Hi Adam, Indeed, I forgot to mention it.
The behavior of Rabbit 4.x is different from 3.x, as in, if not being explicit, nodes wont be disk nodes. Keeping all nodes writing to disk is important when restarting daemons. Which is why there's this patch.
Also, the original fix was wrong, and I rewrote fix-list_users-provider.patch. I tested it more, and I'm sure it's working as expected now.
Let me know if it's ok for an upload, Cheers, Thomas Goirand (zigo)
diff -Nru puppet-module-puppetlabs-rabbitmq-8.5.0/debian/changelog puppet-module-puppetlabs-rabbitmq-8.5.0/debian/changelog --- puppet-module-puppetlabs-rabbitmq-8.5.0/debian/changelog 2025-05-27 15:02:16.000000000 +0200 +++ puppet-module-puppetlabs-rabbitmq-8.5.0/debian/changelog 2025-08-21 16:15:57.000000000 +0200 @@ -1,3 +1,11 @@ +puppet-module-puppetlabs-rabbitmq (8.5.0-8+deb13u1) trixie; urgency=medium + + * fix-list_users-provider.patch: also handle the case when there's no users + at all. + * Add setup-all-nodes-as-disk-nodes.patch. + + -- Thomas Goirand <[email protected]> Thu, 21 Aug 2025 16:15:57 +0200 + puppet-module-puppetlabs-rabbitmq (8.5.0-8) unstable; urgency=medium * Add trixie-support.patch. diff -Nru puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/fix-list_users-provider.patch puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/fix-list_users-provider.patch --- puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/fix-list_users-provider.patch 2025-05-27 15:02:16.000000000 +0200 +++ puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/fix-list_users-provider.patch 2025-08-21 16:15:57.000000000 +0200 @@ -1,16 +1,27 @@ Description: Fix list_users provider Author: Thomas Goirand <[email protected]> -Forwarded: no +Forwarded: https://github.com/voxpupuli/puppet-rabbitmq/pull/1060 Last-Update: 2025-05-29 ---- puppet-module-puppetlabs-rabbitmq-8.5.0.orig/lib/puppet/provider/rabbitmq_user/rabbitmqctl.rb -+++ puppet-module-puppetlabs-rabbitmq-8.5.0/lib/puppet/provider/rabbitmq_user/rabbitmqctl.rb -@@ -15,7 +15,7 @@ Puppet::Type.type(:rabbitmq_user).provid +Index: puppet-module-puppetlabs-rabbitmq/lib/puppet/provider/rabbitmq_user/rabbitmqctl.rb +=================================================================== +--- puppet-module-puppetlabs-rabbitmq.orig/lib/puppet/provider/rabbitmq_user/rabbitmqctl.rb ++++ puppet-module-puppetlabs-rabbitmq/lib/puppet/provider/rabbitmq_user/rabbitmqctl.rb +@@ -15,7 +15,16 @@ Puppet::Type.type(:rabbitmq_user).provid rabbitmqctl_list('users') end - user_list.split(%r{\n}).map do |line| -+ user_list.split(%r{\n})[1..-1].map do |line| ++ lines = user_list.split(%r{\n}) ++ ++ # Skip the header lines in RabbitMQ 4.x ++ # Skip "Listing users ..." and "user tags" lines ++ while lines.any? && (lines.first =~ %r{^Listing users} || lines.first =~ %r{^user\s+tags} || lines.first.strip.empty?) ++ lines.shift ++ end ++ ++ # Process remaining lines (should be user data) ++ lines.map do |line| raise Puppet::Error, "Cannot parse invalid user line: #{line}" unless line =~ %r{^(\S+)\s+\[(.*?)\]$} user = Regexp.last_match(1) tags = Regexp.last_match(2).split(%r{,\s*}) diff -Nru puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/series puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/series --- puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/series 2025-05-27 15:02:16.000000000 +0200 +++ puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/series 2025-08-21 16:15:57.000000000 +0200 @@ -4,3 +4,4 @@ increase-rabbitmq-timeout.patch trixie-support.patch fix-list_users-provider.patch +setup-all-nodes-as-disk-nodes.patch diff -Nru puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/setup-all-nodes-as-disk-nodes.patch puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/setup-all-nodes-as-disk-nodes.patch --- puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/setup-all-nodes-as-disk-nodes.patch 1970-01-01 01:00:00.000000000 +0100 +++ puppet-module-puppetlabs-rabbitmq-8.5.0/debian/patches/setup-all-nodes-as-disk-nodes.patch 2025-08-21 16:15:57.000000000 +0200 @@ -0,0 +1,25 @@ +Description: Setup all nodes as disk nodes + In most production setups, it is advised to set all nodes as disk nodes + bacause of: + * Simplicity – all nodes behave the same; no special recovery procedures. + * Resilience – if one disk node fails, you don’t risk being left with + only RAM nodes (which could cause data loss if all RAM nodes restart). + * Operational flexibility – you can remove or add nodes without worrying + about “last disk node” constraints. + * Modern RabbitMQ performance – RAM nodes used to help in early versions + (pre-3.x) when Mnesia disk I/O was slow, but with current disk and SSD + speeds, the performance difference is often negligible. +Author: Thomas Goirand <[email protected]> +Forwarded: no +Last-Update: 2025-08-11 + +--- puppet-module-puppetlabs-rabbitmq-8.5.0.orig/templates/rabbitmq.config.erb ++++ puppet-module-puppetlabs-rabbitmq-8.5.0/templates/rabbitmq.config.erb +@@ -16,6 +16,7 @@ + <% end -%> + <% if @config_cluster -%> + {cluster_nodes, {[<%= @cluster_nodes.map { |n| "\'rabbit@#{n}\'" }.join(', ') %>], <%= @cluster_node_type %>}}, ++ {rabbit, [{disk_nodes, [<%= @cluster_nodes.map { |n| "\'rabbit@#{n}\'" }.join(', ') %>]}]}, + {cluster_partition_handling, <%= @cluster_partition_handling %>}, + <% end -%> + {tcp_listen_options, [
