ood_core 0.11.1 → 0.13.0

data/lib/ood_core/job/adapters/sge/qstat_xml_r_listener.rb

@@ -24,21 +24,32 @@ class QstatXmlRListener
     @parsed_jobs = []
     @current_job = {
       :tasks => [],
-      :native => {}  # TODO: improve native reporting
+      :native => {
+        :ST_name => ''
+      } 
     }
     @current_text = nil
+    @processing_JB_stdout_path_list = false
 
     @current_request = nil
+    @native_tags = ['JB_job_number', 'JB_job_name', 'JB_version', 'JB_project', 'JB_exec_file', 'JB_script_file', 'JB_script_size', 'JB_submission_time', 'JB_execution_time', 'JB_deadline', 'JB_owner', 'JB_uid', 'JB_group', 'JB_gid', 'JB_account', 'JB_cwd', 'JB_notify', 'JB_type', 'JB_reserve', 'JB_priority', 'JB_jobshare', 'JB_verify', 'JB_checkpoint_attr', 'JB_checkpoint_interval', 'JB_restart']
   end
 
   def tag_start(name, attributes)
     case name
     when 'hard_request'
       start_hard_request(attributes)
+    when "JB_stdout_path_list"
+      @processing_JB_stdout_path_list = true
     end
   end
 
   def tag_end(name)
+    #Add text if in native_tags
+    if (@native_tags.include?(name))
+      @current_job[:native][:"#{name}"] = @current_text
+    end
+    
     case name
     when 'job_list'
       end_job_list
@@ -64,6 +75,10 @@ class QstatXmlRListener
       end_hard_request
     when 'tasks'
       add_child_tasks
+    when 'PN_path'
+      end_PN_path
+    when 'ST_name'
+      end_ST_name
     end
   end
 
@@ -130,6 +145,15 @@ class QstatXmlRListener
     end
   end
 
+  def end_PN_path
+    @current_job[:native][:PN_path] = @current_text if @processing_JB_stdout_path_list
+    @processing_JB_stdout_path_list = false
+  end
+
+  def end_ST_name
+    @current_job[:native][:ST_name] = @current_job[:native][:ST_name] + @current_text + ' '
+  end
+
   # Store a completed job and reset current_job for the next pass
   def end_job_list
     @parsed_jobs << @current_job
@@ -145,4 +169,3 @@ class QstatXmlRListener
     }
   end
 end
-

data/lib/ood_core/job/adapters/slurm.rb

@@ -14,13 +14,17 @@ module OodCore
       # @option config [Object] :conf (nil) Path to the slurm conf
       # @option config [Object] :bin (nil) Path to slurm client binaries
       # @option config [#to_h] :bin_overrides ({}) Optional overrides to Slurm client executables
+      # @option config [Object] :submit_host ("") Submit job on login node via ssh
+      # @option config [Object] :strict_host_checking (true) Whether to use strict host checking when ssh to submit_host
       def self.build_slurm(config)
         c = config.to_h.symbolize_keys
-        cluster = c.fetch(:cluster, nil)
-        conf    = c.fetch(:conf, nil)
-        bin     = c.fetch(:bin, nil)
-        bin_overrides = c.fetch(:bin_overrides, {})
-        slurm = Adapters::Slurm::Batch.new(cluster: cluster, conf: conf, bin: bin, bin_overrides: bin_overrides)
+        cluster              = c.fetch(:cluster, nil)
+        conf                 = c.fetch(:conf, nil)
+        bin                  = c.fetch(:bin, nil)
+        bin_overrides        = c.fetch(:bin_overrides, {})
+        submit_host          = c.fetch(:submit_host, "")
+        strict_host_checking = c.fetch(:strict_host_checking, true)
+        slurm = Adapters::Slurm::Batch.new(cluster: cluster, conf: conf, bin: bin, bin_overrides: bin_overrides, submit_host: submit_host, strict_host_checking: strict_host_checking)
         Adapters::Slurm.new(slurm: slurm)
       end
     end
@@ -62,6 +66,16 @@ module OodCore
           # @return Hash<String, String>
           attr_reader :bin_overrides
 
