ood_core 0.16.1 → 0.17.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 19665b6db28d01da39093dc90d4a5023ca12264f07b932aebc8ec8c443bafa25
-  data.tar.gz: d9c8c6d8f30851ea9138c8325aafd750823534a51f36601a20366265ac4feec2
+  metadata.gz: 266e24cd2e11c2c712b6128bdc3f82bf9781ae9885f8f0eb21439cb80c889c90
+  data.tar.gz: 01e682b6313468371076cdfd4ff2df2f3c06c661af9b4d7d7a65b7dcf3e2d836
 SHA512:
-  metadata.gz: 1ed1eaa873366ad5e825ed29c7401dd3bca4a424ab7a689a19479f297ec20d7e019cd53609006b0919a365dd0002eb0c1e9c0cabcc9f69579cf7ae81b33b3ae7
-  data.tar.gz: 90a4cfa3ee8b1f76ef7e1f28df6d8e64725d1eaff005b4bd4ff7fc8f88e5bfda8a15e706636c18e7b5ac74451071eaea4e6814945ea25e95f6c7ed2de8fd2fec
+  metadata.gz: 22721c9d368ec44533d93914f977576ee77786ad0926976fa24067c7f353104edd32baffac34723b730e7f711c5b7581cf4f72f6232d050c719edcfc1b3cb14f
+  data.tar.gz: 5d99a3c782aad5e420333653e51073cdb6535b7689053452da7ee7ef234292769e70d4010d751aacd4576daf531700136c2aa0ef7a2efc31edcb29e4c01c8be4

data/CHANGELOG.md

@@ -6,6 +6,37 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/)
 and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
+
+## [0.17.0] - 5-26-2021
+
+### Fixed
+
+- All Kubernetes resources now have the same labels in [280](https://github.com/OSC/ood_core/pull/280).
+- Kubernetes does not crash when no configmap is defined in [282](https://github.com/OSC/ood_core/pull/282).
+- Kubernetes will not specify init containers if there are none in
+  [284](https://github.com/OSC/ood_core/pull/284).
+
+### Added
+
+- Kubernetes, Slurm and Torque now support the script option `gpus_per_node` in
+  [266](https://github.com/OSC/ood_core/pull/266).
+- Kubernetes will now save the pod.yml into the staged root in
+  [277](https://github.com/OSC/ood_core/pull/277).
+- Kubernetes now allows for node selector in [264](https://github.com/OSC/ood_core/pull/264).
+- Kubernetes pods now have access the environment variable POD_NAMESPACE in
+  [275](https://github.com/OSC/ood_core/pull/275).
+- Kubernetes pods can now specify the image pull policy in [272](https://github.com/OSC/ood_core/pull/272).
+- Cluster config's batch_connect now support `ssh_allow` to disable sshing to compute
+  nodes per cluster in [286](https://github.com/OSC/ood_core/pull/286).
+- Kubernetes will now add the templated script content to a configmap in
+  [273](https://github.com/OSC/ood_core/pull/273).
+
+### Changed
+
+- Kubernetes username prefix no longer appends a - in [271](https://github.com/OSC/ood_core/pull/271).
+
+
+
 ## [0.16.1] - 2021-04-23
 ### Fixed
 - memorized some allow? variables to have better support around ACLS in
@@ -305,7 +336,8 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 ### Added
 - Initial release!
 
-[Unreleased]: https://github.com/OSC/ood_core/compare/v0.16.1...HEAD
+[Unreleased]: https://github.com/OSC/ood_core/compare/v0.16.0...HEAD
+[0.17.0]: https://github.com/OSC/ood_core/compare/v0.16.1...v0.17.0
 [0.16.1]: https://github.com/OSC/ood_core/compare/v0.16.0...v0.16.1
 [0.16.0]: https://github.com/OSC/ood_core/compare/v0.15.1...v0.16.0
 [0.15.1]: https://github.com/OSC/ood_core/compare/v0.15.0...v0.15.1

data/lib/ood_core/cluster.rb

@@ -147,6 +147,15 @@ module OodCore
       @allow = acls.all?(&:allow?)
     end
 
+    # Whether this cluster supports SSH to batch connect nodes
+    # @return [Boolean] whether cluster supports SSH to batch connect node
+    def batch_connect_ssh_allow?
+      return @batch_connect_ssh_allow if defined?(@batch_connect_ssh_allow)
+      return @batch_connect_ssh_allow = true if batch_connect_config.nil?
+
+      @batch_connect_ssh_allow = batch_connect_config.fetch(:ssh_allow, true)
+    end
+
     # The comparison operator
     # @param other [#to_sym] object to compare against
     # @return [Boolean] whether objects are equivalent

