High front end read/write latency is observed on VMs during host reboot

Products

• VMware vSAN
• VMware vSAN 8.x

Issue/Introduction

During host reboots high front end read/write latency is observed on VMs. This high latency can be detrimental to latency sensitive VMs like databases (Oracle RAC, SQL, ect.) and can lead to potential soft evictions within the guest OS.

Host reboots are being performed via script.

From vSAN Perf Graphs via vCenter > vSAN Cluster > Monitor > vSAN > Performance you observe the following:

Note: The preceding Perf Graphs excerpts are only examples. Date, time, and environmental variables may vary depending on your environment.

Environment

VMware vSAN OSA & ESA 8.x

Cause

• When a host is put into maintenance mode, the VMs running on that host will migrate to other hosts in the cluster. As part of that migration, the ownership of the corresponding vSAN objects of the VM also moves to the same host where the VM is migrated to.
• This colocation of the VM and the ownership of the corresponding vSAN objects is confined to the namespace object and the virtual disk objects that are not multi-writer virtual disks.
• The vSAN objects corresponding to the multi-writer virtual disks may still continue to be owned by the host even after it entered the maintenance mode.
• Later if that host is rebooted, it causes the above mentioned latencies to be observed on those multi-writer virtual disks. 

 

Resolution

To resolve this, upgrade to 9.x as there are improvements to reduce the latency.

The fix will be backported to 8.0U3 P09 tentatively slated to be released Q3 2026.

If additional assistance is required open a case with vSAN Support.