+          # The login node where the job is submitted via ssh
+          # @example owens.osc.edu
+          # @return [String] The login node
+          attr_reader :submit_host
+
+          # Wheter to use strict host checking when ssh to submit_host
+          # @example false
+          # @return [Bool]; true if empty
+          attr_reader :strict_host_checking
+
           # The root exception class that all Slurm-specific exceptions inherit
           # from
           class Error < StandardError; end
@@ -69,11 +83,16 @@ module OodCore
           # @param cluster [#to_s, nil] the cluster name
           # @param conf [#to_s, nil] path to the slurm conf
           # @param bin [#to_s] path to slurm installation binaries
-          def initialize(cluster: nil, bin: nil, conf: nil, bin_overrides: {})
-            @cluster = cluster && cluster.to_s
-            @conf    = conf    && Pathname.new(conf.to_s)
-            @bin     = Pathname.new(bin.to_s)
-            @bin_overrides = bin_overrides
+          # @param bin_overrides [#to_h] a hash of bin ovverides to be used in job
+          # @param submit_host [#to_s] Submits the job on a login node via ssh
+          # @param strict_host_checking [Bool] Whether to use strict host checking when ssh to submit_host
+          def initialize(cluster: nil, bin: nil, conf: nil, bin_overrides: {}, submit_host: "", strict_host_checking: true)
+            @cluster              = cluster && cluster.to_s
+            @conf                 = conf    && Pathname.new(conf.to_s)
+            @bin                  = Pathname.new(bin.to_s)
+            @bin_overrides        = bin_overrides
+            @submit_host          = submit_host.to_s
+            @strict_host_checking = strict_host_checking
           end
 
           # Get a list of hashes detailing each of the jobs on the batch server
@@ -148,9 +167,9 @@ module OodCore
           #TODO: write some barebones test for this? like 2 options and id or no id
           def squeue_args(id: "", owner: nil, options: [])
             args  = ["--all", "--states=all", "--noconvert"]
-            args += ["-o", "#{RECORD_SEPARATOR}#{options.join(UNIT_SEPARATOR)}"]
-            args += ["-u", owner.to_s] unless owner.to_s.empty?
-            args += ["-j", id.to_s] unless id.to_s.empty?
+            args.concat ["-o", "#{RECORD_SEPARATOR}#{options.join(UNIT_SEPARATOR)}"]
+            args.concat ["-u", owner.to_s] unless owner.to_s.empty?
+            args.concat ["-j", id.to_s] unless id.to_s.empty?
             args
           end
 
@@ -275,11 +294,15 @@ module OodCore
             # Call a forked Slurm command for a given cluster
             def call(cmd, *args, env: {}, stdin: "")
               cmd = OodCore::Job::Adapters::Helper.bin_path(cmd, bin, bin_overrides)
+
               args  = args.map(&:to_s)
-              args += ["-M", cluster] if cluster
+              args.concat ["-M", cluster] if cluster
+
               env = env.to_h
               env["SLURM_CONF"] = conf.to_s if conf
-              o, e, s = Open3.capture3(env, cmd, *args, stdin_data: stdin.to_s)
+
+              cmd, args = OodCore::Job::Adapters::Helper.ssh_wrap(submit_host, cmd, args, strict_host_checking)
+              o, e, s = Open3.capture3(env, cmd, *(args.map(&:to_s)), stdin_data: stdin.to_s)
               s.success? ? o : raise(Error, e)
             end
 
