ood_core 0.30.1 → 0.31.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 98cdd24e7d7a73ee1d0dff56b013fdc3a924197ae7dc168e0b7c940c134d99ab
-  data.tar.gz: b22aa4d3199bf0c7865d934aad23c5f5b9a358f19a6eebe77490076234e20461
+  metadata.gz: d1ad113f64d7ee39779802e7386bf753a2323969302963c58456b51a05f132b5
+  data.tar.gz: 472bcdea0fbd7096171a8338507e0e9c6554d6a4ceea3fb12f46a4a59f48cf7e
 SHA512:
-  metadata.gz: bffd4cbfae260072fc1dfdbef4979932bafcf67946d57ffc9a4afd47e5f21eaf68cb9fecd88e3a1eb1989d5e5274145ddbd025a822e5fb378818c7160702b841
-  data.tar.gz: 893d3473c5bd0e87ed93ec5e0d131175bb92b78f381d732a14e62ecd4eeef2198db4d3f0574c7261dbd1c637bcc3494473df29f2bc4ceecdbfcf3142084ede1d
+  metadata.gz: 6436bdfb152bd592627f6fb710d44fc08e10d240a398199081ffd0ba1e3fd3a6f4fbaf8fb56adb88656bb58557eb6630702baea2e07ab1e5072f9180e02da967
+  data.tar.gz: 12eb878a31d953d7fb4c8bb6a0e67febf42b5c04aa397aa9fb17329d9aae33601e2d7358c3799ede4be5ef8af84fc9cb8d44e022847a4d42f3162788e5db828d

data/lib/ood_core/helpers/openstack.rb

@@ -0,0 +1,107 @@
+require "fog/openstack"
+require "json"
+require "etc"
+
+module OodCore
+  class OpenStackHelper
+    attr_reader :auth_url, :openstack_instance
+
+    def initialize(token_file:, openstack_instance:)
+      @token_file = token_file
+      @openstack_instance = openstack_instance
+      @auth_url = "https://identity.#{openstack_instance}/v3"
+    end
+
+    # Load token data from the token file
+    # @return [Hash] Parsed token JSON or nil if file does not exist
+    def load_token_data
+      return nil unless File.exist?(@token_file)
+      JSON.parse(File.read(@token_file))
+    rescue Errno::ENOENT => e
+      puts "Error loading token: #{e}"
+      nil
+    end
+
+    # Get access token from loaded credentials
+    # @return [String] The token ID
+    def access_token
+      load_token_data&.[]("id")
+    end
+
+    # Get user ID from loaded credentials
+    # @return [String] The user ID
+    def user_id
+      load_token_data&.[]("user_id")
+    end
+
+    # Fetch all projects for the authenticated user
+    # @return [Array<Hash>] Array of project hashes with id and name
+    def fetch_user_projects
+      connection_params = {
+        openstack_auth_url: auth_url,
+        openstack_management_url: auth_url,
+        openstack_auth_token: access_token,
+      }
+      identity = Fog::OpenStack::Identity.new(connection_params)
+      identity.list_user_projects(user_id).body["projects"]
+    end
+
+    # Fetch all flavors across all projects for a user
+    # @return [Array<Array>] Sorted array of [display_string, flavor_name, project_id]
+    def fetch_all_flavors
+      flavors = []
+
+      fetch_user_projects.each do |project|
+        scoped_token = scope_token_to_project(access_token, project['id'])
+
+        compute_connection_params = {
+          openstack_auth_url: auth_url,
+          openstack_project_name: project['name'],
+          openstack_management_url: "https://compute.#{openstack_instance}/v2.1/#{project['id']}",
+          openstack_auth_token: scoped_token,
+        }
+        compute = Fog::OpenStack::Compute.new(compute_connection_params)
+
+        compute.flavors.each do |flavor|
+          flavors << [
+            "#{flavor.name} - #{flavor.vcpus}VCPUS, #{flavor.ram/1024}GB RAM, #{flavor.disk}GB disk",
+            flavor.name,
+            project['id']
+          ]
+        end
+      end
+
+      flavors.sort
+    end
+
+    # Convenience method that returns both projects and flavors
+    # @return [Array] Array containing [projects, flavors]
+    def load_projects_and_flavors
+      [fetch_user_projects, fetch_all_flavors]
+    end
+
+    # Scope token to a specific project
+    # @param access_token [String] The unscoped token ID
+    # @param project_id [String] The project ID to scope to
+    # @return [String] The scoped token ID
+    def scope_token_to_project(access_token, project_id)
+      auth = {
+        "auth": {
+          "identity": {
+            "methods": ["token"],
+            "token": { "id": access_token }
+          },
+          "scope": { "project": { "id": project_id } }
+        }
+      }
+
+      connection_params = {
+        openstack_auth_url: auth_url,
+        openstack_management_url: auth_url,
+        openstack_auth_token: access_token,
+      }
+      identity = Fog::OpenStack::Identity.new(connection_params)
+      identity.tokens.authenticate(auth)
+    end
+  end
+end

