confctl 2.0.0 → 2.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 602260a8486eef31801f7fcf2221f09164e4443e66ea94d6fe534f8f68c56fc2
-  data.tar.gz: 24b4b5dabe240f2e93e0a9784f9eda90bbf52cec2a82fab8e5c5644e323be8da
+  metadata.gz: bd99990731daa837f1243feb4a3b098320180badbbb9cc3846c105e7b747b794
+  data.tar.gz: 2ab400c25ea0bfc35f53713826b0e69aeff4667314048aaf6f8f56c942dda4e3
 SHA512:
-  metadata.gz: 87f22fe62a9354b20e711fff429f0a34f6961f9c08489bd42f97ff1d8f9f350bc281e7eca069aecc25081ae71f2ea73f803aa06a0dc4c3d244adbfe10dbbaa86
-  data.tar.gz: 8f44626b3b0e9a97b37171cbc58e94a15b04d87f448110ff211bfa5bc34477d1ce9a0d9f344b19cab93ed8d23f657e44fab3d14c8bad71e62b5d65ea2d3d3d26
+  metadata.gz: 7b2f85762c2c9c8173ffc38fd1fd6c37e2866ae8a6d5a4d1cbfc6d4cbbb8e8037dc9e9d40d0d8f9f12f7a0eae7140c4c52f435ea6e952327803024a6cd08720b
+  data.tar.gz: a3f87cfe12b26ef6c0a520030256a798418b85cee9dbedc2cd63e35b7625aa7e763c4f3d07f9e194826c6d3cffd5f12bc1a55d04201c6963067412a5010c4bad

data/CHANGELOG.md

@@ -1,3 +1,21 @@
+# Fri Jun 06 2025 -- version 2.2.0
+- Handle `pkgs.substituteAll` / `pkgs.replaceVarsWith` compatibility on NixOS unstable
+  and 25.05
+- Change swpin files only when revisions are updated, skip commit when no changes were made
+- Add option `--[no-]editor` to `confctl swpins core/cluster/channel set/update` commands
+
+# Sun May 11 2025 -- version 2.1.0
+- Support for referring to generations by their offset
+- Resolved generations are printed on build/deploy/etc.
+- Automatically rollback faulty configurations
+- Interleave copying of carried machine generations
+- Add `confctl.programs.kexec-netboot`
+- Let the user retry dry activation in interactive mode
+- Fix listing, deletion and garbage collection of carried machines' generations
+- Read and display kernel version for each generation
+- Option to disable the garbage collection in `confctl generation rotate`
+- Shorten titles and entries in netboot menus
+
 # Sun Nov 17 2024 -- version 2.0.0
 - Distinguish machine `config` and `metaConfig` (breaking change)
 - Support for machine carriers and netboot servers
@@ -13,7 +31,7 @@ or it could mean machine metadata from `module.nix`. Machine metadata
 is now accessible as `metaConfig`.
 
 - `confLib.findConfig` has been renamed to `confLib.findMetaConfig`
-- `confLib.confLib.getClusterMachines` returns a list of machines with `metaConfig` attribute
+- `confLib.getClusterMachines` returns a list of machines with `metaConfig` attribute
 
 # Sat Feb 17 2024 -- version 1.0.0
 - Initial release

data/Gemfile

@@ -1,2 +1,8 @@
 source 'https://rubygems.org'
 gemspec
+
+group :development do
+  gem 'overcommit'
+  gem 'rubocop', '~> 1.75.0'
+  gem 'rubocop-rake'
+end

data/README.md

@@ -13,7 +13,8 @@ machines.
   configurations)
 * Query machine state, view changelogs and diffs
 * Run health checks
-* Support for creating netboot servers, see [docs/carrier.md](docs/carrier.md)
+* Automatically roll back faulty configurations
+* Support for creating netboot servers with option to kexec, see [docs/carrier.md](docs/carrier.md)
 
