Class: Arcp::Runtime::JobManager

Inherits:
Object
  • Object
show all
Defined in:
lib/arcp/runtime/job_manager.rb

Overview

Owns agent registry + per-job lifecycle. Submitted jobs run as child ‘Async::Task`s; cancellation propagates via `task.stop`.

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(runtime:, lease_manager:, subscription_manager:, event_log:, clock: Arcp::SystemClock.new) ⇒ JobManager

Returns a new instance of JobManager.



21
22
23
24
25
26
27
28
29
30
31
32
 
# File 'lib/arcp/runtime/job_manager.rb', line 21

def initialize(runtime:, lease_manager:, subscription_manager:, event_log:, clock: Arcp::SystemClock.new)
  @runtime = runtime
  @leases = lease_manager
  @subs = subscription_manager
  @event_log = event_log
  @clock = clock
  @agents = {}                    # name => AgentRegistration
  @jobs = {}                      # job_id => JobRecord
  @event_seq = Hash.new(0)        # job_id => last emitted seq
  @idempotency = {}               # [principal, key] => job_id
  @mutex = Mutex.new
end

Instance Attribute Details

#runtime ⇒ Object (readonly)

Returns the value of attribute runtime.



19
20
21
 
# File 'lib/arcp/runtime/job_manager.rb', line 19

def runtime
  @runtime
end

Instance Method Details

#agent_inventory ⇒ Object 



43
44
45
46
47
48
49
50
51
 
# File 'lib/arcp/runtime/job_manager.rb', line 43

def agent_inventory
  @mutex.synchronize do
    Arcp::Session::AgentInventory.new(
      entries: @agents.values.map do |reg|
        Arcp::Session::AgentEntry.new(name: reg.name, versions: reg.versions, default: reg.default)
      end.freeze
    )
  end
end

#cancel(job_id:, principal_id:, reason: nil) ⇒ Object

Raises:



114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
 
# File 'lib/arcp/runtime/job_manager.rb', line 114

def cancel(job_id:, principal_id:, reason: nil)
  record = @mutex.synchronize { @jobs[job_id] }
  raise Arcp::Errors::JobNotFound, "no such job: #{job_id}" unless record

  unless record.principal_id == principal_id
    raise Arcp::Errors::PermissionDenied.new(
      'only the submitting principal can cancel a job',
      details: { 'job_id' => job_id }
    )
  end

  record.task&.stop
  publish_error(job_id, Arcp::Job::JobError.new(
                          job_id: job_id, final_status: 'cancelled',
                          code: 'CANCELLED', message: reason, retryable: false, details: {}
                        ))
end

#list(principal_id:, filter: {}, limit: 50, cursor: nil) ⇒ Object 



132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
 
# File 'lib/arcp/runtime/job_manager.rb', line 132

def list(principal_id:, filter: {}, limit: 50, cursor: nil)
  offset = cursor ? cursor.to_i : 0
  rows = @mutex.synchronize do
    @jobs.values
         .select { |r| r.principal_id == principal_id }
         .select { |r| filter['status'].nil? || filter['status'].include?(r.status) }
         .select { |r| filter['agent'].nil? || r.agent.start_with?(filter['agent']) }
         .sort_by(&:created_at)
  end

  page = rows[offset, limit] || []
  next_cursor = (offset + page.size) < rows.size ? (offset + page.size).to_s : nil

  summaries = page.map do |r|
    lease = @leases.get(r.job_id)
    counter = @leases.counter(r.job_id)
    Arcp::Job::Summary.new(
      job_id: r.job_id, agent: r.agent, status: r.status, created_at: r.created_at,
      lease_expires_at: lease&.expires_at,
      budget_remaining: counter ? counter.snapshot.transform_values { |v| v.to_s('F') } : nil
    )
  end

  Arcp::Session::JobsResponse.new(
    jobs: summaries.map(&:to_h), next_cursor: next_cursor
  )
end

#lookup(job_id) ⇒ Object 



160
 
# File 'lib/arcp/runtime/job_manager.rb', line 160

def lookup(job_id) = @mutex.synchronize { @jobs[job_id] }

#publish_error(job_id, error) ⇒ Object 



191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
 
# File 'lib/arcp/runtime/job_manager.rb', line 191

def publish_error(job_id, error)
  record = @mutex.synchronize do
    @jobs[job_id] = @jobs[job_id].with(status: error.final_status) if @jobs[job_id]
    @jobs[job_id]
  end
  env = Arcp::Envelope.build(
    type: Arcp::MessageTypes::JOB_ERROR,
    session_id: record&.submitter_session_id || '',
    job_id: job_id, payload: error.to_h
  )
  @event_log.append(env.session_id, env)
  @subs.fanout(job_id, env)
  @subs.clear(job_id)
  @runtime.credential_registry&.revoke_all(job_id: job_id)
  @leases.revoke(job_id)
end

