Security Advisories Latest four
There's no new VMSA today, so these are the four most recently released or updated advisories; Broadcom's advisory portal has the rest. If vCenter isn't yet patched for CVE-2026-59310 in VMSA-2026-0006, that's still the priority, and VMSA-2026-0008.1 still lists "Patch pending" for every SDDC Manager version. On the wider industry front, Broadcom hasn't yet published a statement on Java4Shell (CVE-2026-114938), the OpenJDK Base64 flaw Oracle patched on 17 November; its existing KB 434436 on OpenJDK findings in vCenter 8.x doesn't mention it. There's also still no statement on the Apache HTTP/2 flaw CVE-2026-119618 (KB 443715 covers only earlier Apache CVEs on vCenter 8.x), or on OpenSSH "SSHole" CVE-2026-123158. KB 459989 still says there are "zero affected products in the VMware Cloud Foundation product line" for PHP2Shell, and for the Linux TCP flaw CVE-2026-98432 the only statements remain KB 461242 for Tanzu and Bitnami images and KB 461249 for vDefend.
Node.js, as used in SDDC Manager, has a Brotli decoder out-of-bounds write that an attacker with network access to SDDC Manager could use to run code. It is critical (9.8) on SDDC Manager 9.1.1.x and important (7.5) on 9.0.x, 9.1.0 and 5.2.x; revision .1 on 28 Oct clarified the impacted products in the response matrix. Every fixed version is still "Patch pending", and SDDC Manager 9.1.1.0200 (12 Oct) is still the newest 9.1.1 patch on TechDocs. Broadcom's portfolio statement KB 460791 says that, apart from SDDC Manager in VCF 9.1.1, none of its products are critically affected. Press reports say CISA added the underlying Brotli flaw to its KEV catalogue on 27 Oct; the Broadcom advisory itself makes no statement on exploitation. Issue date 27 Oct 2026.
- CVE-2026-111497 9.8 SDDC Manager Node.js Brotli · 9.8 on 9.1.1, 7.5 elsewhere · reported KEV 27 Oct
DoApply KB 460732 on 9.1.1 or KB 460734 on 5.2.x, 9.0.x and 9.1.0, keep SDDC Manager reachable only from trusted admin hosts, and subscribe to the advisory so you see the patch the day it lands.
For desktop hypervisors: Workstation and Fusion 25H2 and 26H1, CVSSv3 8.1 to 9.3. Both CVEs need local admin rights inside a guest and can lead to code execution on the host. There are no workarounds and the advisory makes no statement on exploitation.
- CVE-2026-59346 9.3 VMXNET3 integer overflow
- CVE-2026-59347 8.1 HGFS stack overflow
DoUpdate Workstation and Fusion to 26H1u1, especially if you run nested labs on a laptop.
Still the one to prioritise. It covers vCenter, ESX, Workstation and Fusion (and so VCF, VVF and Telco Cloud), CVSSv3 2.7 to 9.8, with no workarounds. CISA added vCenter CVE-2026-59310 to its KEV catalogue on 18 Aug 2026 with known ransomware use. SecurityWeek [Press] has reported that CISA added the ESX logging flaw CVE-2026-41709 on 28 Oct. The Broadcom advisory itself makes no statement on exploitation. First published 29 Jul 2026; revision .2 on 19 Aug added 7.0 guidance for the critical issues.
- CVE-2026-59309 9.8 vCenter VMDir auth bypass
- CVE-2026-59310 9.8 vCenter Syslog traversal · KEV
- CVE-2026-47876 9.3 ESX VMXNET3
- CVE-2026-41703 7.6 ESX OOB read
- CVE-2026-41709 2.7 ESX logging · reported KEV 28 Oct
DoPatch vCenter to 9.1.0.0300, 9.0.2.0100 or 8.0 U3k (8.0 U2f is also listed; VCF 5.x: async patch to 8.0 U3k per KB 88287). On ESX 9.0, ESXi-9.0.2.0100-25595025 covers all three ESX CVEs. On 9.1 the matrix lists ESXi-9.1.0.0200-25557999 for CVE-2026-47876 and ESXi-9.1.0.0-25370933 for the other two. On 8.0, ESXi80U3k-25595708 fixes CVE-2026-47876, U3i-25205845 CVE-2026-41703 and U3j-25429389 CVE-2026-41709. vCenter 7.0 fixes need an extended support contract.
