How to identify specific processes or worlds (e.g. VM worlds) which are generating high I/O on a particular storage device.
This is useful where a particular device is generating unexpectedly high I/O.
While high or unexpected I/O is observed on a device on an ESXi host, run:
esxtop
Type u, to select the devices view.
Type e, and paste the device id at the prompt Device to expand/rollup (device):
Metrics are displayed per world for the device:
The world IDs can be correlated with VM worlds by correlating with the output of:
esxcli vm process list
Non-VM worlds can be identified, for example, by correlating them with the output of:
ps -c -P -J -s -t
esxcli storage core device world list -d <device naa id>
In the esxtop adapter view (type d) , I/O metrics can be viewed on a per path basis in similar manner:
Type e, and paste an adapter name at the prompt.