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

@@ -23,7 +23,7 @@ class OodCore::Job::Adapters::Kubernetes::Batch
     @cluster = options.fetch(:cluster, 'open-ondemand')
     @mounts = options.fetch(:mounts, []).map { |m| m.to_h.symbolize_keys }
     @all_namespaces = options.fetch(:all_namespaces, false)
-    @username_prefix = options.fetch(:username_prefix, nil)
+    @username_prefix = options.fetch(:username_prefix, '')
     @namespace_prefix = options.fetch(:namespace_prefix, '')
 
     @using_context = false
@@ -45,6 +45,9 @@ class OodCore::Job::Adapters::Kubernetes::Batch
     raise ArgumentError, 'Must specify the script' if script.nil?
 
     resource_yml, id = generate_id_yml(script)
+    if !script.workdir.nil? && Dir.exist?(script.workdir)
+      File.open(File.join(script.workdir, 'pod.yml'), 'w') { |f| f.write resource_yml }
+    end
     call("#{formatted_ns_cmd} create -f -", stdin: resource_yml)
 
     id
@@ -146,7 +149,7 @@ class OodCore::Job::Adapters::Kubernetes::Batch
   end
 
   def k8s_username
-    username_prefix.nil? ? username : "#{username_prefix}-#{username}"
+    "#{username_prefix}#{username}"
   end
 
   def user
@@ -189,10 +192,12 @@ class OodCore::Job::Adapters::Kubernetes::Batch
     native_data = script.native
     container = helper.container_from_native(native_data[:container], default_env)
     id = generate_id(container.name)
-    configmap = helper.configmap_from_native(native_data, id)
+    configmap = helper.configmap_from_native(native_data, id, script.content)
     init_containers = helper.init_ctrs_from_native(native_data[:init_containers], container.env)
     spec = OodCore::Job::Adapters::Kubernetes::Resources::PodSpec.new(container, init_containers: init_containers)
     all_mounts = native_data[:mounts].nil? ? mounts : mounts + native_data[:mounts]
+    node_selector = native_data[:node_selector].nil? ? {} : native_data[:node_selector]
+    gpu_type = native_data[:gpu_type].nil? ? "nvidia.com/gpu" : native_data[:gpu_type]
 
     template = ERB.new(File.read(resource_file), nil, '-')
 

data/lib/ood_core/job/adapters/kubernetes/helper.rb

@@ -53,6 +53,7 @@ class OodCore::Job::Adapters::Kubernetes::Helper
       cpu: container[:cpu],
       working_dir: container[:working_dir],
       restart_policy: container[:restart_policy],
+      image_pull_policy: container[:image_pull_policy],
       image_pull_secret: container[:image_pull_secret]
     )
   end
@@ -80,10 +81,18 @@ class OodCore::Job::Adapters::Kubernetes::Helper
   #   the input configmap hash
   # @param id [#to_s]
   #   the id to use for giving the configmap a name
+  # @param script_content [#to_s]
+  #   the batch script content
   # @return  [OodCore::Job::Adapters::Kubernetes::Resources::ConfigMap]
-  def configmap_from_native(native, id)
-    configmap = native.fetch(:configmap, nil)
-    return nil if configmap.nil?
+  def configmap_from_native(native, id, script_content)
+    configmap = native.fetch(:configmap, {})
+    configmap[:files] ||= []
+    configmap[:files] << {
+      filename: 'script.sh',
+      data: script_content,
+      mount_path: '/ood/script.sh',
+      sub_path: 'script.sh',
+    } unless configmap[:files].any? { |f| f[:filename] == 'script.sh' }
 
     OodCore::Job::Adapters::Kubernetes::Resources::ConfigMap.new(
       configmap_name(id),

data/lib/ood_core/job/adapters/kubernetes/resources.rb

@@ -35,11 +35,11 @@ module OodCore::Job::Adapters::Kubernetes::Resources
 
   class Container
     attr_accessor :name, :image, :command, :port, :env, :memory, :cpu, :working_dir,
-                  :restart_policy, :image_pull_secret, :supplemental_groups
+                  :restart_policy, :image_pull_policy, :image_pull_secret, :supplemental_groups
 
     def initialize(
         name, image, command: [], port: nil, env: {}, memory: "4Gi", cpu: "1",
-        working_dir: "", restart_policy: "Never", image_pull_secret: nil, supplemental_groups: []
+        working_dir: "", restart_policy: "Never", image_pull_policy: nil, image_pull_secret: nil, supplemental_groups: []
       )
       raise ArgumentError, "containers need valid names and images" unless name && image
 
