# Vates / XCP-ng 日跟踪 · 2026-09-03
> 日期：2026-09-03（Asia/Shanghai）。原料：archive 原文，非旧日报。

## 摘要
- **XO 原生 MCP**（自 XO 6.2）：一等公民 `@xen-orchestra/mcp`，可用自然语言问基础设施，可完全 on-prem。
- **XO 6.8**：相关 VM 同步快照、XO 6 主机管理增强、XO Lite 建网、**官方 Veeam 支持 XCP-ng**。
- **XCP-ng 8.3 LTS 2026-08 维护包**：需重启；累计更新。
- **V2V 文档**：XO 5.110+ 用 nbdkit + VDDK 替代旧 NFS 路径。
- 看过：定价四档（Essential → Enterprise）、GPU/PCI passthrough 文档、NetBox 插件、RISC-V 实验博文。

## 原文摘录

### [MCP meets Xen Orchestra](https://xen-orchestra.com/blog/mcp-meets-xen-orchestra/)
- 来源：Xen Orchestra 博客 / `vates/mcp-meets-xen-orchestra-45f87986.txt` / 抓取 2026-09-03T00:44:06Z
> Talk to your infrastructure in plain language, get instant answers, and keep everything on-prem if you want. With XO 6.2, we shipped something we've been excited about for a while: native support for the Model Context Protocol (MCP). Not a third-party plugin, not a community experiment: @xen-orchestra/mcp is a first-party module maintained by the Vates team as part of the Xen Orchestra codebase. It's a small package, but it opens up a fundamentally new way of interacting with your virtualization stack: instead of clicking through dashboards or writing API calls, you can now ask questions in plain English (or French, or any language your AI assistant understands) and get instant, accurate answers drawn directly from your live infrastructure. And if you want to go further (like running the AI model itself on your own GPUs, inside your own VMs, with zero external dependency) you can do that too. We'll get to that! Let's walk through what this looks like in practice. MCP is an open standard, originally introduced by Anthropic and now governed by the Linux Foundation's Agentic AI…

### [XO 6.8](https://xen-orchestra.com/blog/xen-orchestra-6-8/)
- 来源：XO 发布说明（文件名 xen-orchestra-6-8） / `vates/xen-orchestra-6-8-529be7cd.txt` / 抓取 2026-09-03T00:43:00Z 左右
> XO 6.8 is out: synchronized snapshots for related VMs, more host management in XO 6, network creation in XO Lite, plus official Veeam support for XCP-ng. This month is about what surrounds a migration rather than the migration itself: the backup tooling you already own, the hardware you can avoid buying, and the numbers you need before the decision gets made. Alongside Xen Orchestra 6.8, Veeam Backup & Replication now officially supports XCP-ng, TwinStor opened as a Technology Preview for two-host pools, and we published a calculator that puts a Vates VMS environment side by side with its VMware equivalent. On the product side, Backup gains synchronized snapshots for VMs that belong to the same application, XO 6 keeps closing the operations that still required a return to XO 5, and XO Lite can now create networks instead of only displaying them. Add XOA sizing its own memory, a declarative way to share a VM from Terraform or Go, and a round of documentation cleanup, and 6.8 is a release with as much happening around Xen Orchestra as inside it. As usual, this announcement is…

### [August 2026 Updates #1 for XCP-ng 8.3 LTS](https://xcp-ng.org/blog/2026/08/18/august-2026-updates-1-for-xcp-ng-8-3-lts/)
- 来源：XCP-ng 博客 / `vates/august-2026-updates-1-for-xcp-ng-8-3-lts-b5d15e15.txt` / 抓取 2026-09-03T00:36:13Z
> New maintenance updates are available for XCP-ng 8.3 LTS. This is one of these batches of updates we group together to limit the number of maintenance operations on XCP-ng hosts. They are ready to be applied to your pools. Let me introduce the changes to you. But first, a word about the update process. If your update habits, based on the fact that it "usually works", are to first update all hosts then reboot them one by one, you'll have a bad surprise, because in this case VMs will refuse to migrate, due to a temporary inconsistency between software expectations and local state. This is expected behaviour, and is why the documented process (implemented in Xen Orchestra as the Rolling Pool Update feature) is to update hosts one by one, starting with the pool master. Disable, evacuate, update, reboot, next host. This guarantees that software and local state always agree. Also note that if LINSTOR storage is involved, any manual update process involves additional steps. Reading this too late? Fortunately, restarting the toolstack on every host (xe-toolstack-restart) should be enough to…

### [VMware migration (V2V) | Xen Orchestra](https://docs.xen-orchestra.com/xo5/v2v-migration-guide)
- 来源：XO 文档 / `vates/vmware-migration-v2v-…-9ff4d03e.txt` / 抓取 2026-09-03T00:44:21Z
> Xen Orchestra 5.110 introduced a new migration backend for moving virtual machines from VMware to XCP-ng (V2V). It replaces the old NFS-based method with nbdkit and VMware's VDDK libraries, delivering better performance and reliability. The whole process is driven from Xen Orchestra, in Import → VM → From VMware. XO connects to your vCenter or ESXi host, reads the VM disks directly, and imports them into an XCP-ng storage repository: This guide walks you through the entire migration process: You will also find troubleshooting advice and best practices to maximize speed and minimize downtime. The V2V import is currently available in the XO 5 interface. The new XO 6 web UI does not expose it yet: even if you use XO 6 daily, switch to the XO 5 interface for the migration itself. The new V2V backend replaces the old approach, and delivers: The new backend only reads allocated blocks, which accelerates transfers, and snapshots have minimal impact on performance. There is no longer any need for temporary NFS storage or remote VSAN targets. Very large disks are handled natively: disks…

## 对基础档案的影响
- **AI/Agent**：XO MCP 是开源虚拟化里少见的「官方 MCP 运维面」，对标 Nutanix Agent Gateway / VMware Tanzu agent 层，但是运维助手而非推理网关。
- **备份生态**：XO 6.8 官方 Veeam for XCP-ng——企业替代的关键缺口。
- **版本**：XCP-ng 8.3 LTS 仍在维护；XO 演进到 6.8。
- **迁移**：VDDK/nbdkit V2V 路径写入工具链。