#publish_event(job_id, event) ⇒ Object 



162
163
164
165
166
167
168
169
170
171
172
 
# File 'lib/arcp/runtime/job_manager.rb', line 162

def publish_event(job_id, event)
  seq = @mutex.synchronize { @event_seq[job_id] += 1 }
  env = Arcp::Envelope.build(
    type: Arcp::MessageTypes::JOB_EVENT,
    session_id: @mutex.synchronize { @jobs[job_id]&.submitter_session_id || '' },
    job_id: job_id, event_seq: seq, payload: event.to_h
  )
  @event_log.append(env.session_id, env)
  @subs.fanout(job_id, env)
  seq
end

#publish_result(job_id, result) ⇒ Object 



174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
 
# File 'lib/arcp/runtime/job_manager.rb', line 174

def publish_result(job_id, result)
  record = @mutex.synchronize do
    @jobs[job_id] = @jobs[job_id].with(status: 'succeeded') if @jobs[job_id]
    @jobs[job_id]
  end
  env = Arcp::Envelope.build(
    type: Arcp::MessageTypes::JOB_RESULT,
    session_id: record&.submitter_session_id || '',
    job_id: job_id, payload: result.to_h
  )
  @event_log.append(env.session_id, env)
  @subs.fanout(job_id, env)
  @subs.clear(job_id)
  @runtime.credential_registry&.revoke_all(job_id: job_id)
  @leases.revoke(job_id)
end

#register_agent(name:, versions:, default:, handler:) ⇒ Object 



34
35
36
37
38
39
40
41
 
# File 'lib/arcp/runtime/job_manager.rb', line 34

def register_agent(name:, versions:, default:, handler:)
  @mutex.synchronize do
    @agents[name] = AgentRegistration.new(
      name: name, versions: Array(versions).freeze,
      default: default, handler: handler
    )
  end
end

#resolve_agent(ref_str) ⇒ Object

Raises:



53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
 
# File 'lib/arcp/runtime/job_manager.rb', line 53

def resolve_agent(ref_str)
  ref = Arcp::Job::AgentRef.parse(ref_str)
  reg = @mutex.synchronize { @agents[ref.name] }
  raise Arcp::Errors::AgentNotAvailable, "agent not registered: #{ref.name}" if reg.nil?

  version = ref.version || reg.default
  if version.nil? || (!reg.versions.empty? && !reg.versions.include?(version))
    raise Arcp::Errors::AgentVersionNotAvailable.new(
      "agent #{ref.name} has no version #{version.inspect}",
      details: { 'agent' => ref.name, 'version' => version, 'available' => reg.versions }
    )
  end

  [reg, "#{ref.name}@#{version}"]
end

#submit(submit:, principal_id:, session_id:, session_actor:) ⇒ Object 



69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
 
# File 'lib/arcp/runtime/job_manager.rb', line 69

def submit(submit:, principal_id:, session_id:, session_actor:)
  reg, resolved = resolve_agent(submit.agent)

  if submit.idempotency_key
    key = [principal_id, submit.idempotency_key]
    if (existing = @mutex.synchronize { @idempotency[key] })
      existing_record = @mutex.synchronize { @jobs[existing] }
      if existing_record && existing_record.agent != resolved
        raise Arcp::Errors::DuplicateKey.new(
          'idempotency key reused with different agent',
          details: { 'job_id' => existing }
        )
      end
      return existing
    end
  end

  job_id = Arcp::Ids.job_id

  lease = build_lease(submit, job_id)
  @leases.register(job_id, lease) if lease
  credentials = issue_credentials(
    job_id: job_id, lease: lease, agent: resolved, principal_id: principal_id
  )

  record = JobRecord.new(
    job_id: job_id, agent: resolved, principal_id: principal_id,
    status: 'pending', created_at: @clock.now.iso8601,
    input: submit.input, submitter_session_id: session_id, task: nil
  )
  @mutex.synchronize do
    @jobs[job_id] = record
    @idempotency[[principal_id, submit.idempotency_key]] = job_id if submit.idempotency_key
  end

  @subs.register_owner(job_id, principal_id, session_id, session_actor.outbox)

  task = Async do |t|
    run_agent(t, reg, job_id, submit, lease)
  end
  @mutex.synchronize { @jobs[job_id] = @jobs[job_id].with(task: task, status: 'running') }

  [job_id, resolved, lease, credentials]
end