Symptoms

Cluster management IP was running on Node-A.The user disconnected both management ports for an appliance that holds the cluster management IP and reconnected the ports.User can ping Node-B management IP.User cannot ping Cluster Management IP when IP resides on Node-B.

Cause

Cluster IP is a floating IP that can move between the cluster nodes. When we move the cluster IP to the peer node, we also send a special ARP packet (Using Gratuitous ARP (GARP)) to notify the switch to update its ARP table so the next frame forwarding is done to the right MAC address.Cluster IP preferentially runs on a secondary node unless the secondary node is either down or its management is disconnected. If the secondary node resumes or management is reconnected, cluster management IP trespasses back to the secondary node.If both nodes management ports are down for any reason, Cluster management IP always moves back to the secondary node. However, in scenarios where the secondary node is not connected to the management switch, it cannot send the GARP request to update the switch ARP table.This creates a misalignment in the switch ARP table as ARP packet is not sent to the switch (port is down) and the switch ARP table has the wrong MAC address associated to the cluster IP once the physical port is up.

Resolution

This issue is planned to be addressed in a future release of PowerStore.WorkaroundUse any of the below options to work around the issue: If both management ports are up, unplug node-B's management cable, wait 1 minute and plug it back in (floating IP moves to node-A and then moves back to node-B).Remove the stale ARP entry of the management IP from the switch ARP table.Wait about 10 minutes for ARP information to the switches to sync.Try leveraging tools such as arping to the Cluster IP to force the hosts to attempt refreshing the ARP table.Contact Dell Technical Support to confirm this issue has occurred. Reference this article ID in the Service Request for expedited handling.

Support Cases