@@ -358,30 +381,31 @@ module OodCore
           # Set sbatch options
           args = []
           # ignore args, don't know how to do this for slurm
-          args += ["-H"] if script.submit_as_hold
-          args += (script.rerunnable ? ["--requeue"] : ["--no-requeue"]) unless script.rerunnable.nil?
-          args += ["-D", script.workdir.to_s] unless script.workdir.nil?
-          args += ["--mail-user", script.email.join(",")] unless script.email.nil?
+          args.concat ["-H"] if script.submit_as_hold
+          args.concat (script.rerunnable ? ["--requeue"] : ["--no-requeue"]) unless script.rerunnable.nil?
+          args.concat ["-D", script.workdir.to_s] unless script.workdir.nil?
+          args.concat ["--mail-user", script.email.join(",")] unless script.email.nil?
           if script.email_on_started && script.email_on_terminated
-            args += ["--mail-type", "ALL"]
+            args.concat ["--mail-type", "ALL"]
           elsif script.email_on_started
-            args += ["--mail-type", "BEGIN"]
+            args.concat ["--mail-type", "BEGIN"]
           elsif script.email_on_terminated
-            args += ["--mail-type", "END"]
+            args.concat ["--mail-type", "END"]
           elsif script.email_on_started == false && script.email_on_terminated == false
-            args += ["--mail-type", "NONE"]
+            args.concat ["--mail-type", "NONE"]
           end
-          args += ["-J", script.job_name] unless script.job_name.nil?
-          args += ["-i", script.input_path] unless script.input_path.nil?
-          args += ["-o", script.output_path] unless script.output_path.nil?
-          args += ["-e", script.error_path] unless script.error_path.nil?
-          args += ["--reservation", script.reservation_id] unless script.reservation_id.nil?
-          args += ["-p", script.queue_name] unless script.queue_name.nil?
-          args += ["--priority", script.priority] unless script.priority.nil?
-          args += ["--begin", script.start_time.localtime.strftime("%C%y-%m-%dT%H:%M:%S")] unless script.start_time.nil?
-          args += ["-A", script.accounting_id] unless script.accounting_id.nil?
-          args += ["-t", seconds_to_duration(script.wall_time)] unless script.wall_time.nil?
-          args += ['-a', script.job_array_request] unless script.job_array_request.nil?
+          args.concat ["-J", script.job_name] unless script.job_name.nil?
+          args.concat ["-i", script.input_path] unless script.input_path.nil?
+          args.concat ["-o", script.output_path] unless script.output_path.nil?
+          args.concat ["-e", script.error_path] unless script.error_path.nil?
+          args.concat ["--reservation", script.reservation_id] unless script.reservation_id.nil?
+          args.concat ["-p", script.queue_name] unless script.queue_name.nil?
+          args.concat ["--priority", script.priority] unless script.priority.nil?
+          args.concat ["--begin", script.start_time.localtime.strftime("%C%y-%m-%dT%H:%M:%S")] unless script.start_time.nil?
+          args.concat ["-A", script.accounting_id] unless script.accounting_id.nil?
+          args.concat ["-t", seconds_to_duration(script.wall_time)] unless script.wall_time.nil?
+          args.concat ['-a', script.job_array_request] unless script.job_array_request.nil?
+          args.concat ['--qos', script.qos] unless script.qos.nil?
           # ignore nodes, don't know how to do this for slurm
 
           # Set dependencies
@@ -390,17 +414,14 @@ module OodCore
           depend << "afterok:#{afterok.join(":")}"       unless afterok.empty?
           depend << "afternotok:#{afternotok.join(":")}" unless afternotok.empty?
           depend << "afterany:#{afterany.join(":")}"     unless afterany.empty?
-          args += ["-d", depend.join(",")]               unless depend.empty?
+          args.concat ["-d", depend.join(",")]               unless depend.empty?
 
           # Set environment variables
           env = script.job_environment || {}
-          unless (script.job_environment.nil? || script.job_environment.empty?)
-            prefix = script.copy_environment? ? "ALL," : "NONE,"  # NONE if false or nil
-            args += ["--export", prefix + script.job_environment.keys.join(",")]
-          end
+          args.concat ["--export", export_arg(env, script.copy_environment?)]
 
           # Set native options
-          args += script.native if script.native
+          args.concat script.native if script.native
 
           # Set content
           content = if script.shell_path.nil?
@@ -630,6 +651,25 @@ module OodCore
 
             Info.new(**parent_task_hash)
           end
+
+
+          # we default to export NONE, but SLURM defaults to ALL.
+          # we do this bc SLURM setups a new environment, loading /etc/profile
+          # and all giving 'module' function (among other things shells give),
+          # where the PUN did not.
+          # --export=ALL export the PUN's environment.
+          def export_arg(env, copy_environment)
+            if !env.empty? && !copy_environment
+              env.keys.join(",")
+            elsif !env.empty? && copy_environment
+              "ALL," + env.keys.join(",")
+            elsif env.empty? && copy_environment
+              # only this option changes behaivor dramatically
+              "ALL"
+            else
+              "NONE"
+            end
+          end
       end
     end
   end

data/lib/ood_core/job/adapters/torque.rb

@@ -1,5 +1,6 @@
 require "ood_core/refinements/hash_extensions"
 require "ood_core/job/adapters/helper"
+require 'shellwords'
 
