You can inspect the virtual machine instances backing your nodes using kubectl against the Fleet Cluster:

# List all VMIs in the namespace
kubectl -n <namespace> get vmi

# Describe a specific node's VMI
kubectl -n <namespace> describe vmi node-<xyz>

WARNING

The Fleet Cluster is managed by the Elastic Fleet provider and may not be accessible to you. Contact your provider if you don't have access.

Node Console

If you need to debug a node directly, you can access its console using virtctl against the Fleet Cluster:

virtctl -n <namespace> console node-<xyz>

To log in, use the Recovery User and Recovery Password shown in the Elastic Node Pool's Info tab under Linux. The recovery credentials are shared across all nodes in the pool.

Once logged in, useful commands to investigate:

# Kubernetes service logs (Elastic Nodes do not run a machine agent)
journalctl -u kubelet
journalctl -u containerd

# Kernel messages (hardware, driver, boot issues)
dmesg | tail -100

# Network status (Flatcar uses systemd-networkd)
ip addr
ip route
networkctl status
journalctl -u systemd-networkd