cpl 1.3.0 → 2.0.0.rc.0

data/lib/command/run.rb

@@ -1,7 +1,15 @@
 # frozen_string_literal: true
 
 module Command
-  class Run < Base
+  class Run < Base # rubocop:disable Metrics/ClassLength
+    INTERACTIVE_COMMANDS = [
+      "bash",
+      "rails console",
+      "rails c",
+      "rails dbconsole",
+      "rails db"
+    ].freeze
+
     NAME = "run"
     USAGE = "run COMMAND"
     REQUIRES_ARGS = true
@@ -9,34 +17,53 @@ module Command
     OPTIONS = [
       app_option(required: true),
       image_option,
+      log_method_option,
       workload_option,
       location_option,
       use_local_token_option,
-      terminal_size_option
+      terminal_size_option,
+      interactive_option,
+      detached_option,
+      cpu_option,
+      memory_option,
+      entrypoint_option
     ].freeze
-    DESCRIPTION = "Runs one-off **_interactive_** replicas (analog of `heroku run`)"
+    DESCRIPTION = "Runs one-off interactive or non-interactive replicas (analog of `heroku run`)"
     LONG_DESCRIPTION = <<~DESC
-      - Runs one-off **_interactive_** replicas (analog of `heroku run`)
-      - Uses `Standard` workload type and `cpln exec` as the execution method, with CLI streaming
-      - If `fix_terminal_size` is `true` in the `.controlplane/controlplane.yml` file, the remote terminal size will be fixed to match the local terminal size (may also be overriden through `--terminal-size`)
-
-      > **IMPORTANT:** Useful for development where it's needed for interaction, and where network connection drops and
-      > task crashing are tolerable. For production tasks, it's better to use `cpl run:detached`.
+      - Runs one-off interactive or non-interactive replicas (analog of `heroku run`)
+      - Uses `Cron` workload type and either:
+      - - `cpln workload exec` for interactive mode, with CLI streaming
+      - - log async fetching for non-interactive mode
+      - The Dockerfile entrypoint is used as the command by default, which assumes `exec "${@}"` to be present,
+        and the args ["bash", "-c", cmd_to_run] are passed
+      - The entrypoint can be overriden through `--entrypoint`, which must be a single command or a script path that exists in the container,
+        and the args ["bash", "-c", cmd_to_run] are passed,
+        unless the entrypoint is `bash`, in which case the args ["-c", cmd_to_run] are passed
+      - Providing `--entrypoint none` sets the entrypoint to `bash` by default
+      - If `fix_terminal_size` is `true` in the `.controlplane/controlplane.yml` file,
+        the remote terminal size will be fixed to match the local terminal size (may also be overriden through `--terminal-size`)
     DESC
     EXAMPLES = <<~EX
       ```sh
       # Opens shell (bash by default).
       cpl run -a $APP_NAME
 
-      # Need to quote COMMAND if setting ENV value or passing args.
-      cpl run -a $APP_NAME -- 'LOG_LEVEL=warn rails db:migrate'
+      # Runs interactive command, keeps shell open, and stops job when exiting.
+      cpl run -a $APP_NAME --interactive -- rails c
+
+      # Some commands are automatically detected as interactive, so no need to pass `--interactive`.
+      #{INTERACTIVE_COMMANDS.map { |cmd| "cpl run -a $APP_NAME -- #{cmd}" }.join("\n      ")}
 
-      # Runs command, displays output, and exits shell.
-      cpl run -a $APP_NAME -- ls /
-      cpl run -a $APP_NAME -- rails db:migrate:status
+      # Runs non-interactive command, outputs logs, exits with the exit code of the command and stops job.
+      cpl run -a $APP_NAME -- rails db:migrate
 
-      # Runs command and keeps shell open.
-      cpl run -a $APP_NAME -- rails c
+      # Runs non-iteractive command, detaches, exits with 0, and prints commands to:
+      # - see logs from the job
+      # - stop the job
+      cpl run -a $APP_NAME --detached -- rails db:migrate
+
+      # The command needs to be quoted if setting an env variable or passing args.
+      cpl run -a $APP_NAME -- 'SOME_ENV_VAR=some_value rails db:migrate'
 
