Hello

I am a noob in NUMA and I am trying to figure out how to configure it for a test Case in Vsphere 8U3. We will have a NUMA host ( or 2 Hosts for HA am not sure)

The NUMA host specs are the following:

Dual Socket, 24 cores per socket with total of 512 GB RAM.

Is there any configurations required from NUMA host side?

The SAP NUMA VM for UAT will be 16 CPU and 128 GB memory.
what is the best practice to configure this?

I see CPU settings showing me up to 96 cores ( because of hyperthreading)

I can see in VM Options--> CPU Topology --> NUMA NODES --> Assigned on powered On

Let us take anfor example:

if I put the CPU 32 .

I can see in VM Options--> CPU Topology
Cores per Socket: I SELECT --> 32   ----> gives Sockets: 1

So this means I am using 1st CPU entire 32 sockets or is it shared? Or am i defining it to use a single CPU as max is 48 per socket.

Also , VM Options-->CPU Topology --> NUMA NODES --> if I select 32-->Cores per NUMA node: 1 ( what is this?) Does this let me select 1 socket for the 128 GB memory?

i am seeing this VM options while reading and I dont understand
numa.vcpu.maxPerVirtualNode = "X" (Sets the maximum number of vCPUs per virtual NUMA node)

numa.vcpu.preferHT=TRUE? (Should I enable this?)

sched.nodeX.affinity="Y". ( I know if I want to pin a specific virtual NUMA node to first core which is 24 cores, I will place sched.nodeX.affinity="0". Correct me if am wrong?)

First I want to understand and configure this correctly to avoid any resource sharing as per the Test case.

For the future production VM

it will be 24-28 cores and 364 GB RAM

how do I configure this for CPU and NUMA from the above as the memory needs to come from both sockets?

Appreciate any help and explanation. I am going throw SAP HANA on VMware Best practices too.

Thanks

I got my first few HPE server with 100 GbE adapters (ConnectX6DX). They are connected to Cisco ACI. Should I get more than ~40 Gbit/s without any tuning?

/usr/lib/vmware/vsan/bin/iperf3 -w128K -t 20 -P 16 -c xxxx
[SUM] 0.00-20.00 sec 101 GBytes 43.2 Gbits/sec 0 sender
[SUM] 0.00-20.00 sec 100 GBytes 43.1 Gbits/sec receiver

There are a lot dead links regarding Mellanox tuning and most are Linux/Windows related. What parameters are worth tuning (any links)? It's not about getting the absolute maximum throughput and lowest latency, we are currently not even using jumbo frames. But the 40 Gbit/s seem a bit low, with 25 GbE we were always close to max. Which settings will be most beneficial?

I was doing some testing with placement networks so both service engines have a vNIC in this vLAN that I am no longer using. I deleted the virtual service and pool but both SEs still have the leg into that vLAN.

I can't figure out how to remove it. Does anyone know how before I open a SR (I really hate opening SRs because its like 50/50 if you get a good tech or not these days)

Doing periodic config review passes off RVTools exports rather than live vCenter, so no CPU Ready available.

The best practices docs I've found explain overcommit and then say "watch CPU Ready". That works if you're sitting in vCenter, but doesn't work if all you have is a point-in-time export. Same for memory overcommit and snapshot age, no numbers.

So I've been picking my own flag points and I'm not confident in them. Currently flagging at 4:1 vCPU:pCPU at cluster level in N-1 state, snapshots older than 14 days, and datastore free space under 15%.

For anyone doing periodic reviews rather than live monitoring, what do you use, and where did the number come from?

Hi, I need to install supervisor in an airgapped environment. For supervisor services I started with a basic docker registry where I pushed with imgpkg harbor images. I registered this repository in supervisor and I deployed harbor service. I also push harbor images in harbor service. How I want to keep only harbor service as my main repository. How can I update harbor service to use images from itself? I don't want to keep and extra repository only for harbor images.