@@ -52,6 +52,7 @@ module OodCore::Job::Adapters::Kubernetes::Resources
       @cpu = cpu.nil? ? "1" : cpu
       @working_dir = working_dir.nil? ? "" : working_dir
       @restart_policy = restart_policy.nil? ? "Never" : restart_policy
+      @image_pull_policy = image_pull_policy.nil? ? "IfNotPresent" : image_pull_policy
       @image_pull_secret = image_pull_secret
       @supplemental_groups = supplemental_groups.nil? ? [] : supplemental_groups
     end
@@ -66,6 +67,7 @@ module OodCore::Job::Adapters::Kubernetes::Resources
         cpu == other.cpu &&
         working_dir == other.working_dir &&
         restart_policy == other.restart_policy &&
+        image_pull_policy == other.image_pull_policy &&
         image_pull_secret == other.image_pull_secret &&
         supplemental_groups == other.supplemental_groups
     end

data/lib/ood_core/job/adapters/kubernetes/templates/pod.yml.erb

@@ -39,7 +39,7 @@ spec:
   containers:
   - name: "<%= spec.container.name %>"
     image: <%= spec.container.image %>
-    imagePullPolicy: IfNotPresent
+    imagePullPolicy: <%= spec.container.image_pull_policy %>
     <%- unless spec.container.working_dir.empty? -%>
     workingDir: "<%= spec.container.working_dir %>"
     <%- end -%>
@@ -48,6 +48,10 @@ spec:
       valueFrom:
         fieldRef:
           fieldPath: metadata.name
+    - name: POD_NAMESPACE
+      valueFrom:
+        fieldRef:
+          fieldPath: metadata.namespace
     <%- spec.container.env.each_pair do |name, value| -%>
     - name: <%= name %>
       value: "<%= value %>"
@@ -62,8 +66,9 @@ spec:
     ports:
     - containerPort: <%= spec.container.port %>
     <%- end -%>
-    <%- if configmap.mounts? || !all_mounts.empty? -%>
+    <%- if !all_mounts.empty? || (!configmap.nil? && configmap.mounts?) -%>
     volumeMounts:
+    <%- unless configmap.nil? -%>
     <%- configmap.files.each do |file| -%>
       <%- next if file.mount_path.nil? -%>
     - name: configmap-volume
@@ -72,6 +77,7 @@ spec:
       subPath: <%= file.sub_path %>
       <%- end # end unless file.sub_path.nil? -%>
     <%- end # end configmap.files.each -%>
+    <%- end # unless configmap.nil? -%>
     <%- all_mounts.each do |mount| -%>
     - name: <%= mount[:name] %>
       mountPath: <%= mount[:destination_path] %>
@@ -81,25 +87,36 @@ spec:
       limits:
         memory: "<%= spec.container.memory %>"
         cpu: "<%= spec.container.cpu %>"
+        <%- unless script.gpus_per_node.nil? -%>
+        <%= gpu_type %>: <%= script.gpus_per_node %>
+        <%- end -%>
       requests:
         memory: "<%= spec.container.memory %>"
         cpu: "<%= spec.container.cpu %>"
+        <%- unless script.gpus_per_node.nil? -%>
+        <%= gpu_type %>: <%= script.gpus_per_node %>
+        <%- end -%>
     securityContext:
       allowPrivilegeEscalation: false
       capabilities:
         drop:
         - all
       privileged: false
-  <%- unless spec.init_containers.nil? -%>
+  <%- unless spec.init_containers.empty? -%>
   initContainers:
   <%- spec.init_containers.each do |ctr| -%>
   - name: "<%= ctr.name %>"
     image: "<%= ctr.image %>"
+    imagePullPolicy: <%= ctr.image_pull_policy %>
     env:
     - name: POD_NAME
       valueFrom:
         fieldRef:
           fieldPath: metadata.name
+    - name: POD_NAMESPACE
+      valueFrom:
+        fieldRef:
+          fieldPath: metadata.namespace
     <%- ctr.env.each_pair do |name, value| -%>
     - name: <%= name %>
       value: "<%= value %>"
@@ -108,8 +125,9 @@ spec:
     <%- ctr.command.each do |cmd| -%>
     - "<%= cmd %>"
     <%- end # command loop -%>
-    <%- if configmap.init_mounts? || !all_mounts.empty? -%>
+    <%- if !all_mounts.empty? || (!configmap.nil? && configmap.init_mounts?) -%>
     volumeMounts:
+    <%- unless configmap.nil? -%>
     <%- configmap.files.each do |file| -%>
     <%- next if file.init_mount_path.nil? -%>
     - name: configmap-volume
@@ -118,6 +136,7 @@ spec:
       subPath: <%= file.init_sub_path %>
       <%- end # end unless file.sub_path.nil? -%>
     <%- end # end configmap.files.each -%>
+    <%- end # unless configmap.nil? -%>
     <%- all_mounts.each do |mount| -%>
     - name: <%= mount[:name] %>
       mountPath: <%= mount[:destination_path] %>
@@ -152,6 +171,12 @@ spec:
   <%- end # if mount is [host,nfs] -%>
   <%- end # for each mount -%>
   <%- end # (configmap.to_s.empty? || all_mounts.empty?) -%>
+  <%- unless node_selector.empty? -%>
+  nodeSelector:
+  <%- node_selector.each_pair do |key, value| -%>
+    <%= key %>: "<%= value %>"
+  <%- end # node_selector.each_pair -%>
+  <%- end #unless node_selector.empty? -%>
 ---
 <%- unless spec.container.port.nil? -%>
 apiVersion: v1
@@ -161,6 +186,8 @@ metadata:
   namespace: <%= namespace %>
   labels:
     job: <%= id %>
+    app.kubernetes.io/name: <%= container.name %>
+    app.kubernetes.io/managed-by: open-ondemand
 spec:
   selector:
     job: <%= id %>
@@ -170,8 +197,8 @@ spec:
     targetPort: <%= spec.container.port %>
   type: NodePort
 <%- end # end for service -%>
----
 <%- unless configmap.nil? -%>
+---
 apiVersion: v1
 kind: ConfigMap
 metadata:
@@ -179,6 +206,8 @@ metadata:
   namespace: <%= namespace %>
   labels:
     job: <%= id %>
+    app.kubernetes.io/name: <%= container.name %>
+    app.kubernetes.io/managed-by: open-ondemand
 data:
 <%- configmap.files.each do |file| -%>
   <%- next if file.data.nil? || file.filename.nil? -%>