       # Uses a different image (which may not be promoted yet).
       cpl run -a $APP_NAME --image appimage:123 -- rails db:migrate # Exact image name
@@ -48,97 +75,432 @@ module Command
       # Overrides remote CPLN_TOKEN env variable with local token.
       # Useful when superuser rights are needed in remote container.
       cpl run -a $APP_NAME --use-local-token -- bash
+
+      # Replaces the existing Dockerfile entrypoint with `bash`.
+      cpl run -a $APP_NAME --entrypoint none -- rails db:migrate
+
+      # Replaces the existing Dockerfile entrypoint.
+      cpl run -a $APP_NAME --entrypoint /app/alternative-entrypoint.sh -- rails db:migrate
       ```
     EX
 
-    attr_reader :location, :workload_to_clone, :workload_clone, :container
+    MAGIC_END = "---cpl run command finished---"
+
+    attr_reader :interactive, :detached, :location, :original_workload, :runner_workload,
+                :container, :image_link, :image_changed, :job, :replica, :command
+
+    def call # rubocop:disable Metrics/CyclomaticComplexity, Metrics/MethodLength, Metrics/PerceivedComplexity
+      @interactive = config.options[:interactive] || interactive_command?
+      @detached = config.options[:detached]
+      @log_method = config.options[:log_method]
 
-    def call # rubocop:disable Metrics/MethodLength
       @location = config.location
-      @workload_to_clone = config.options["workload"] || config[:one_off_workload]
-      @workload_clone = "#{workload_to_clone}-run-#{random_four_digits}"
+      @original_workload = config.options[:workload] || config[:one_off_workload]
+      @runner_workload = "#{original_workload}-runner"
+
+      unless interactive
+        @internal_sigint = false
+
+        # Catch Ctrl+C in the main process
+        trap("SIGINT") do
+          unless @internal_sigint
+            print_detached_commands
+            exit(ExitCode::INTERRUPT)
+          end
+        end
+      end
 
-      step("Cloning workload '#{workload_to_clone}' on app '#{config.options[:app]}' to '#{workload_clone}'") do
-        clone_workload
+      if cp.fetch_workload(runner_workload).nil?
+        create_runner_workload
+        wait_for_runner_workload_create
       end
+      update_runner_workload
+      wait_for_runner_workload_update
+
+      # NOTE: need to wait some time before starting the job,
+      # otherwise the image may not be updated yet
+      # TODO: need to figure out if there's a better way to do this
+      sleep 1 if image_changed
+
+      start_job
+      wait_for_replica_for_job
 
-      wait_for_workload(workload_clone)
-      wait_for_replica(workload_clone, location)
-      run_in_replica
-    ensure
       progress.puts
-      ensure_workload_deleted(workload_clone)
+      if interactive
+        run_interactive
+      else
+        run_non_interactive
+      end
     end
 
     private
 
-    def clone_workload # rubocop:disable Metrics/MethodLength
-      # Create a base copy of workload props
-      spec = cp.fetch_workload!(workload_to_clone).fetch("spec")
-      container_spec = spec["containers"].detect { _1["name"] == workload_to_clone } || spec["containers"].first
-      @container = container_spec["name"]
+    def interactive_command?
+      INTERACTIVE_COMMANDS.include?(args_join(config.args))
+    end
+
+    def app_workload_replica_args
+      ["-a", config.app, "--workload", runner_workload, "--replica", replica]
+    end
 
-      # remove other containers if any
-      spec["containers"] = [container_spec]
+    def create_runner_workload # rubocop:disable Metrics/MethodLength
+      step("Creating runner workload '#{runner_workload}' based on '#{original_workload}'") do
+        spec, container_spec = base_workload_specs(original_workload)
 
-      # Stub workload command with dummy server that just responds to port
-      # Needed to avoid execution of ENTRYPOINT and CMD of Dockerfile
-      container_spec["command"] = "ruby"
-      container_spec["args"] = ["-e", Scripts.http_dummy_server_ruby]
+        # Remove other containers if any
+        spec["containers"] = [container_spec]
 
-      # Ensure one-off workload will be running
-      spec["defaultOptions"]["suspend"] = false
+        # Default to using existing Dockerfile entrypoint
+        container_spec.delete("command")
 