 ## Requirements
 

data/docs/carrier.md

@@ -136,3 +136,15 @@ images are deployed to it:
   '';
 }
 ```
+
+## kexec from the netboot server
+Set `confctl.programs.kexec-netboot.enable = true;` within your netbooted machine
+configurations. This will add program `kexec-netboot` to your system path.
+
+This program can be used to load kernel and initrd from the netboot server for kexec.
+It automatically discovers the netboot server it was booted from, by default it uses
+the latest generation found on the netboot server at the moment of execution. Use
+`kexec-netboot --interactive` to select which machine/generation/variant should be
+loaded. See `kexec-netboot --help` for all available options.
+
+The loaded kernel can be run either by `kexec-netboot -e` or `kexec -e` directly.

data/lib/confctl/cli/app.rb

@@ -11,6 +11,13 @@ module ConfCtl::Cli
     end
 
     def self.run
+      # Workaround for nix-build error:
+      #   error: creating directory '/tmp/nix-shell-5100-0/nix-build-1986594-0': No such file or directory
+      if ENV['IN_NIX_SHELL']
+        ENV['TMP'] = '/tmp'
+        ENV['TMPDIR'] = '/tmp'
+      end
+
       cli = get
       exit(cli.run(ARGV))
     end
@@ -213,6 +220,12 @@ module ConfCtl::Cli
         c.desc 'Do not activate copied closures'
         c.switch 'copy-only', negatable: false
 
+        c.desc 'Enable auto-rollback'
+        c.switch 'enable-auto-rollback', default_value: false, negatable: false
+
+        c.desc 'Disable auto-rollback'
+        c.switch 'disable-auto-rollback', default_value: false, negatable: false
+
         c.desc 'Reboot target systems after deployment'
         c.switch :reboot
 
@@ -447,6 +460,9 @@ module ConfCtl::Cli
           c.desc 'List remote machine generations'
           c.switch %i[r remote]
 
+          c.desc 'Do not run the garbage collector if enabled in configuration'
+          c.switch %i[gc collect-garbage], default_value: true
+
           c.desc 'Max number of concurrent nix-collect-garbage processes'
           c.flag 'max-concurrent-gc', arg_name: 'n', type: Integer,
                                       default_value: 5
@@ -521,6 +537,9 @@ module ConfCtl::Cli
         c.desc 'Include changelog in the commit message'
         c.switch :changelog, default_value: true
 
+        c.desc 'Open $EDITOR with commit message'
+        c.switch :editor, default_value: true
+
         c.desc 'Generate changelog for downgrade'
         c.switch %i[d downgrade], default_value: false
 

data/lib/confctl/cli/cluster.rb

@@ -333,7 +333,7 @@ module ConfCtl::Cli
 
       if opts[:interactive]
         host_generations.each do |host, gen|
-          if copy_to_host(nix, host, machines[host], gen.toplevel) == :skip
+          if copy_to_host(nix, host, machines[host], gen) == :skip
             puts Rainbow("Skipping #{host}").yellow
             skipped_copy << host
           end
@@ -351,7 +351,7 @@ module ConfCtl::Cli
           next
         end
 
-        if deploy_to_host(nix, host, machines[host], gen.toplevel, action) == :skip
+        if deploy_to_host(nix, host, machines[host], gen, action) == :skip
           puts Rainbow("Skipping #{host}").yellow
           skipped_activation << host
           next
@@ -396,14 +396,14 @@ module ConfCtl::Cli
       host_generations.each do |host, gen|
         machine = machines[host]
 
-        if copy_to_host(nix, host, machine, gen.toplevel) == :skip
+        if copy_to_host(nix, host, machine, gen) == :skip
           puts Rainbow("Skipping #{host}").yellow
           next
         end
 
         next if opts['copy-only']
 
-        if deploy_to_host(nix, host, machine, gen.toplevel, action) == :skip
+        if deploy_to_host(nix, host, machine, gen, action) == :skip
           puts Rainbow("Skipping #{host}").yellow
           next
         end
@@ -421,7 +421,7 @@ module ConfCtl::Cli
       end
     end
 
-    def copy_to_host(nix, host, machine, toplevel)
+    def copy_to_host(nix, host, machine, build_generation)
       puts Rainbow("Copying configuration to #{host} (#{machine.target_host})").yellow
 