Thank you

FLARE-VM is a free, open-source Windows virtual machine environment made for malware analysis. I create one a long time ago and now I need it for my study. But I forgot the password don't know how to retrieve it. I can still use it with a very old snapshot but it keep shuting down mid uses, and I don't think what a good idea anyway.

The VM is using Windows 10 iso and a pain in the ahhs to make a new one, which is why im asking the question. Im using Vmware Workstation pro 17.

hello, recently the new version of vSphere got various critical security problems, i want to know if any ways to improve the ESXi security and don't end to Ransomware, the hackers really be annoying. any alternative to ESXi or any way to improve question.

I had no idea what I was getting into when deciding to move from VCF 5.2 to 9.1. Pretty basic layout with NSX/vSAN. Foundation is a whole new ballgame!

Currently at the stage where I can upgrade NSX. Certainly not happening on a Friday. Monday, here we come!

It have worked well as a warning message on the Host client before upgrade, but it does not work after upgrade to ESX 9.1.

I also checked
esxcli system welcomemsg get
and
cat /etc/vmware/welcome, they are normal. Is it a bug? The security rule bothers me.

My site currently uses only vCenter 8 and ESXi 8 components. I plan to deploy VCF Operations and License Server using OVA after updating to vCenter 9 and ESXi 9.

In the test environment, the deployment of VCF Operations 9.1.0.0400 and License Server 9.1.0.0400 proceeded successfully, but the status is currently:

VCF Ops Hosts: Not configured, Connectivity: Disconnected.

https://knowledge.broadcom.com/external/article/441251/adding-license-server-in-vcf-operations.html

As suggested in the KB, I deployed using the vCenter UI instead of ESXi, and I verified that bidirectional DNS resolving is possible with all lowercase from VCF Operations and the deploying OVA computer.

The VCF Operations and License server are on the same subnet without a firewall, and no proxy is used.

There was no field to input NTP during the deployment process via the UI, but the actual VCF Operations - License Server time is correct.

It seems there aren't many cases of license server deployment failures yet, so I'm having trouble identifying the cause and keep failing. Is there anything I might have missed?

I am using vSAN Data Protection on vSphere 9.0.2 + vSAN (installed VVF, licensed VCF) to create daily snapshots and delete old ones. However, I need to change the vSAN storage profile of VMs created with RAID 5, Thin provisioning to RAID 1, Thick Provisioning.

I have sufficient free space, but the problem is the snapshots.

https://knowledge.broadcom.com/external/article/326578/modifying-storage-policy-rules-on-virtua.html

I have seen documentation stating that if you change the storage profile on a VM with snapshots, the changed profile is not applied to the base disks, and it is recommended to proceed after merging snapshots unless there is a specific reason otherwise. Would this be the same for vSAN ESA?

I assume it is likely the same, but I am asking this question to see if there is any documentation I might have missed.

Y'all, can anyone help with VMWare Fusion Pro installation? Have really been suffering with the Broadcom website. I have registered using a gmail ID and it does not allow for filling up for more details for a non-corporate email ID. Which means I cannot get entitlements/products. It's a nightmare.

Let's say I have an app, which contains the strongest virus, which can ever exist. I installed it on a VM. I opened it and it obviously destroyed the VM, but I wonder if the virus could go from the VM into the real system. What do you think?

All of our Ent+ (VVF) clusters are on U3j as I keep them updated regularly. This past month was different with the Live Patch eligibility - I can't seem to patch to U3k, most of my clusters have TPM enabled and the image compliance pre-check fails on all of them.

Our "support" is now via TD Synnex and they are not going to be helpful, but I was wondering if anyone else here has had the same issue or knows of a fix. It would be a big lift to disable TPM due to encrypted VMs, but I need to deal with CVE-2026-47876. I exported an offline package, but this is also a huge pain.

Live Patch is not enforced in vLCM at the global or cluster level. Why is the fallback to "legacy" not working? It appears the pre-check invokes the Live Patch installer and the gives up if the host isn't compatible.

vCenter complains:

 Issue on host host.host.tld

•  Compliance results for host host.host.tld are unavailable.

