From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate001.proxmox.com (gate001.proxmox.com [45.144.208.40]) by lore.proxmox.com (Postfix) with ESMTPS id A494C1FF0AF for ; Fri, 20 Nov 2026 19:51:18 +0100 (CET) Received: from gate001.proxmox.com (localhost.localdomain [127.0.0.1]) by gate001.proxmox.com (Proxmox) with ESMTP id 79A81216A5; Fri, 20 Nov 2026 19:51:16 +0100 (CET) Date: Fri, 20 Nov 2026 19:51:10 +0100 From: Gabriel Goller To: Thomas Lamprecht Subject: Re: [PATCH common/manager/network/proxmox{-ve-rs,-perl-rs,-widget-toolkit} v3 00/28] SDN VRF support Message-ID: References: <20261119194116.575520-1-g.goller@proxmox.com> <179515721902.2281522.12185275346766103866.b4-ty@b4> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <179515721902.2281522.12185275346766103866.b4-ty@b4> User-Agent: NeoMutt/20260616 X-Bm-Milter-Handled: 55990f41-d878-4baa-be0a-ee34c49e34d2 X-Bm-Transport-Timestamp: 1795200671708 X-SPAM-LEVEL: Spam detection results: 0 AWL 0.214 Adjusted score from AWL reputation of From: address DMARC_MISSING 0.1 Missing DMARC policy KAM_DMARC_STATUS 0.01 Test Rule for DKIM or SPF Failure with Strict Alignment (newer systems) RCVD_IN_DNSWL_MED -2.3 Sender listed at https://www.dnswl.org/, medium trust SPF_HELO_NONE 0.001 SPF: HELO does not publish an SPF Record SPF_PASS -0.001 SPF: sender matches SPF record Message-ID-Hash: WSE26M73C6VWQSB7DKJ5JINN2UZ42I5F X-Message-ID-Hash: WSE26M73C6VWQSB7DKJ5JINN2UZ42I5F X-MailFrom: g.goller@proxmox.com X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header CC: pve-devel@lists.proxmox.com X-Mailman-Version: 3.3.10 Precedence: list List-Id: Proxmox VE development discussion List-Help: List-Owner: List-Post: List-Subscribe: List-Unsubscribe: On 20.11.2026 08:42, Thomas Lamprecht wrote: > On Thu, 19 Nov 2026 20:40:46 +0100, Gabriel Goller wrote: > > This series adds VRF support to SDN zones and fabrics. > > > > A simple zone can now use either the default VRF or a dedicated VRF. The > > simple zone dialog has a checkbox for this and uses the default VRF by > > default. If the checkbox is disabled, the zone creates a dedicated VRF named > > `vrf_`. EVPN zones always create a dedicated VRF. BGP and OSPF > > fabrics can select a zone and run inside its VRF. > > > > [...] > > Applied, thanks! > > Tested on a three-node cluster with two VRF zones that use the same > subnet, with SNAT, DHCP and a default route via a VLAN uplink. A setup > like the existing ones still produces the same config, apart from the > new IPv6 unreachable route in EVPN VRFs. > > What I changed in your patches: > - Static routes now require a dedicated VRF as without one, the routes > went into the main routing table of the host, and unchecking the VRF > moved existing routes there. I went that way as this we can still > relax easily later without breaking existing configurations. If routes > in the main table are wanted, a separate SDN object for static routes > would be the cleaner way. > - The permission check for routes skipped zones with an empty node list, > which means all nodes. It now checks all nodes in that case. > - Several OSPF fabrics in one VRF could set different options for the > same kind of redistributed routes, so the FRR config flipped back and > forth on every apply. Now the first fabric wins. In the default VRF, > this also means that the other fabrics' redistributions apply now, > not just the first one's. > - Since your series sets up DHCP after ifreload, one failing interface > skipped the DHCP config of all zones, and a DHCP error skipped the FRR > config. Now each step runs even if the other one fails. > - A VRF on an OVS bridge or internal port was rejected, which blocked > all network changes on a node with such a manual config. Only OVS > ports and bonds are rejected now, and the UI matches that. > - The dnsmasq override files for the VRF were removed before stopping > the service, so systemd warned on every apply. They are now removed > after stopping it. > - I also did some rewording and formatting. > > What I added on top: > - ifupdown2 does not enable IPv4 forwarding on VLAN interfaces by > default, which broke VLAN uplinks for the zone. So I added an explicit > forwarding option to the network API and UI. This way an admin can > turn it on without editing the config by hand, and we don't have to > enable and disable it automatically. > - I reworked the VRF docs to walk through setting up an uplink, and > documented the new options. Thanks for all these fixes, had a quick look at them and they looked good. > Still open: > - Host interfaces can still point to the VRF of a zone after the zone or > its VRF is removed. A warning on the node would help. https://lore.proxmox.com/pve-devel/20261120183116.478378-1-g.goller@proxmox.com/ > - A VRF set on a fabric interface makes the whole SDN network config of > the node fail. A warning might be enough. https://lore.proxmox.com/pve-devel/20261120185037.493949-1-g.goller@proxmox.com/ > - Conflicting OSPF settings between fabrics could be rejected in the API > instead of the first fabric silently winning. Hmm this is not so easily fixable without making a breaking change. Prepared a series, but this is for something after the release and will take a while to get right. > - If an OSPF or BGP fabric in a VRF uses a WireGuard interface, the > WireGuard peer routes still go into the main routing table. https://lore.proxmox.com/pve-devel/20261120152657.275504-1-g.goller@proxmox.com/ > - Removing all static routes and unchecking the VRF in one go fails in > the UI. Doing it in two steps works. https://lore.proxmox.com/pve-devel/20261120143944.219256-1-g.goller@proxmox.com/ > > [snip]