       return :skip if opts[:interactive] && !ask_confirmation(always: true)
@@ -435,7 +435,7 @@ module ConfCtl::Cli
           width: 80
         )
 
-        ret = nix.copy(machine, toplevel) do |i, n, path|
+        ret = nix_copy(nix, machine, build_generation) do |i, n, path|
           lw << "[#{i}/#{n}] #{path}"
 
           lw.sync_console do
@@ -451,8 +451,10 @@ module ConfCtl::Cli
     end
 
     def concurrent_copy(machines, host_generations, nix)
+      sorted_generations = sort_generations_for_copy(machines, host_generations)
+
       LogView.open(
-        header: "#{Rainbow("Copying to #{host_generations.length} machines").bright}\n",
+        header: "#{Rainbow("Copying to #{sorted_generations.length} machines").bright}\n",
         title: Rainbow('Live view').bright
       ) do |lw|
         multibar = TTY::ProgressBar::Multi.new(
@@ -461,13 +463,13 @@ module ConfCtl::Cli
         )
         executor = ConfCtl::ParallelExecutor.new(opts['max-concurrent-copy'])
 
-        host_generations.each do |host, gen|
+        sorted_generations.each do |host, gen|
           pb = multibar.register(
             "#{host} [:bar] :current/:total (:percent)"
           )
 
           executor.add do
-            ret = nix.copy(machines[host], gen.toplevel) do |i, n, path|
+            ret = nix_copy(nix, machines[host], gen) do |i, n, path|
               lw << "#{host}> [#{i}/#{n}] #{path}"
 
               lw.sync_console do
@@ -504,67 +506,136 @@ module ConfCtl::Cli
       end
     end
 
-    def deploy_to_host(nix, host, machine, toplevel, action)
+    # Sort generations for copy to target machines
+    #
+    # This algorithm interleaves carried machines from different carriers,
+    # so that we copy to multiple carriers at the same time, and not many carried
+    # machines to one carrier.
+    def sort_generations_for_copy(machines, host_generations)
+      carried, standalone = machines.to_a.partition(&:carried?)
+
+      carried_groups = carried.group_by { |m| m.carrier_machine.name }
+
+      sorted_generations = standalone.map do |m|
+        [m.name, host_generations[m.name]]
+      end
+
+      until carried_groups.empty?
+        carried_groups.delete_if do |_, machines|
+          m = machines.shift
+          next(true) if m.nil?
+
+          sorted_generations << [m.name, host_generations[m.name]]
+
+          machines.empty?
+        end
+      end
+
+      if sorted_generations.length != host_generations.length
+        raise 'programming error: sorted generations length != original generations'
+      end
+
+      sorted_generations
+    end
+
+    def deploy_to_host(nix, host, machine, generation, action)
+      ret = nil
+
       LogView.open_with_logger(
         header: "#{Rainbow('Deploying to').bright} #{Rainbow(host).yellow}\n",
         title: Rainbow('Live view').bright,
         size: :auto,
         reserved_lines: 10
       ) do |lw|
-        if machine.carried?
-          deploy_carried_to_host(lw, nix, host, machine, toplevel, action)
-        else
-          deploy_standalone_to_host(lw, nix, host, machine, toplevel, action)
-        end
+        ret =
+          if machine.carried?
+            deploy_carried_to_host(lw, nix, host, machine, generation, action)
+          else
+            deploy_standalone_to_host(lw, nix, host, machine, generation, action)
+          end
 
         lw.flush
       end
-    end
 
-    def deploy_standalone_to_host(lw, nix, host, machine, toplevel, action)
-      if opts['dry-activate-first']
-        lw.sync_console do
-          puts Rainbow(
-            "Trying to activate configuration on #{host} " \
-            "(#{machine.target_host})"
-          ).yellow
-        end
-
-        raise "Error while activating configuration on #{host}" unless nix.activate(machine, toplevel, 'dry-activate')
-      end
-
-      lw.sync_console do
-        puts Rainbow(
-          "Activating configuration on #{host} (#{machine.target_host}): " \
-          "#{action}"
-        ).yellow
-      end
+      ret
+    end
 