-      # Ensure no scaling
-      spec["defaultOptions"]["autoscaling"]["minScale"] = 1
-      spec["defaultOptions"]["autoscaling"]["maxScale"] = 1
-      spec["defaultOptions"]["capacityAI"] = false
+        # Remove props that conflict with job
+        container_spec.delete("ports")
+        container_spec.delete("lifecycle")
+        container_spec.delete("livenessProbe")
+        container_spec.delete("readinessProbe")
 
-      # Override image if specified
-      image = config.options[:image]
-      image = latest_image if image == "latest"
-      container_spec["image"] = "/org/#{config.org}/image/#{image}" if image
+        # Ensure cron workload won't run per schedule
+        spec["defaultOptions"]["suspend"] = true
 
-      # Set runner
-      container_spec["env"] ||= []
-      container_spec["env"] << { "name" => "CONTROLPLANE_RUNNER", "value" => runner_script }
+        # Ensure no scaling
+        spec["defaultOptions"]["autoscaling"] = {}
+        spec["defaultOptions"]["capacityAI"] = false
 
-      if config.options["use_local_token"]
-        container_spec["env"] << { "name" => "CONTROLPLANE_TOKEN", "value" => ControlplaneApiDirect.new.api_token }
+        # Set cron job props
+        spec["type"] = "cron"
+
+        # Next job set to run on January 1st, 2029
+        spec["job"] = { "schedule" => "0 0 1 1 1", "restartPolicy" => "Never" }
+
+        # Create runner workload
+        cp.apply_hash("kind" => "workload", "name" => runner_workload, "spec" => spec)
       end
+    end
+
+    def update_runner_workload # rubocop:disable Metrics/MethodLength
+      step("Updating runner workload '#{runner_workload}' based on '#{original_workload}'") do
+        _, original_container_spec = base_workload_specs(original_workload)
+        spec, container_spec = base_workload_specs(runner_workload)
 
-      # Create workload clone
-      cp.apply_hash("kind" => "workload", "name" => workload_clone, "spec" => spec)
+        # Override image if specified
+        image = config.options[:image]
+        if image
+          image = latest_image if image == "latest"
+          @image_link = "/org/#{config.org}/image/#{image}"
+        else
+          @image_link = original_container_spec["image"]
+        end
+        @image_changed = container_spec["image"] != image_link
+        container_spec["image"] = image_link
+
+        # Container overrides
+        container_spec["cpu"] = config.options[:cpu] if config.options[:cpu]
+        container_spec["memory"] = config.options[:memory] if config.options[:memory]
+
+        # Update runner workload
+        cp.apply_hash("kind" => "workload", "name" => runner_workload, "spec" => spec)
+      end
     end
 
-    def runner_script # rubocop:disable Metrics/MethodLength
-      script = Scripts.helpers_cleanup
+    def wait_for_runner_workload_create
+      step("Waiting for runner workload '#{runner_workload}' to be created", retry_on_failure: true) do
+        cp.fetch_workload(runner_workload)
+      end
+    end
+
+    def wait_for_runner_workload_update
+      step("Waiting for runner workload '#{runner_workload}' to be updated", retry_on_failure: true) do
+        _, container_spec = base_workload_specs(runner_workload)
+        container_spec["image"] == image_link
+      end
+    end
+
+    def start_job
+      job_start_yaml = build_job_start_yaml
+
+      step("Starting job for runner workload '#{runner_workload}'", retry_on_failure: true) do
+        result = cp.start_cron_workload(runner_workload, job_start_yaml, location: location)
+        @job = result&.dig("items", 0, "id")
+
+        job || false
+      end
+    end
+
+    def wait_for_replica_for_job
+      step("Waiting for replica to start, which runs job '#{job}'", retry_on_failure: true) do
+        result = cp.fetch_workload_replicas(runner_workload, location: location)
+        @replica = result["items"].find { |item| item.include?(job) }
+
+        replica || false
+      end
+    end
+
+    def run_interactive
+      progress.puts("Connecting to replica '#{replica}'...\n\n")
+      cp.workload_exec(runner_workload, replica, location: location, container: container, command: command)
+    end
 