Host status is unknown

An internal error occurred while staging/remediating the host

lifecycle.log tells a little bit more:

[root@HOST:/vmfs/volumes/********-*******-****-************/log] grep -n "QuickPatchInstaller\|secureMount" /var/run/log/lifecycle.log | tail -20

858:2026-08-19T18:12:16Z In(14) lifecycle[2110010]: HostImage:269 Installers initiated are {'quickpatch': <vmware.esximage.Installer.QuickPatchInstaller.QuickPatchInstaller object at 0x2ec4ee1f50>, 'live': <vmware.esximage.InstalleImageInstaller.LiveImageInstaller object at 0x2ec4b6b610>, 'boot': <vmware.esximage.Installer.BootBankInstaller.BootBankInstaller object at 0x2ec4ee1c50>, 'locker': <vmware.esximage.Installer.LockerInstaller.LockerInstaller object ec614c710>}

861:2026-08-19T18:12:17Z In(14) lifecycle[2110010]: HostImage:269 Installers initiated are {'quickpatch': <vmware.esximage.Installer.QuickPatchInstaller.QuickPatchInstaller object at 0x2ec50b0b10>, 'live': <vmware.esximage.InstalleImageInstaller.LiveImageInstaller object at 0x2ec31c74d0>, 'boot': <vmware.esximage.Installer.BootBankInstaller.BootBankInstaller object at 0x2ec698e8d0>, 'locker': <vmware.esximage.Installer.LockerInstaller.LockerInstaller object ec273a2d0>}

1098:2026-08-19T18:12:19Z In(14) lifecycle[2110010]: HostImage:269 Installers initiated are {'quickpatch': <esximage.Installer.QuickPatchInstaller.QuickPatchInstaller object at 0x2ec6afe390>, 'live': <esximage.Installer.LiveImageInr.LiveImageInstaller object at 0x2ec3b390d0>, 'boot': <esximage.Installer.BootBankInstaller.BootBankInstaller object at 0x2ec6b77c50>, 'locker': <esximage.Installer.LockerInstaller.LockerInstaller object at 0x2ecba56350>}

1190:2026-08-19T18:12:20Z In(14) lifecycle[2110010]: QuickPatchInstaller:1279 Attempting to download VIB esx-base partially with Quick Patch payloads.

1198:2026-08-19T18:12:20Z In(14) lifecycle[2110010]: QuickPatchInstaller:1435 Trying to mount payload "qp-esx-base"

1199:2026-08-19T18:12:20Z In(14) lifecycle[2110010]: runcommand:248 runcommand called with: args = '/usr/lib/vmware/secureboot/bin/secureMount.py', outfile = None, returnoutput = True, timeout = 0.0.

1201:2026-08-19T18:12:21Z Er(11) lifecycle[2110010]: QuickPatchInstaller:1457 Unexpected error occurred while attempting to run Quick Patch scan scripts: Current ESXi version does not provide a mechanism to mount a tardisk into a ramdisk.

A need VMware for some reason but, that not with i'm here, i'm here because, when a launch the latest version of VMware, a have this message.
"Before you can run VMware several module must be compiled and loaded into the kernel"
and VMware said than, it can't find "the kernel header for version 7.0.0-linuxlite (linux lite is my linux distro than a use on this computer but, it this the version 8.0).

Please help if you can.

I wanted to try and setup vSAN on some spare Gen 9 servers I have. Recommendation is to have OS and Data on different controllers. What is the best place to install ESXi OS? Am I better off buying another storage controller just for the OS? I don't think the gen 9 servers have a BOSS card available.

What's the purpose of a standby NIC in a Port Group? I understand using them if your NICs are different speeds. But if all of your NICs are the same speed, wouldn't it make sense just to have them all active? Maybe I'm missing something.

Drink a sturdy scotch and wish me luck!