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

@@ -423,6 +423,7 @@ module OodCore
           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?
+          args.concat ['--gpus-per-node', script.gpus_per_node] unless script.gpus_per_node.nil?
           # ignore nodes, don't know how to do this for slurm
 
           # Set dependencies

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

@@ -159,6 +159,8 @@ module OodCore
             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?
+            args.concat ['-l', "gpus=#{script.gpus_per_node}"] unless script.gpus_per_node.nil?
+
             # Set environment variables
             env = script.job_environment.to_h
             args.concat ["-v", env.keys.join(",")] unless env.empty?

data/lib/ood_core/job/script.rb

@@ -103,6 +103,10 @@ module OodCore
       # @return [String, nil] qos
       attr_reader :qos
 
+      # The GPUs per node for the job
+      # @return [Integer, nil] gpus per node
+      attr_reader :gpus_per_node
+
       # Object detailing any native specifications that are implementation specific
       # @note Should not be used at all costs.
       # @return [Object, nil] native specifications
@@ -136,6 +140,7 @@ module OodCore
       # @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 gpus_per_node [#to_i, nil] gpus per node
       # @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,
@@ -145,7 +150,7 @@ 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,
-                     qos: nil, native: nil, copy_environment: nil, **_)
+                     qos: nil, gpus_per_node: nil, native: nil, copy_environment: nil, **_)
         @content = content.to_s
 
         @submit_as_hold      = submit_as_hold
@@ -170,6 +175,7 @@ module OodCore
         @accounting_id      = accounting_id     && accounting_id.to_s
         @job_array_request  = job_array_request && job_array_request.to_s
         @qos                = qos               && qos.to_s
+        @gpus_per_node      = gpus_per_node     && gpus_per_node.to_i
         @native             = native
         @copy_environment   = (copy_environment.nil?) ? nil : !! copy_environment
       end
@@ -200,6 +206,7 @@ module OodCore
           accounting_id:       accounting_id,
           job_array_request:   job_array_request,
           qos:                 qos,
+          gpus_per_node:       gpus_per_node,
           native:              native,
           copy_environment:    copy_environment
         }

data/lib/ood_core/version.rb

@@ -1,4 +1,4 @@
 module OodCore
   # The current version of {OodCore}
-  VERSION = "0.16.1"
+  VERSION = "0.17.0"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: ood_core
 version: !ruby/object:Gem::Version
-  version: 0.16.1
+  version: 0.17.0
 platform: ruby
 authors:
 - Eric Franz
@@ -10,7 +10,7 @@ authors:
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2021-04-23 00:00:00.000000000 Z
+date: 2021-05-26 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ood_support