Hosts Flap Between Connected and Not Responding States Due to DNS Resolution Conflict

Products

• VMware vSphere ESXi

Issue/Introduction

Multiple ESXi hosts are intermittently going into a "Not Responding" state in vCenter. The host connectivity status keeps fluctuating between normal and disconnected. 

Environment

vSphere 7.x/8.x

Cause

The issue occurs due to incorrect or conflicting DNS resolution for the ESXi hosts. Specifically, the hostnames of the affected ESXi servers are associated with multiple A (address) records in the DNS server. As a result, when vCenter attempts to resolve the hostname, it may receive different IP addresses at different times. This causes communication inconsistencies between vCenter and the host, leading to the host being intermittently marked as "Not Responding".

This behavior can be observed by performing an nslookup on the hostnames from the vCenter Server, which returns two or more IP addresses for the same hostname. If one of the IPs is not reachable or does not belong to the intended ESXi host, vCenter will lose connectivity to the host until the next successful resolution, resulting in flapping host connectivity status.

This issue commonly arises in environments where:

• DNS records were manually created or cloned incorrectly.

• Duplicate A records exist in different DNS zones.

• Hosts were re-IP’d or re-purposed without proper DNS cleanup.

This DNS misconfiguration leads to unstable communication between vCenter and the ESXi hosts.

Resolution

 

• Permanent Fix:

  • Review the DNS configuration for each affected ESXi host.

  • Remove the unintended or additional A record (IP address) from the DNS server, ensuring each hostname resolves to a single, correct IP.

• Temporary Workaround:

  • Take a backup of the etc/hosts file on the vCenter.

  • Add a static entry for the ESXi host in the vCenter's /etc/hosts file using the correct IP address.
    Example:

     

    <Correct_IP> esxi-hostname.domain.com

  • Note: This is a temporary workaround. Once DNS is corrected, the static entry should be removed to avoid long-term inconsistency.