oxygen914/kubeoncall/network-connectivity-triage

Network Connectivity Triage

Diagnostic objective

Locate failure at DNS resolution, endpoint selection, connection establishment, TLS, or packet transport without changing network state.

Required live evidence

Identify source, destination, port, protocol, affected Pods and nodes, first failure time, DNS result, Service selectors, ready endpoints, connection outcome, retransmits, packet loss, and conntrack utilization.

Decision branches

• DNS resolution fails: compare resolver reachability, response code, search path, and affected namespaces.
• Service resolves but has no ready endpoints: inspect selectors and Pod readiness; treat it as workload evidence.
• Connections time out or reset: compare node locality, policy evidence, retransmits, and destination health.
• Conntrack is high: confirm entries, limit, growth, and node concentration before assigning root cause.

Safe next step

Use the relevant network or conntrack runbook and report the exact failed hop with independent evidence.

Prohibited actions

Do not edit NetworkPolicy, flush conntrack, restart networking, change DNS, or bypass TLS from this Skill.

Recovery and escalation

Require stable resolution, ready endpoints, successful connections, and normal loss/retransmit indicators. Escalate for multi-node impact, control-plane isolation, or sustained conntrack exhaustion.