-      return :skip if opts[:interactive] && !ask_confirmation(always: true)
+    def deploy_standalone_to_host(lw, nix, host, machine, generation, action)
+      return :skip if pre_host_activate(lw, nix, host, machine, generation, action) == :skip
 
       # rubocop:disable Style/GuardClause
 
-      unless nix.activate(machine, toplevel, action)
-        raise "Error while activating configuration on #{host}"
-      end
+      result =
+        if %w[test switch].include?(action) && enable_auto_rollback?(machine, generation)
+          nix.activate_with_rollback(machine, generation, action)
+        else
+          nix.activate(machine, generation, action)
+        end
+
+      raise "Error while activating configuration on #{host}" unless result
 
-      if %w[boot switch].include?(action) && !nix.set_profile(machine, toplevel)
+      if %w[boot switch].include?(action) && !nix.set_profile(machine, generation.toplevel)
         raise "Error while setting profile on #{host}"
       end
 
       # rubocop:enable Style/GuardClause
     end
 
-    def deploy_carried_to_host(lw, nix, host, machine, toplevel, action)
+    def deploy_carried_to_host(lw, nix, host, machine, generation, action)
       return if action != 'switch'
 
       # rubocop:disable Style/GuardClause
-      unless nix.set_carried_profile(machine, toplevel)
+      unless nix.set_carried_profile(machine, generation.toplevel)
         raise "Error while setting carried profile for #{host} on #{machine.carrier_machine}"
       end
       # rubocop:enable Style/GuardClause
     end
 
+    def pre_host_activate(lw, nix, host, machine, generation, action)
+      loop do
+        if opts['dry-activate-first']
+          lw.sync_console do
+            puts Rainbow(
+              "Trying to activate configuration on #{host} " \
+              "(#{machine.target_host})"
+            ).yellow
+          end
+
+          unless nix.activate(machine, generation, 'dry-activate')
+            raise "Error while activating configuration on #{host}"
+          end
+        end
+
+        lw.sync_console do
+          puts Rainbow(
+            "Activating configuration on #{host} (#{machine.target_host}): " \
+            "#{action}"
+          ).yellow
+        end
+
+        return unless opts[:interactive]
+
+        options = {}
+        options['y'] = 'Continue'
+        options['r'] = 'Retry' if opts['dry-activate-first']
+        options['s'] = 'Skip'
+        options['a'] = 'Abort'
+
+        answer = ask_action(options:, default: nil)
+
+        case answer
+        when 'y'
+          return
+        when 'r'
+          next
+        when 's'
+          return :skip
+        when 'a'
+          raise 'Aborting'
+        end
+      end
+    end
+
     def reboot_host(host, machine)
       if machine.localhost?
         puts Rainbow("Skipping reboot of #{host} as it is localhost").yellow
@@ -896,12 +967,15 @@ module ConfCtl::Cli
     def find_generations(machines, generation_name)
       host_generations = {}
       missing_hosts = []
+      generation_offset = generation_name.to_i if /\A-?\d+\z/ =~ generation_name
 
       machines.each_key do |host|
         list = ConfCtl::Generation::BuildList.new(host)
 
         gen =
-          if generation_name == 'current'
+          if generation_offset
+            list.at_offset(generation_offset)
+          elsif generation_name == 'current'
             list.current
           else
             list[generation_name]
@@ -916,8 +990,17 @@ module ConfCtl::Cli
 
       raise 'No generation found' if host_generations.empty?
 
+      puts 'Resolved host generations:'
+      list_generations(host_generations, missing_hosts:)
+
+      if opts[:interactive]
+        ask_confirmation!
+      else
+        puts
+      end
+
       if missing_hosts.any?
-        ask_confirmation! do
+        ask_confirmation!(always: opts[:interactive]) do
           puts "Generation '#{generation_name}' was not found on the following hosts:"
           missing_hosts.each { |host| puts "  #{host}" }
           puts