data/lib/ood_core/job/adapters/coder/openstack_credentials.rb

@@ -97,6 +97,7 @@ class OpenStackCredentials < CredentialsInterface
       credentials_to_destroy.destroy
     rescue Excon::Error::Forbidden => e
       puts "Error destroying application credentials with id #{os_app_credentials['id']} #{e}"
+      raise JobAdapterError, e.message
     end
   end
 
@@ -115,4 +116,4 @@ class OpenStackCredentials < CredentialsInterface
   def find_os_application_credentials(connection, os_app_credentials)
     connection.application_credentials.find_by_id(os_app_credentials['id'], os_app_credentials['user_id'])
   end
-end
+end

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

@@ -79,10 +79,7 @@ module OodCore
         #   adapters can get by without populating the entire Info object
         # @return [Array<Info>] information describing submitted jobs
         def info_all(attrs: nil)
-        # TODO - implement info all for namespaces?
-          batch.method_missing(attrs: attrs)
-        rescue Batch::Error => e
-          raise JobAdapterError, e.message
+          raise NotImplementedError, 'subclass did not define #info_all'
         end
     
         # Whether the adapter supports job arrays
@@ -125,4 +122,4 @@ module OodCore
       end
     end
   end
-end
+end

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

@@ -1,4 +1,5 @@
 require "time"
+require "json"
 require "ood_core/refinements/hash_extensions"
 require "ood_core/job/adapters/helper"
 
@@ -85,6 +86,50 @@ module OodCore
             @bin_overrides        = bin_overrides
           end
 
+          # Get a ClusterInfo object containing information about the given cluster
+          # @return [ClusterInfo] object containing cluster details
+          def get_cluster_info
+            args = ["-a", "-F", "json"]
+            stdout = call("pbsnodes", *args)
+            node_info = JSON.parse(stdout)
+    
+            # Initialize cluster info values
+            total_nodes     = 0
+            allocated_nodes = 0
+            total_cpus      = 0
+            allocated_cpus  = 0
+            total_gpus      = 0
+            allocated_gpus  = 0 
+
+            nodes = node_info.fetch('nodes', {})
+
+            nodes.each do |_node_name, node|
+              total_nodes += 1
+              resources_avail = node.fetch('resources_available', {})
+              total_cpus += get_node_resource(resources_avail, 'ncpus')
+              total_gpus += get_node_resource(resources_avail, 'ngpus')
+
+              # Resources assigned (currently allocated to jobs)
+              resources_assigned = node.fetch('resources_assigned', {})
+              ncpus_assigned     = get_node_resource(resources_assigned, 'ncpus')
+              ngpus_assigned     = get_node_resource(resources_assigned, 'ngpus')
+
+              allocated_cpus += ncpus_assigned
+              allocated_gpus += ngpus_assigned
+
+              # A node is allocated if at least one CPU has been assigned to a job
+              allocated_nodes += 1 if ncpus_assigned > 0
+            end
+
+            ClusterInfo.new(active_nodes: allocated_nodes,
+                            total_nodes: total_nodes,
+                            active_processors: allocated_cpus,
+                            total_processors: total_cpus,
+                            active_gpus: allocated_gpus,
+                            total_gpus: total_gpus
+            )
+          end
+
           # Get a list of hashes detailing each of the jobs on the batch server
           # @example Status info for all jobs
           #   my_batch.get_jobs
@@ -174,6 +219,13 @@ module OodCore
           end
 
