Network partitioned or witness host not found in vSAN Stretch Cluster

Products

• VMware vSphere ESXi

Issue/Introduction

• vSAN health check reports "Network partitioned" or "Witness host not found."

• vmkping -I vmkX <Remote_IP> succeeds to other data nodes, where vmkX is the vmkernel for vSAN.

• vmkping -I vmkX <Remote_IP> fails to the witness node, where vmkX is the vmkernel for vSAN.

• tcpdump-uw on the witness shows no vSAN clustering traffic (ports 12345/12321) for some data nodes.

• Data nodes and the witness nodes are on different subnets (Layer 3).

Environment

• VMware vSAN 8.x

• VMware vSAN 7.x

Cause

Successful ICMP (ping) between data nodes on the same subnet are expected, as no routing is required. vSAN clustering traffic requires static routes when the data nodes and witness node are in different subnets.

Resolution

Configure static routes for the subnets used by the vSAN data and witness nodes.

1. Log in to the ESXi host via SSH.

2. Add a static route to the remote subnet: esxcli network ip route ipv4 add -g <Gateway_IP> -n <Remote_Subnet/Mask> Example: esxcli network ip route ipv4 add -g ###.####.####.### -n ###.####.####.0/24

3. Verify the route is active: esxcli network ip route ipv4 list

4. Repeat these steps on all the witness and data node appliance to ensure a route exists.

For example, witness node(s) should have a route for data nodes and data nodes should have a route for witness node(s).

For more information, see Creating Static Routes for Virtual SAN Networking.

To receive updates on this topic, subscribe to this article (reference: Subscribe to a Broadcom knowledge article by article or product).

Additional Information

Creating Static Routes for vSAN Networking