If you run Avi Load Balancer (including for VKS or VCF Automation), this one matters. Seven CVEs, CVSSv3 7.1 to 9.8, no workarounds; the worst, CVE-2026-47865, is an authentication bypass. The advisory makes no statement on exploitation. First published 14 Jul 2026; revision .1 on 7 Aug added the 22.1.7-2p12 patch.
- CVE-2026-47865 9.8 auth bypass
- CVE-2026-47871 8.8 directory traversal
- CVE-2026-47867 8.7 RCE
- CVE-2026-47869 8.7 authenticated RCE
- CVE-2026-47866 8.3 authorisation bypass
- CVE-2026-47868 7.8 local privilege escalation
- CVE-2026-47870 7.1 privilege escalation
DoUpgrade 32.1.1 to 32.1.2 or later (the advisory's recommended version), 31.1.1–31.2.2 to 31.2.2-2p3, and 30.1.1–30.2.6 to 30.2.7. On 22.1.x, 22.1.7-2p12 fixes CVE-2026-47865, but the matrix lists 30.2.7 for the other six.
All VMware security advisories on Broadcom's support portal →
Deep Dive of the Day vSAN storage clusters and compute clusters in VCF 9.1: what changed since 9.0, and the choice you make at creation
Most VCF domains start life as vSAN HCI clusters, where every host gives both compute and storage. The VCF 9.1 storage models page adds a disaggregated pair: a vSAN storage cluster that only serves storage (what used to be called vSAN Max, as Pranay Jha notes), and a vSAN compute cluster that has no local vSAN storage and mounts a datastore from the storage cluster instead. The point is to let compute and storage grow on their own timelines, so a CPU-hungry workload domain doesn't force you to buy capacity drives, and vice versa.
There are firm rules about where they can go. Storage clusters need vSAN ESA, so NVMe drives only, and Broadcom doesn't recommend running VMs on them. Neither type can be the management domain's default cluster; both are supported for additional management clusters and workload domains. A compute cluster must sit in the same workload domain as its storage cluster, and a single-rack compute cluster can be stretched across two availability zones while a multi-rack compute cluster supports a single zone only.
VCF 9.1 relaxed some limits from 9.0. The 9.0 sharing rules say OSA and ESA can't share datastores, a storage cluster can't be mounted by an OSA client, and data-in-transit encryption between client and server clusters isn't supported. The 9.1 what's new page says compute-only clusters can now mount OSA and storage-cluster datastores at the same time, stretched storage clusters can serve storage across vCenter boundaries, and disaggregated vSAN gains in-transit encryption between client and storage clusters. If you read a 9.0-era design guide, check it against 9.1 before you rule anything out.
Some numbers don't change. Per the sharing rules, a datastore can be shared with up to 10 client clusters, a client cluster can mount up to 5 remote datastores, and no more than 128 hosts in total can mount one datastore. A storage cluster needs at least 4 hosts, and Broadcom recommends 6 or more; its design guidance says 7 or more lets the cluster get its resilience back after a sustained host failure and supports RAID-6. Client-to-server latency must stay under 5 ms. On bandwidth, 25 GbE between storage hosts is supported and 100 GbE recommended (smaller storage clusters may use 10 GbE), while clients can connect at 10 GbE with 25 GbE recommended.
The decision you can't undo is the cluster type. The sharing rules say it's set when the cluster is created, and turning an HCI cluster into a storage cluster means turning vSAN off and reconfiguring it. Two other traps: set vSphere HA's APD response on client clusters to power off and restart VMs, and remember that storage policies are per vCenter, so when client and storage clusters sit under different vCenters it's the policy on the client cluster's vCenter that applies. Pranay Jha's design notes add a sensible sizing habit: plan on usable capacity, not raw, and start with four devices per host rather than the two-device minimum, so a drive failure leaves room to rebuild locally.
For a nested lab, a small storage cluster plus a compute cluster is a good way to learn the mount, policy and failure behaviour without much hardware. Just remember that a 4-host minimum, ESA and NVMe-only devices still apply, so nested NVMe controllers and the usual HCL workarounds come into play, and that what you see in a lab says nothing about performance.