           private
+            # Get a resource value from a node's resources hash, returning 0 if the
+            # resource is not present
+            def get_node_resource(resources, key)
+             val = resources.fetch(key, 0)
+             val.to_i
+            end
+ 
             # Call a forked PBS Pro command for a given batch server
             def call(cmd, *args, env: {}, stdin: "", chdir: nil)
               cmd = cmd.to_s
@@ -299,6 +351,10 @@ module OodCore
           raise JobAdapterError, e.message
         end
 
+        def cluster_info
+          @pbspro.get_cluster_info
+        end
+
         # Retrieve info for all jobs from the resource manager
         # @raise [JobAdapterError] if something goes wrong getting job info
         # @return [Array<Info>] information describing submitted jobs

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

@@ -47,7 +47,7 @@ module OodCore
         # calculated from gres string
         # @return [Integer] the number of gpus in gres
         def self.gpus_from_gres(gres)
-          gres.to_s.scan(/gpu[^(,]*[:=](\d+)/).flatten.map(&:to_i).sum
+          gres.to_s.scan(/gpu[s:]*[\w()-]*[=:]?(\d+)(?:[(,]|$)/).flatten.map(&:to_i).sum
         end
 
         # Object used for simplified communication with a Slurm batch server
@@ -121,19 +121,28 @@ module OodCore
           # Get a ClusterInfo object containing information about the given cluster
           # @return [ClusterInfo] object containing cluster details
           def get_cluster_info
-            node_cpu_info = call("sinfo", "-aho %A/%D/%C").strip.split('/')
+            node_cpu_info = call("sinfo", "-aho %F/%C").strip.split('/').map(&:to_i)
             gres_length = call("sinfo", "-o %G").lines.map(&:strip).map(&:length).max + 2
-            gres_lines = call("sinfo", "-ahNO ,nodehost,gres:#{gres_length},gresused:#{gres_length}")
-                         .lines.uniq.map(&:split)
-            ClusterInfo.new(active_nodes: node_cpu_info[0].to_i,
-                            total_nodes: node_cpu_info[2].to_i,
-                            active_processors: node_cpu_info[3].to_i,
-                            total_processors: node_cpu_info[6].to_i,
+            gres_lines = call("sinfo", "-ahNO", "nodehost,gres:#{gres_length},gresused:#{gres_length},statelong")
+                         .lines.uniq.reject { |line| line.match?(/maint|drain|down/i) }.map(&:split)
+
+            node_info = sinfo_headers.zip(node_cpu_info).to_h
+            ClusterInfo.new(active_nodes: node_info['nodes_allocated'],
+                            total_nodes: node_info['nodes_total'],
+                            active_processors: node_info['cpus_allocated'],
+                            total_processors: node_info['cpus_total'],
                             active_gpus: gres_lines.sum { |line| Slurm.gpus_from_gres(line[2]) },
                             total_gpus: gres_lines.sum { |line| Slurm.gpus_from_gres(line[1]) }
             )
           end
 
+          def sinfo_headers
+            [
+              'nodes_allocated', 'nodes_idle', 'nodes_other', 'nodes_total',
+              'cpus_allocated', 'cpus_idle', 'cpus_other', 'cpus_total'
+            ]
+          end
+
           # Get a list of hashes detailing each of the jobs on the batch server
           # @example Status info for all jobs
           #   my_batch.get_jobs
@@ -568,7 +577,7 @@ module OodCore
           'OOM' => :completed, # OUT_OF_MEMORY
 
           'BOOT_FAIL'     => :completed,
-          'CANCELED'      => :completed,
+          'CANCELLED'     => :completed,
           'COMPLETED'     => :completed,
           'DEADLINE'      => :completed,
           'FAILED'        => :completed,
@@ -870,11 +879,11 @@ module OodCore
             "%02d:%02d:%02d" % [time/3600, time/60%60, time%60]
           end
 
