FIELD NOTES ON VIRTUALIZATION, STORAGE AND INFRASTRUCTURE

Written from production, not from datasheets.

Three hundred and fourteen posts on vSphere internals, storage latency, automation and the failure modes that only show up at three in the morning. If a walkthrough here says something works, it is because it was run, broken, and run again.

tapas@vmwareblogs:~$ site --summary

  author     Tapas Kumar Mahanta, Solutions Architect
  focus      vSphere internals, storage, automation, infrastructure
  posts      314
  subjects   6
  linkedin   /in/tapas-kumar-mahanta
  method     reproduce it, break it, then write it down
  promise    every command in every post has been run

tapas@vmwareblogs:~$ _

What you will find here

Six subject areas, built up over nine years of support escalations, lab builds and customer engagements.

Latest posts

• Designing vSphere for scale: where vCenter becomes the bottleneck, and how to prove it before you get thereAugust 16, 2026
  Configuration maximums tell you what is supported, not what performs. Property Collector architecture, SSO topology failure modes, evacuation arithmetic and a scale probe that measures your own ceiling.
• vSphere cluster design: the sizing arithmetic, and four reference builds that follow from itAugust 16, 2026
  Host counts are derived, not chosen. The admission control maths, failure domain reasoning and four worked designs: edge, general purpose production, stretched, and GPU.
• RWX volumes on VMstore: the CSI call sequence, and the five places it strandsAugust 16, 2026
  RWX staging is reference counted per node, not per pod. That single difference from RWO explains stuck ContainerCreating, orphaned leases, ESTALE and pods that will not terminate.
• VCF certificate rotation: the trust chain, the four ways it strands, and how to recoverAugust 16, 2026
  SDDC Manager holds one view of the trust chain and each component holds another. When rotation fails halfway they disagree, and the UI stops working. Detection and recovery.
• Correlating vSphere and VMstore latency in one table: the triage scriptAugust 16, 2026
  Guest, host and array latency for the same VM in the same window. A PowerShell and Python join across the vSphere and Tintri APIs that names the guilty layer.
• Tintri per-VM latency: why one sample lies, and how to measure it properlyAugust 16, 2026
  Storage latency is heavy tailed, so the mean is close to meaningless. Percentiles, sustained-breach detection with hysteresis, integrated component attribution and noisy-neighbour correlation.

Three worth starting with

• Wiring Grafana into vCenter
  A telemetry pipeline from vSphere performance counter to rendered pixel, and the ten design decisions that decide whether it survives a production estate.
• Building VMware Cloud Foundation 9 on Tintri NFS storage
  VCF on NFS, end to end, including the parts the deployment guide leaves out.
• ESXi Log Analyser
  Turning an ESXi support bundle into an answer instead of an afternoon.

About me

I am Tapas Kumar Mahanta, a Solutions Architect working on storage and virtualization infrastructure, based in Bangalore. Most of what appears on this site starts as a customer escalation, a lab build or a proof of concept that did not behave the way the documentation said it would.

The day job covers vSphere and ESXi at scale, Tintri and vSAN storage, Cisco UCS, Kubernetes and CSI driver development, and OpenStack and OpenShift integration work. The writing here is the residue of that: the specific thing that was wrong, how it was found, and what to do about it.

Nothing here is vendor approved and none of it is official guidance. It is what worked, written down while it was still fresh. Corrections and disagreements are welcome, and more useful to me than agreement.

Get in touch

• LinkedIn  linkedin.com/in/tapas-kumar-mahanta
• Email  tmahanta@vmwareblogs.co.in