-      if config.options["use_local_token"]
-        script += <<~SHELL
-          CPLN_TOKEN=$CONTROLPLANE_TOKEN
-          unset CONTROLPLANE_TOKEN
-        SHELL
+    def run_non_interactive
+      if detached
+        print_detached_commands
+        exit(ExitCode::SUCCESS)
       end
 
+      case @log_method
+      when 1 then run_non_interactive_v1
+      when 2 then run_non_interactive_v2
+      when 3 then run_non_interactive_v3
+      else raise "Invalid log method: #{@log_method}"
+      end
+    end
+
+    def run_non_interactive_v1 # rubocop:disable Metrics/MethodLength
+      logs_pid = Process.fork do
+        # Catch Ctrl+C in the forked process
+        trap("SIGINT") do
+          exit(ExitCode::SUCCESS)
+        end
+
+        Cpl::Cli.start(["logs", *app_workload_replica_args])
+      end
+      Process.detach(logs_pid)
+
+      exit_status = wait_for_job_status
+
+      # We need to wait a bit for the logs to appear,
+      # otherwise it may exit without showing them
+      Kernel.sleep(30)
+
+      @internal_sigint = true
+      Process.kill("INT", logs_pid)
+      exit(exit_status)
+    end
+
+    def run_non_interactive_v2
+      current_cpl = File.expand_path("cpl", "#{__dir__}/../..")
+      logs_pipe = IO.popen([current_cpl, "logs", *app_workload_replica_args])
+
+      exit_status = wait_for_job_status_and_log(logs_pipe)
+
+      @internal_sigint = true
+      Process.kill("INT", logs_pipe.pid)
+      exit(exit_status)
+    end
+
+    def run_non_interactive_v3
+      exit(show_logs_waiting)
+    end
+
+    def base_workload_specs(workload)
+      spec = cp.fetch_workload!(workload).fetch("spec")
+      container_spec = spec["containers"].detect { _1["name"] == original_workload } || spec["containers"].first
+      @container = container_spec["name"]
+
+      [spec, container_spec]
+    end
+
+    def build_job_start_yaml # rubocop:disable Metrics/CyclomaticComplexity, Metrics/MethodLength, Metrics/PerceivedComplexity
+      job_start_hash = { "name" => container }
+
+      if config.options[:use_local_token]
+        job_start_hash["env"] ||= []
+        job_start_hash["env"].push({ "name" => "CPL_TOKEN", "value" => ControlplaneApiDirect.new.api_token[:token] })
+      end
+
+      entrypoint = nil
+      if config.options[:entrypoint]
+        entrypoint = config.options[:entrypoint] == "none" ? "bash" : config.options[:entrypoint]
+      end
+
+      job_start_hash["command"] = entrypoint if entrypoint
+      job_start_hash["args"] ||= []
+      job_start_hash["args"].push("bash") unless entrypoint == "bash"
+      job_start_hash["args"].push("-c")
+      job_start_hash["env"] ||= []
+      job_start_hash["env"].push({ "name" => "CPL_RUNNER_SCRIPT", "value" => runner_script })
+      if interactive
+        job_start_hash["env"].push({ "name" => "CPL_MONITORING_SCRIPT", "value" => interactive_monitoring_script })
+
+        job_start_hash["args"].push('eval "$CPL_MONITORING_SCRIPT"')
+        @command = %(bash -c 'eval "$CPL_RUNNER_SCRIPT"')
+      else
+        job_start_hash["args"].push('eval "$CPL_RUNNER_SCRIPT"')
+      end
+
+      job_start_hash.to_yaml
+    end
+
+    def interactive_monitoring_script
+      <<~SCRIPT
+        primary_pid=""
+
+        check_primary() {
+          if ! kill -0 $primary_pid 2>/dev/null; then
+            echo "Primary process has exited. Shutting down."
+            exit 0
+          fi
+        }
+
+        while true; do
+          if [[ -z "$primary_pid" ]]; then
+            primary_pid=$(ps -eo pid,etime,cmd --sort=etime | grep -v "$$" | grep -v 'ps -eo' | grep -v 'grep' | grep 'CPL_RUNNER_SCRIPT' | head -n 1 | awk '{print $1}')
+            if [[ ! -z "$primary_pid" ]]; then
+              echo "Primary process set with PID: $primary_pid"
+            fi
+          else
+            check_primary
+          fi
+
+          sleep 1
+        done
+      SCRIPT
+    end
+
+    def interactive_runner_script
+      script = ""
+
       # NOTE: fixes terminal size to match local terminal
       if config.current[:fix_terminal_size] || config.options[:terminal_size]
         if config.options[:terminal_size]
           rows, cols = config.options[:terminal_size].split(",")
         else
+          # NOTE: cannot use `Shell.cmd` here, as `stty size` has to run in a terminal environment
           rows, cols = `stty size`.split(/\s+/)
         end
-        script += "stty rows #{rows}\nstty cols #{cols}\n" if rows && cols
+        script += "stty rows #{rows}\nstty cols #{cols}\n"
       end
 