 module OodCore
   module Job
@@ -9,16 +10,18 @@ module OodCore
       # Build the Torque adapter from a configuration
       # @param config [#to_h] the configuration for job adapter
       # @option config [#to_s] :host The batch server host
+      # @option config [#to_s] :submit_host The login node to submit the job via ssh
       # @option config [#to_s] :lib ('') Path to torque client libraries
       # @option config [#to_s] :bin ('') Path to torque client binaries
       # @option config [#to_h] :custom_bin ({}) Optional overrides to Torque client executables
       def self.build_torque(config)
         c = config.to_h.symbolize_keys
         host = c.fetch(:host) { raise ArgumentError, "No host specified. Missing argument: host" }.to_s
+        submit_host = c.fetch(:submit_host, "").to_s
         lib  = c.fetch(:lib, "").to_s
         bin  = c.fetch(:bin, "").to_s
         custom_bin = c.fetch(:custom_bin, {})
-        pbs  = Adapters::Torque::Batch.new(host: host, lib: lib, bin: bin, custom_bin: custom_bin)
+        pbs  = Adapters::Torque::Batch.new(host: host, submit_host: submit_host, lib: lib, bin: bin, custom_bin: custom_bin)
         Adapters::Torque.new(pbs: pbs)
       end
     end
@@ -85,7 +88,7 @@ module OodCore
           depend << "afterany:#{afterany.join(':')}"     unless afterany.empty?
 
           # Set mailing options
-          mail_points  = ""
+          mail_points = ""
           mail_points += "b" if script.email_on_started
           mail_points += "e" if script.email_on_terminated
 
@@ -129,40 +132,44 @@ module OodCore
               envvars.merge!   script.native.fetch(:envvars, {})
             end
 
+            # Destructively change envvars to shellescape values
+            envvars.transform_values! { |v| Shellwords.escape(v) }
+
             # Submit job
             @pbs.submit_string(script.content, queue: script.queue_name, headers: headers, resources: resources, envvars: envvars)
           else
             # Set qsub arguments
             args = []
-            args += ["-F", script.args.join(" ")] unless script.args.nil?
-            args += ["-h"] if script.submit_as_hold
-            args += ["-r", script.rerunnable ? "y" : "n"] unless script.rerunnable.nil?
-            args += ["-M", script.email.join(",")] unless script.email.nil?
-            args += ["-m", mail_points] unless mail_points.empty?
-            args += ["-N", script.job_name] unless script.job_name.nil?
-            args += ["-S", script.shell_path] unless script.shell_path.nil?
+            args.concat ["-F", script.args.join(" ")] unless script.args.nil?
+            args.concat ["-h"] if script.submit_as_hold
+            args.concat ["-r", script.rerunnable ? "y" : "n"] unless script.rerunnable.nil?
+            args.concat ["-M", script.email.join(",")] unless script.email.nil?
+            args.concat ["-m", mail_points] unless mail_points.empty?
+            args.concat ["-N", script.job_name] unless script.job_name.nil?
+            args.concat ["-S", script.shell_path] unless script.shell_path.nil?
             # ignore input_path (not defined in Torque)
-            args += ["-o", script.output_path] unless script.output_path.nil?
-            args += ["-e", script.error_path] unless script.error_path.nil?
-            args += ["-W", "x=advres:#{script.reservation_id}"] unless script.reservation_id.nil?
-            args += ["-q", script.queue_name] unless script.queue_name.nil?
-            args += ["-p", script.priority] unless script.priority.nil?
-            args += ["-a", script.start_time.localtime.strftime("%C%y%m%d%H%M.%S")] unless script.start_time.nil?
-            args += ["-A", script.accounting_id] unless script.accounting_id.nil?
-            args += ["-W", "depend=#{depend.join(",")}"] unless depend.empty?
-            args += ["-l", "walltime=#{seconds_to_duration(script.wall_time)}"] unless script.wall_time.nil?
-            args += ['-t', script.job_array_request] unless script.job_array_request.nil?
+            args.concat ["-o", script.output_path] unless script.output_path.nil?
+            args.concat ["-e", script.error_path] unless script.error_path.nil?
+            args.concat ["-W", "x=advres:#{script.reservation_id}"] unless script.reservation_id.nil?
+            args.concat ["-q", script.queue_name] unless script.queue_name.nil?
+            args.concat ["-p", script.priority] unless script.priority.nil?
+            args.concat ["-a", script.start_time.localtime.strftime("%C%y%m%d%H%M.%S")] unless script.start_time.nil?
+            args.concat ["-A", script.accounting_id] unless script.accounting_id.nil?
+            args.concat ["-W", "depend=#{depend.join(",")}"] unless depend.empty?
+            args.concat ["-l", "walltime=#{seconds_to_duration(script.wall_time)}"] unless script.wall_time.nil?
+            args.concat ['-t', script.job_array_request] unless script.job_array_request.nil?
+            args.concat ['-l', "qos=#{script.qos}"] unless script.qos.nil?
             # Set environment variables
             env = script.job_environment.to_h
-            args += ["-v", env.keys.join(",")] unless env.empty?
-            args += ["-V"] if script.copy_environment?
+            args.concat ["-v", env.keys.join(",")] unless env.empty?
+            args.concat ["-V"] if script.copy_environment?
 
