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 339111FF0E1 for ; Mon, 16 Nov 2026 10:18:54 +0100 (CET) Received: from gate001.proxmox.com (localhost.localdomain [127.0.0.1]) by gate001.proxmox.com (Proxmox) with ESMTP id D6DF8216D1; Mon, 16 Nov 2026 10:18:49 +0100 (CET) From: Kefu Chai To: m.carrara@proxmox.com Subject: Re: [PATCH ceph master/ceph squid/manager v4 0/7] Fix #6816: Prevent ceph-exporter Daemon from Crashing on Startup Date: Mon, 16 Nov 2026 17:18:39 +0800 Message-ID: <20261116091839.1804896-1-k.chai@proxmox.com> X-Mailer: git-send-email 2.47.3 In-Reply-To: <20260623134454.710441-1-m.carrara@proxmox.com> References: <20260623134454.710441-1-m.carrara@proxmox.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Bm-Milter-Handled: 55990f41-d878-4baa-be0a-ee34c49e34d2 X-Bm-Transport-Timestamp: 1794820723656 X-SPAM-LEVEL: Spam detection results: 0 AWL 0.206 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: ZA4DX2UITFPS6YJI7FZ6YNG5636QR7ZY X-Message-ID-Hash: ZA4DX2UITFPS6YJI7FZ6YNG5636QR7ZY X-MailFrom: k.chai@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: Hi Max, finally got around to this, sorry for the delay. I applied the pve-manager patches to current master and they still apply cleanly, which after five months is a pleasant surprise. I also like that the keyring is only rewritten when its contents actually changed, and that pve-init-ceph-crash comes down to a one-liner now. Most of the comments below are about the ceph patches. patch 3/7: > + if test -n "$2" && dpkg --compare-versions "$2" 'lt' '9.2.4'; then > + # Don't fail in case user has "exotic" configuration where RADOS > + # isn't available on all nodes for some reason > + /usr/share/pve-manager/helpers/pve-ceph-keyring --init client.exporter || true > + fi 9.2.4 is long out by now, so for most people upgrading this branch never triggers. I guess this wants a bump to whatever version the series lands in. Also, update_ceph_conf() restarts ceph-crash.service when it is enabled. Maybe do the same here, otherwise an exporter that already crash-looped into its start limit stays down even once the keyring is there. patch 6/7: > +diff --git a/systemd/ceph-exporter.service.in b/systemd/ceph-exporter.service.in Upstream renamed this to a template, ceph-exporter@.service.in, with the Tentacle 20.2.4 import, so on current master this does not apply anymore. The patches/series context moved as well, but that part is just appending at the end again. The squid patch still has the right file name. Since the branch exists now, the series could use a tentacle-stable-9 patch too. With the template unit, --id being hardcoded to 'exporter' means the instance name does not matter for the arguments, but something still has to enable an instance. Does the packaging do that? I did not dig into how the unit gets enabled these days. One tiny nit on 5/7: #!/bin/bash can be #!/bin/sh for the one-liner. And maybe a line for the commit message: the in-tree exporter never talks to the cluster at all, it only scans the admin sockets, so the mon 'allow r' cap is mostly there to have a valid key. Which I think is fine, just good to know when someone later wonders about the caps.