-      script += args_join(config.args)
       script
     end
 
-    def run_in_replica
-      progress.puts("Connecting...\n\n")
-      command = %(bash -c 'eval "$CONTROLPLANE_RUNNER"')
-      cp.workload_exec(workload_clone, location: location, container: container, command: command)
+    def runner_script # rubocop:disable Metrics/MethodLength
+      script = <<~SCRIPT
+        unset CPL_RUNNER_SCRIPT
+        unset CPL_MONITORING_SCRIPT
+
+        if [ -n "$CPL_TOKEN" ]; then
+          CPLN_TOKEN=$CPL_TOKEN
+          unset CPL_TOKEN
+        fi
+      SCRIPT
+
+      script += interactive_runner_script if interactive
+
+      script +=
+        if @log_method == 1
+          args_join(config.args)
+        else
+          <<~SCRIPT
+            ( #{args_join(config.args)} )
+            CPL_EXIT_CODE=$?
+            echo '#{MAGIC_END}'
+            exit $CPL_EXIT_CODE
+          SCRIPT
+        end
+
+      script
+    end
+
+    def wait_for_job_status
+      Kernel.sleep(1) until (exit_code = resolve_job_status)
+      exit_code
+    end
+
+    def wait_for_job_status_and_log(logs_pipe) # rubocop:disable Metrics/MethodLength
+      no_logs_counter = 0
+
+      loop do
+        no_logs_counter += 1
+        break if no_logs_counter > 60 # 30s
+        break if logs_pipe.eof?
+        next Kernel.sleep(0.5) unless logs_pipe.ready?
+
+        no_logs_counter = 0
+        line = logs_pipe.gets
+        break if line.chomp == MAGIC_END
+
+        puts(line)
+      end
+
+      resolve_job_status
+    end
+
+    def print_detached_commands
+      app_workload_replica_config = app_workload_replica_args.join(" ")
+      progress.puts(
+        "\n\n" \
+        "- To view logs from the job, run:\n  `cpl logs #{app_workload_replica_config}`\n" \
+        "- To stop the job, run:\n  `cpl ps:stop #{app_workload_replica_config}`\n"
+      )
+    end
+
+    def resolve_job_status
+      result = cp.fetch_cron_workload(runner_workload, location: location)
+      job_details = result&.dig("items")&.find { |item| item["id"] == job }
+      status = job_details&.dig("status")
+
+      case status
+      when "failed"
+        ExitCode::ERROR_DEFAULT
+      when "successful"
+        ExitCode::SUCCESS
+      end
+    end
+
+    ###########################################
+    ### temporary extaction from run:detached
+    ###########################################
+    def show_logs_waiting # rubocop:disable Metrics/MethodLength
+      retries = 0
+      begin
+        job_finished_count = 0
+        loop do
+          case print_uniq_logs
+          when :finished
+            break
+          when :changed
+            next
+          else
+            job_finished_count += 1 if resolve_job_status
+            break if job_finished_count > 5
+
+            sleep(1)
+          end
+        end
+
+        resolve_job_status
+      rescue RuntimeError => e
+        raise "#{e} Exiting..." unless retries < 10 # MAX_RETRIES
+
+        progress.puts(Shell.color("ERROR: #{e} Retrying...", :red))
+        retries += 1
+        retry
+      end
+    end
+
+    def print_uniq_logs
+      status = nil
+
+      @printed_log_entries ||= []
+      ts = Time.now.to_i
+      entries = normalized_log_entries(from: ts - 60, to: ts)
+
+      (entries - @printed_log_entries).sort.each do |(_ts, val)|
+        status ||= :changed
+        val.chomp == MAGIC_END ? status = :finished : progress.puts(val)
+      end
+
+      @printed_log_entries = entries # as well truncate old entries if any
+
+      status || :unchanged
+    end
+
+    def normalized_log_entries(from:, to:)
+      log = cp.log_get(workload: runner_workload, from: from, to: to, replica: replica)
+
+      log["data"]["result"]
+        .each_with_object([]) { |obj, result| result.concat(obj["values"]) }
+        .select { |ts, _val| ts[..-10].to_i > from }
     end
   end
 end

