Excessive Storage DRS tasks in vCenter Server 7.0

Products

• VMware vCenter Server

Issue/Introduction

Symptoms:
After upgrading to vCenter Server 7.0, you notice that storage DRS is initiating a very high number of storage vMotion tasks.
As a result a lot of these tasks are stuck in queued status.
In certain situations this might result in Storage DRS overloading a datastore to a point where no free space is left.

Environment

VMware vCenter Server 7.0.x

Cause

SDRS checks the datastore usage every 10 min and if an overloaded datastore is identified or one which is close to be overloaded, a load balance session will be scheduled in 10 min no matter if there is already a scheduled one or not. Because Storage vMotion however is both time- and resource-consuming, the number of parallel Storage vMotion tasks on any given ESXi cannot go over 2 at the same time. Due to this, these LB sessions can pile up easily and as a result, some tasks may get stuck leading many of them to pile up.

Resolution

A fix for this issue has been published with vCenter Server 7.0 Update 3o.
Should you be still experiencing this issue with this or a newer vCenter version, please contact Broadcom support by opening a support request.

Workaround:
As a workaround you can reduce the aggressiveness for Storage DRS by setting the "minimum space utilization difference" value to a higher percentage, as shown in the image below.


This will result in SDRS applying a more conservative strategy when balancing datastore usage, leading to a lower number of Storage vMotion tasks and allowing vCenter to execute these tasks with a far lower risk in them queueing up.