vSAN iSCSI targets imbalanced in stretched cluster when affinity site is down

Products

• VMware vSAN

Issue/Introduction

Symptoms:

• In vSphere 7.0U1, support for vSAN iSCSI targets in stretched clusters was introduced.

• In a stretched cluster, iSCSI targets could be set to have site-affinity, e.g. primary site or secondary site.

• If there are many targets with site-affinity set to a site, these iSCSI targets are automatically balanced across the hosts in this site.

• If some iSCSI targets haven't been assigned any affinity attributes, these targets will be automatically balanced across all hosts in both sites through target owner change. The automatic balancing of owners is to avoid iSCSI traffic overwhelming a single host.

• If all target objects are set affinity to a site, and this site is down, all targets will fail over to the other site, in this scenario, the load balancing cannot function well, since load balancing cannot balance targets with opposite site affinity.

Environment

VMware vSAN 7.0.x

Cause

This issue is caused when the affinity site is down, these targets are failed over to the other site which is not the affinity site.

The assumption is that the site with affinity will be recovered soon and be back online. In order to reduce iSCSI target owner movements, the iSCSI target service will not balance the targets with the opposite site affinity as it will be moved to the affinity site soon after the failed site is recovered.

Resolution

VMware by Broadcom is aware of the situation. 

If the above symptoms and issue matches, please contact Broadcom Support to investigate the issue.