data/lib/command/setup_app.rb

@@ -4,16 +4,19 @@ module Command
   class SetupApp < Base
     NAME = "setup-app"
     OPTIONS = [
-      app_option(required: true)
+      app_option(required: true),
+      skip_secret_access_binding_option
     ].freeze
     DESCRIPTION = "Creates an app and all its workloads"
     LONG_DESCRIPTION = <<~DESC
       - Creates an app and all its workloads
       - Specify the templates for the app and workloads through `setup_app_templates` in the `.controlplane/controlplane.yml` file
       - This should only be used for temporary apps like review apps, never for persistent apps like production (to update workloads for those, use 'cpl apply-template' instead)
+      - Automatically binds the app to the secrets policy, as long as both the identity and the policy exist
+      - Use `--skip-secret-access-binding` to prevent the automatic bind
     DESC
 
-    def call
+    def call # rubocop:disable Metrics/MethodLength
       templates = config[:setup_app_templates]
 
       app = cp.fetch_gvc
@@ -24,6 +27,23 @@ module Command
       end
 
       Cpl::Cli.start(["apply-template", *templates, "-a", config.app])
+
+      return if config.options[:skip_secret_access_binding]
+
+      progress.puts
+
+      if cp.fetch_identity(app_identity).nil? || cp.fetch_policy(app_secrets_policy).nil?
+        raise "Can't bind identity to policy: identity '#{app_identity}' or " \
+              "policy '#{app_secrets_policy}' doesn't exist. " \
+              "Please create them or use `--skip-secret-access-binding` to ignore this message." \
+              "You can also set a custom secrets name with `secrets_name` " \
+              "and a custom secrets policy name with `secrets_policy_name` " \
+              "in the `.controlplane/controlplane.yml` file."
+      end
+
+      step("Binding identity to policy") do
+        cp.bind_identity_to_policy(app_identity_link, app_secrets_policy)
+      end
     end
   end
 end

data/lib/constants/exit_code.rb

@@ -0,0 +1,7 @@
+# frozen_string_literal: true
+
+module ExitCode
+  SUCCESS = 0
+  ERROR_DEFAULT = 64
+  INTERRUPT = 130
+end

data/lib/core/config.rb

@@ -34,6 +34,10 @@ class Config # rubocop:disable Metrics/ClassLength
     @app ||= load_app_from_options || load_app_from_env
   end
 
+  def app_prefix
+    current&.fetch(:name)
+  end
+
   def location
     @location ||= load_location_from_options || load_location_from_env || load_location_from_file
   end
@@ -111,8 +115,12 @@ class Config # rubocop:disable Metrics/ClassLength
 
   def find_app_config(app_name1)
     @app_configs ||= {}
-    @app_configs[app_name1] ||= apps.find do |app_name2, app_config|
-                                  app_matches?(app_name1, app_name2, app_config)
+
+    @app_configs[app_name1] ||= apps.filter_map do |app_name2, app_config|
+                                  next unless app_matches?(app_name1, app_name2, app_config)
+
+                                  app_config[:name] = app_name2
+                                  app_config
                                 end&.last
   end
 
@@ -166,7 +174,7 @@ class Config # rubocop:disable Metrics/ClassLength
   end
 
   def config_file_path # rubocop:disable Metrics/MethodLength
-    @config_file_path ||= begin
+    @config_file_path ||= ENV["CONFIG_FILE_PATH"] || begin
       path = Pathname.new(".").expand_path
 
       loop do