             # If error_path is not specified we join stdout & stderr (as this
             # mimics what the other resource managers do)
-            args += ["-j", "oe"] if script.error_path.nil?
+            args.concat ["-j", "oe"] if script.error_path.nil?
 
             # Set native options
-            args += script.native if script.native
+            args.concat script.native if script.native
 
             # Submit job
             @pbs.submit(script.content, args: args, env: env, chdir: script.workdir)

data/lib/ood_core/job/adapters/torque/batch.rb

@@ -9,6 +9,18 @@ class OodCore::Job::Adapters::Torque
     # @return [String] the batch server host
     attr_reader :host
 
+    # The login node where job is submitted via ssh
+    # @example OSC's owens login node
+    #   my_conn.submit_host #=> "owens.osc.edu"
+    # @return [String] the login node
+    attr_reader :submit_host
+
+    # Determines whether to use strict_host_checking for ssh
+    # @example 
+    #   my_conn.strict_host_checking.to_s #=> "owens.osc.edu"
+    # @return [Bool] 
+    attr_reader :strict_host_checking
+
     # The path to the Torque client installation libraries
     # @example For Torque 5.0.0
     #   my_conn.lib.to_s #=> "/usr/local/Torque/5.0.0/lib"
@@ -32,19 +44,23 @@ class OodCore::Job::Adapters::Torque
     class Error < StandardError; end
 
     # @param host [#to_s] the batch server host
+    # @param submit_host [#to_s] the login node
+    # @param strict_host_checking [bool] use strict host checking when ssh to submit_host
     # @param lib [#to_s] path to FFI installation libraries
     # @param bin [#to_s] path to FFI installation binaries
-    def initialize(host:, lib: "", bin: "", bin_overrides: {}, **_)
-      @host    = host.to_s
-      @lib     = Pathname.new(lib.to_s)
-      @bin     = Pathname.new(bin.to_s)
-      @bin_overrides = bin_overrides
+    def initialize(host:, submit_host: "", strict_host_checking: true, lib: "", bin: "", bin_overrides: {}, **_)
+      @host                 = host.to_s
+      @submit_host          = submit_host.to_s
+      @strict_host_checking = strict_host_checking
+      @lib                  = Pathname.new(lib.to_s)
+      @bin                  = Pathname.new(bin.to_s)
+      @bin_overrides        = bin_overrides
     end
 
     # Convert object to hash
     # @return [Hash] the hash describing this object
     def to_h
-      {host: host, lib: lib, bin: bin}
+      {host: host, submit_host: submit_host, strict_host_checking: strict_host_checking, lib: lib, bin: bin}
     end
 