-        # Parse date time string ignoring unknown values returned by Slurm
+        # safely parse date time string, return nil when there are errors.
         def parse_time(date_time)
-          return nil if date_time.empty? || %w[N/A NONE UNKNOWN].include?(date_time.to_s.upcase)
-
           Time.parse(date_time)
+        rescue ArgumentError
+          nil
         end
 
           # Convert host list string to individual nodes
@@ -901,7 +910,7 @@ module OodCore
 
           # Determine state from Slurm state code
           def get_state(st)
-            STATE_MAP.fetch(st, :undetermined)
+            STATE_MAP.fetch(st.split.first, :undetermined)
           end
 
           # Parse hash describing Slurm job status
@@ -928,8 +937,8 @@ module OodCore
               wallclock_time: duration_in_seconds(v[:time_used]),
               wallclock_limit: duration_in_seconds(v[:time_limit]),
               cpu_time: nil,
-              submission_time: v[:submit_time] ? Time.parse(v[:submit_time]) : nil,
-              dispatch_time: (v[:start_time].nil? || v[:start_time] == "N/A") ? nil : Time.parse(v[:start_time]),
+              submission_time: parse_time(v[:submit_time]),
+              dispatch_time: parse_time(v[:start_time]),
               native: v,
               gpus: self.class.gpus_from_gres(v[:gres])
             )

data/lib/ood_core/version.rb

@@ -1,4 +1,4 @@
 module OodCore
   # The current version of {OodCore}
-  VERSION = "0.30.1"
+  VERSION = "0.31.0"
 end

data/lib/ood_core.rb

@@ -4,6 +4,7 @@ require "ood_core/cluster"
 require "ood_core/clusters"
 require "ood_core/invalid_cluster"
 require "ood_core/data_formatter"
+require "ood_core/helpers/openstack"
 
 # The main namespace for ood_core
 module OodCore
@@ -42,4 +43,8 @@ module OodCore
     require "ood_core/batch_connect/template"
     require "ood_core/batch_connect/factory"
   end
+  
+  module Helpers
+    require "ood_core/helpers/openstack"
+  end
 end

data/ood_core.gemspec

@@ -27,11 +27,11 @@ Gem::Specification.new do |spec|
   spec.add_runtime_dependency "fog-openstack", "~> 1.1.5"
   spec.add_runtime_dependency "rexml", "~> 3.2"
   spec.add_development_dependency "bundler", "~> 2.1"
-  spec.add_development_dependency "rake", "~> 13.3.0"
+  spec.add_development_dependency "rake", "~> 13.4.1"
   spec.add_development_dependency "rspec", "~> 3.0"
   spec.add_development_dependency "pry", "~> 0.10"
   spec.add_development_dependency "timecop", "~> 0.8"
   spec.add_development_dependency "climate_control", "~> 1.2.0"
   spec.add_development_dependency "minitest", "~> 5"
-  spec.add_development_dependency "mocha", "~> 2.4"
+  spec.add_development_dependency "mocha", "~> 3.0"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: ood_core
 version: !ruby/object:Gem::Version
-  version: 0.30.1
+  version: 0.31.0
 platform: ruby
 authors:
 - Eric Franz
@@ -10,7 +10,7 @@ authors:
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2025-12-03 00:00:00.000000000 Z
+date: 2026-04-16 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ood_support
@@ -88,14 +88,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 13.3.0
+        version: 13.4.1
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 13.3.0
+        version: 13.4.1
 - !ruby/object:Gem::Dependency
   name: rspec
   requirement: !ruby/object:Gem::Requirement
@@ -172,14 +172,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '2.4'
+        version: '3.0'
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '2.4'
+        version: '3.0'
 description: Open OnDemand core library that provides support for an HPC Center to
   globally define HPC services that web applications can then take advantage of.
 email:
@@ -214,6 +214,7 @@ files:
 - lib/ood_core/clusters.rb
 - lib/ood_core/data_formatter.rb
 - lib/ood_core/errors.rb
+- lib/ood_core/helpers/openstack.rb
 - lib/ood_core/invalid_cluster.rb
 - lib/ood_core/job/account_info.rb
 - lib/ood_core/job/adapter.rb