From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from firstgate.proxmox.com (firstgate.proxmox.com [212.224.123.68]) by lore.proxmox.com (Postfix) with ESMTPS id 7ACDB1FF13C for ; Thu, 02 Apr 2026 16:24:13 +0200 (CEST) Received: from firstgate.proxmox.com (localhost [127.0.0.1]) by firstgate.proxmox.com (Proxmox) with ESMTP id C281D1C5EB; Thu, 2 Apr 2026 16:24:42 +0200 (CEST) Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=UTF-8 Date: Thu, 02 Apr 2026 16:24:09 +0200 Message-Id: Subject: Re: [PATCH cluster/ha-manager/manager v4 00/28] dynamic scheduler + load rebalancer From: "Dominik Rusovac" To: "Daniel Kral" , X-Mailer: aerc 0.20.0 References: <20260402124817.416232-1-d.kral@proxmox.com> In-Reply-To: <20260402124817.416232-1-d.kral@proxmox.com> X-Bm-Milter-Handled: 55990f41-d878-4baa-be0a-ee34c49e34d2 X-Bm-Transport-Timestamp: 1775139790607 X-SPAM-LEVEL: Spam detection results: 0 AWL -1.007 Adjusted score from AWL reputation of From: address BAYES_00 -1.9 Bayes spam probability is 0 to 1% DMARC_MISSING 0.1 Missing DMARC policy KAM_DMARC_STATUS 0.01 Test Rule for DKIM or SPF Failure with Strict Alignment RCVD_IN_VALIDITY_CERTIFIED_BLOCKED 1 ADMINISTRATOR NOTICE: The query to Validity was blocked. See https://knowledge.validity.com/hc/en-us/articles/20961730681243 for more information. RCVD_IN_VALIDITY_RPBL_BLOCKED 1 ADMINISTRATOR NOTICE: The query to Validity was blocked. See https://knowledge.validity.com/hc/en-us/articles/20961730681243 for more information. RCVD_IN_VALIDITY_SAFE_BLOCKED 1 ADMINISTRATOR NOTICE: The query to Validity was blocked. See https://knowledge.validity.com/hc/en-us/articles/20961730681243 for more information. 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: QWSRBKD4SS2H5YESCMOHKMIB6MOQ45VH X-Message-ID-Hash: QWSRBKD4SS2H5YESCMOHKMIB6MOQ45VH X-MailFrom: d.rusovac@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 X-Mailman-Version: 3.3.10 Precedence: list List-Id: Proxmox VE development discussion List-Help: List-Owner: List-Post: List-Subscribe: List-Unsubscribe: Reviewed all of the patches from v1 up to v4. Tested the behavior=20 of the CRS in a 3-node-cluster and in a 7-node-cluster regarding: * disarmed HA and re-armed HA * maintenance mode of nodes * fenced nodes * affinity rules Moreover: * used all of the variations (static or dynamic with bruteforce or topsis) * played around with a bunch of different thresholds, margins and hold durations for the purpose of fine tuning the scheduler * verified that hostnames can include hyphens, for example * verified that minimum requirements for number fields are detected * used UI for setting different auto-rebalance parameters Observations: * scoring of best migration cannot happen in the same round as enabling maintenance mode, obtained warning: "unable to score best balancing migration - leader 'ct:205' is not pres= ent in the cluster usage" * sustained imbalance round counter is not reset in case of early returns, which, e.g., can cause auto rebalance immediately after re-arming or disabling maintenance mode Looks good to me overall, I think the tiny things related to my observations could be fixed in a small follow-up. On Thu Apr 2, 2026 at 2:43 PM CEST, Daniel Kral wrote: > Here's the v4 of the load balancer patches for the HA Manager. > > Most of the patches here are already R-b'd by @Dominik (many, many > thanks!) and only a few things have changed, the biggest of course is > changing the default node imbalance threshold from '0.7' to '0.3' and > adding the pve-manager patches. > > I'm already half-way there with the pve-docs patches, but will send them > in a separate patch series (as the changes are also updating the CRS > section in general). > > Thank you very much for the feedback @Dominik, @Thomas, @Maximiliano, > and @Jillian Morgan! > [snip] Consider this as:=20 Tested-by: Dominik Rusovac