Read I/O spike every 5 mins on LUN used as scratch partition for many hosts

Products

• VMware vSphere ESXi

Issue/Introduction

• Storage array experiences high read I/O spike every 5 minutes
• Spikes seen on LUN configured for use as scratch LUN for many hosts

Environment

VMware vSphere ESXi 8.0.x

Cause

Each host queries its scratch storage every 5 minutes. When many hosts use the same LUN for scratch, a larger amount of metadata must be read by each host accessing it every 5 minutes, causing a spike in read I/O on the LUN.

 

Resolution

Limit the number of hosts using a shared LUN as a scratch location, and limit the size of the shared LUN accordingly.