VMware Cloud Community

PowerCLI to access Cluster Level DRS imbalance information?

I did a quick search on the forums here and only found some folks asking for CPU/MEM/Disk stats.

I'm reading through the vSphere 6.7 Clustering Deep Dive and I'm going through the DRS section.

In the past, I wanted to find a way to collect DRS information to see what it is doing.  Not reservations, or limits, but more around entitlements.  A cluster may be imbalanced but unable to satisfy that imbalance because of "XYZ".    I could log into the vCenter client and hit the Cluster/DRS window to see recommendations (if any exist) but that does not give me all of the data.

Is there anyway, via API, PowerCLI, SSH, that can give me "Normalized Entitlement" or CHLSD-Current Host Load Standard Deviation for each host in a cluster?

Maybe this is bubbled up into vRops?

I understand, we should not care about this information and should just let DRS take care of things.  But, I love data, and being able to see an imbalance, even if slight, with trending analysis would be interesting.  Knowing I can adjust the slider to be more aggressive and it would balance VMs without a mass vmotion storm every evening would be great.  Even more so, being able to create automation around imbalance metrics (cross cluster) would be amazing...


Tags (2)
0 Kudos
1 Reply

I did a try to get the Normalized Entitlement in the past, but rather unsuccessful I have to admit :smileycry:

In the VMware Distributed Resource Management: Design, Implementation, and Lessons Learned paper, section 3.1.1, there is a mention of how DRS calculates the normalised entitlement values.
But there is not enough info, at least that I found, to provide further details.

And from the Deep Dive book(s) it should be clear that there are many advanced settings that seem to influence the calculations.

Plus, details in the calculations seem to change between releases.

The best I came up with was to calculate numbers that were more or less following the same ratio as the numbers DRS shows over time.
But the actual values the DRS Summary shows was way beyond what I was able to calculate.

All in all, I assume that the actual calculations DRS uses are to be considered the crown jewels, which one does not simply leave at the front door :smileygrin:

I do wish that there will be one day metrics that show/collect these DRS numbers.

That would at least allow us to look at the DRS history, see critical points in time and perhaps extrapolate to somewhat predict the (near) future of clusters.

Blog: lucd.info  Twitter: @LucD22  Co-author PowerCLI Reference

0 Kudos