@@ -928,22 +1011,59 @@ module ConfCtl::Cli
       host_generations
     end
 
+    def list_generations(host_generations, missing_hosts:)
+      swpin_names = []
+
+      host_generations.each_value do |gen|
+        gen.swpin_names.each do |name|
+          swpin_names << name unless swpin_names.include?(name)
+        end
+      end
+
+      rows = host_generations.map do |host, gen|
+        row = {
+          'name' => host,
+          'generation' => gen.name,
+          'kernel' => gen.kernel_version
+        }
+
+        gen.swpin_specs.each do |name, spec|
+          row[name] = spec.version
+        end
+
+        row
+      end
+
+      missing_hosts.each do |host|
+        rows << { 'name' => host, 'generation' => 'not found' }
+      end
+
+      OutputFormatter.print(
+        rows,
+        %w[name generation kernel] + swpin_names,
+        layout: :columns,
+        sort: %w[name generation]
+      )
+    end
+
     def do_build(machines)
       nix = ConfCtl::Nix.new(
         show_trace: opts['show-trace'],
         max_jobs: opts['max-jobs'],
         cores: opts['cores']
       )
-      hosts_swpin_paths = {}
 
       autoupdate_swpins(machines)
       host_swpin_specs = check_swpins(machines)
 
       raise 'one or more swpins need to be updated' unless host_swpin_specs
 
-      machines.each do |host, d|
-        puts Rainbow("Evaluating swpins for #{host}...").bright
-        hosts_swpin_paths[host] = nix.eval_host_swpins(host).update(d.nix_paths)
+      puts Rainbow("Evaluating swpins for #{machines.length} machines...").bright
+
+      hosts_swpin_paths = nix.eval_host_swpins(machines.map { |host, _| host })
+
+      machines.each do |host, m|
+        hosts_swpin_paths[host].update(m.nix_paths)
       end
 
       grps = swpin_build_groups(hosts_swpin_paths)
@@ -1273,5 +1393,21 @@ module ConfCtl::Cli
 
       raise ArgumentError, "invalid time duration '#{interval}'"
     end
+
+    # @param nix [Nix]
+    # @param machine [Machine]
+    # @param generation [Generation::Build]
+    def nix_copy(nix, machine, generation, &)
+      paths = [generation.toplevel]
+      paths << generation.auto_rollback if enable_auto_rollback?(machine, generation)
+
+      nix.copy(machine, paths, &)
+    end
+
+    def enable_auto_rollback?(machine, generation)
+      !opts['disable-auto-rollback'] \
+        && (opts['enable-auto-rollback'] || machine.auto_rollback?) \
+        && generation.auto_rollback
+    end
   end
 end

data/lib/confctl/cli/generation.rb

@@ -11,6 +11,7 @@ module ConfCtl::Cli
     def remove
       machines = select_machines(args[0])
       gens = select_generations(machines, args[1])
+      changed_hosts = []
 
       if gens.empty?
         puts 'No generations found'
@@ -27,21 +28,36 @@ module ConfCtl::Cli
       gens.each do |gen|
         puts "Removing #{gen.presence_str} generation #{gen.host}@#{gen.name}"
         gen.destroy
+
+        changed_hosts << gen.host unless changed_hosts.include?(gen.host)
       end
 
       return unless opts[:remote] && opts[:gc]
 
-      machines_gc = machines.runnable.select do |host, _machine|
-        gens.detect { |gen| gen.host == host }
+      machines_gc = {}
+
+      machines.each do |host, machine|
+        next unless changed_hosts.include?(host)
+
+        m =
+          if machine.carried?
+            machine.carrier_machine
+          else
+            machine
+          end
+
+        machines_gc[m.name] = m if m.target_host
       end
 
-      run_gc(machines_gc)
+      run_gc(ConfCtl::MachineList.new(machines: machines_gc))
     end
 
     def rotate
       machines = select_machines(args[0])
 
       to_delete = []
+      changed_hosts = []
+      enable_gc = opts[:remote] && opts[:gc]
 
       to_delete.concat(host_generations_rotate(machines)) if opts[:remote]
 
@@ -56,29 +72,37 @@ module ConfCtl::Cli
         puts 'The following generations will be removed:'
         OutputFormatter.print(to_delete, %i[host name type id], layout: :columns)
         puts
-        puts "Garbage collection: #{opts[:remote] ? 'when enabled in configuration' : 'no'}"