     # The comparison operator
@@ -437,10 +453,10 @@ class OodCore::Job::Adapters::Torque
       # NB: The binary includes many useful filters and is preferred
       def qsub_submit(script, queue, headers, resources, envvars)
         params  = []
-        params += ["-q", "#{queue}"] unless queue.empty?
-        params += headers.map {|k,v| qsub_arg(k,v)}.flatten
-        params += resources.map{|k,v| ["-l", "#{k}=#{v}"]}.flatten
-        params += ["-v", envvars.map{|k,v| "#{k}=#{v}"}.join(",")] unless envvars.empty?
+        params.concat ["-q", "#{queue}"] unless queue.empty?
+        params.concat headers.map {|k,v| qsub_arg(k,v)}.flatten
+        params.concat resources.map{|k,v| ["-l", "#{k}=#{v}"]}.flatten
+        params.concat ["-v", envvars.map{|k,v| "#{k}=#{v}"}.join(",")] unless envvars.empty?
         params << script
 
         env = {
@@ -448,6 +464,7 @@ class OodCore::Job::Adapters::Torque
           "LD_LIBRARY_PATH" => "#{lib}:#{ENV['LD_LIBRARY_PATH']}"
         }
         cmd = OodCore::Job::Adapters::Helper.bin_path('qsub', bin, bin_overrides)
+        cmd, params = OodCore::Job::Adapters::Helper.ssh_wrap(submit_host, cmd, params, strict_host_checking, env)
         o, e, s = Open3.capture3(env, cmd, *params)
         raise Error, e unless s.success?
         o.chomp
@@ -456,14 +473,14 @@ class OodCore::Job::Adapters::Torque
       # Call a forked PBS command for a given host
       def call(cmd, *args, env: {}, stdin: "", chdir: nil)
         cmd = OodCore::Job::Adapters::Helper.bin_path(cmd, bin, bin_overrides)
-        args = args.map(&:to_s)
         env  = env.to_h.each_with_object({}) {|(k,v), h| h[k.to_s] = v.to_s}.merge({
           "PBS_DEFAULT"     => host,
           "LD_LIBRARY_PATH" => %{#{lib}:#{ENV["LD_LIBRARY_PATH"]}}
         })
+        cmd, args = OodCore::Job::Adapters::Helper.ssh_wrap(submit_host, cmd, args, strict_host_checking, env)
         stdin = stdin.to_s
         chdir ||= "."
-        o, e, s = Open3.capture3(env, cmd, *args, stdin_data: stdin, chdir: chdir.to_s)
+        o, e, s = Open3.capture3(env, cmd, *(args.map(&:to_s)), stdin_data: stdin, chdir: chdir.to_s)
         s.success? ? o : raise(Error, e)
       end
   end

data/lib/ood_core/job/script.rb

@@ -99,6 +99,10 @@ module OodCore
       # @return [String, nil] job array request
       attr_reader :job_array_request
 
+      # The qos selected for the job
+      # @return [String, nil] qos
+      attr_reader :qos
+
       # Object detailing any native specifications that are implementation specific
       # @note Should not be used at all costs.
       # @return [Object, nil] native specifications
@@ -130,7 +134,10 @@ module OodCore
       # @param start_time [#to_i, nil] eligible start time
       # @param wall_time [#to_i, nil] max real time
       # @param accounting_id [#to_s, nil] accounting id
+      # @param job_array_request [#to_s, nil] job array request
+      # @param qos [#to_s, nil] qos
       # @param native [Object, nil] native specifications
+      # @param copy_environment [Boolean, nil] copy the environment
       def initialize(content:, args: nil, submit_as_hold: nil, rerunnable: nil,
                      job_environment: nil, workdir: nil, email: nil,
                      email_on_started: nil, email_on_terminated: nil,
@@ -138,7 +145,7 @@ module OodCore
                      output_path: nil, error_path: nil, reservation_id: nil,
                      queue_name: nil, priority: nil, start_time: nil,
                      wall_time: nil, accounting_id: nil, job_array_request: nil,
-                     native: nil, copy_environment: nil, **_)
+                     qos: nil, native: nil, copy_environment: nil, **_)
         @content = content.to_s
 
         @submit_as_hold      = submit_as_hold
@@ -162,6 +169,7 @@ module OodCore
         @wall_time          = wall_time         && wall_time.to_i
         @accounting_id      = accounting_id     && accounting_id.to_s
         @job_array_request  = job_array_request && job_array_request.to_s
+        @qos                = qos               && qos.to_s
         @native             = native
         @copy_environment   = (copy_environment.nil?) ? nil : !! copy_environment
       end
@@ -191,6 +199,7 @@ module OodCore
           wall_time:           wall_time,
           accounting_id:       accounting_id,
           job_array_request:   job_array_request,
+          qos:                 qos,
           native:              native,
           copy_environment:    copy_environment
         }