+        puts "Garbage collection: #{enable_gc ? 'when enabled in configuration' : 'no'}"
       end
 
       to_delete.each do |gen|
         puts "Removing #{gen[:type]} generation #{gen[:host]}@#{gen[:name]}"
         gen[:generation].destroy
+
+        changed_hosts << gen[:host] unless changed_hosts.include?(gen[:host])
       end
 
-      return unless opts[:remote]
+      return unless enable_gc
 
       global = ConfCtl::Settings.instance.host_generations
+      machines_gc = {}
 
-      machines_gc = machines.runnable.select do |_host, machine|
-        gc = machine['buildGenerations']['collectGarbage']
+      machines.each do |host, machine|
+        next unless changed_hosts.include?(host)
 
-        if gc.nil?
-          global['collectGarbage']
-        else
-          gc
-        end
+        m =
+          if machine.carried?
+            machine.carrier_machine
+          else
+            machine
+          end
+
+        next if !m.target_host || (!m['buildGenerations']['collectGarbage'] && !global['collectGarbage'])
+
+        machines_gc[m.name] = m
       end
 
-      run_gc(machines_gc) if machines_gc.any?
+      run_gc(ConfCtl::MachineList.new(machines: machines_gc)) if machines_gc.any?
     end
 
     def collect_garbage
@@ -131,12 +155,16 @@ module ConfCtl::Cli
       select_older_than =
         (Time.now - (::Regexp.last_match(1).to_i * 24 * 60 * 60) if !select_old && /\A(\d+)d\Z/ =~ pattern)
 
+      gen_at = gens.at_offset(pattern.to_i) if /\A-?\d+\z/ =~ pattern
+
       if pattern
         gens.delete_if do |gen|
           if select_old
             gen.current
           elsif select_older_than
             gen.date >= select_older_than
+          elsif gen_at
+            gen != gen_at
           else
             !ConfCtl::Pattern.match?(pattern, gen.name)
           end
@@ -316,7 +344,8 @@ module ConfCtl::Cli
           'name' => gen.name,
           'id' => gen.id,
           'presence' => gen.presence_str,
-          'current' => gen.current_str
+          'current' => gen.current_str,
+          'kernel' => gen.kernel_version
         }
 
         gen.swpin_specs.each do |name, spec|
@@ -328,7 +357,7 @@ module ConfCtl::Cli
 
       OutputFormatter.print(
         rows,
-        %w[host name id presence current] + swpin_names,
+        %w[host name id presence current kernel] + swpin_names,
         layout: :columns,
         sort: %w[name host]
       )

data/lib/confctl/cli/swpins/channel.rb

@@ -37,11 +37,12 @@ module ConfCtl::Cli
 
       channels.each(&:save)
 
-      return unless opts[:commit]
+      return if !opts[:commit] || !change_set.any_changes?
 
       change_set.commit(
         type: opts[:downgrade] ? :downgrade : :upgrade,
-        changelog: opts[:changelog]
+        changelog: opts[:changelog],
+        editor: opts[:editor]
       )
     end
 
@@ -62,11 +63,12 @@ module ConfCtl::Cli
 
       channels.each(&:save)
 
-      return unless opts[:commit]
+      return if !opts[:commit] || !change_set.any_changes?
 
       change_set.commit(
         type: opts[:downgrade] ? :downgrade : :upgrade,
-        changelog: opts[:changelog]
+        changelog: opts[:changelog],
+        editor: opts[:editor]
       )
     end
   end

data/lib/confctl/cli/swpins/cluster.rb

@@ -42,11 +42,12 @@ module ConfCtl::Cli
 
       cluster_names.each(&:save)
 
-      return unless opts[:commit]
+      return if !opts[:commit] || !change_set.any_changes?
 
       change_set.commit(
         type: opts[:downgrade] ? :downgrade : :upgrade,
-        changelog: opts[:changelog]
+        changelog: opts[:changelog],
+        editor: opts[:editor]
       )
     end
 
@@ -69,11 +70,12 @@ module ConfCtl::Cli
 
       cluster_names.each(&:save)
 
-      return unless opts[:commit]
+      return if !opts[:commit] || !change_set.any_changes?
 
       change_set.commit(
         type: opts[:downgrade] ? :downgrade : :upgrade,
-        changelog: opts[:changelog]
+        changelog: opts[:changelog],
+        editor: opts[:editor]
       )
     end
   end