ace-assign 0.69.0 → 0.70.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (38) hide show
  1. checksums.yaml +4 -4
  2. data/.ace-defaults/assign/presets/deliver-release.yml +11 -8
  3. data/.ace-defaults/assign/presets/integrate-batch.yml +10 -5
  4. data/.ace-defaults/assign/presets/supervise-delivery.yml +6 -4
  5. data/CHANGELOG.md +7 -0
  6. data/lib/ace/assign/atoms/assignment_scope.rb +5 -0
  7. data/lib/ace/assign/authority/assignment_definition.rb +35 -0
  8. data/lib/ace/assign/authority/candidate_transfer.rb +2 -2
  9. data/lib/ace/assign/authority/endcap_attempts.rb +1 -1
  10. data/lib/ace/assign/authority/launch_control_exclusions.rb +3 -17
  11. data/lib/ace/assign/authority/ordinary_attempt_start.rb +137 -0
  12. data/lib/ace/assign/authority/ordinary_client.rb +70 -0
  13. data/lib/ace/assign/authority/ordinary_completion.rb +94 -0
  14. data/lib/ace/assign/authority/ordinary_effect_settlement.rb +118 -0
  15. data/lib/ace/assign/authority/ordinary_lifecycle.rb +191 -0
  16. data/lib/ace/assign/authority/ordinary_peer.rb +54 -0
  17. data/lib/ace/assign/authority/ordinary_progress.rb +162 -0
  18. data/lib/ace/assign/authority/ordinary_queue.rb +115 -0
  19. data/lib/ace/assign/authority/ordinary_results.rb +196 -0
  20. data/lib/ace/assign/authority/ordinary_server.rb +75 -0
  21. data/lib/ace/assign/authority/prepared_work.rb +2 -2
  22. data/lib/ace/assign/authority/protected_assignment_context.rb +3 -3
  23. data/lib/ace/assign/authority/router.rb +11 -6
  24. data/lib/ace/assign/authority/server.rb +11 -180
  25. data/lib/ace/assign/authority/socket_server.rb +255 -0
  26. data/lib/ace/assign/authority/transfer_codec.rb +2 -2
  27. data/lib/ace/assign/cli/commands/attempt/evidence.rb +2 -2
  28. data/lib/ace/assign/molecules/canonical_read_commands.rb +84 -0
  29. data/lib/ace/assign/molecules/canonical_read_snapshot.rb +2 -63
  30. data/lib/ace/assign/molecules/evidence_journal.rb +141 -34
  31. data/lib/ace/assign/molecules/installed_journal_read.rb +143 -0
  32. data/lib/ace/assign/molecules/journal_mutation.rb +5 -3
  33. data/lib/ace/assign/molecules/lifecycle_exclusion.rb +13 -2
  34. data/lib/ace/assign/molecules/proposal_journal.rb +32 -4
  35. data/lib/ace/assign/organisms/attempt_coordinator.rb +51 -36
  36. data/lib/ace/assign/organisms/ordinary_merge_consumer.rb +5 -5
  37. data/lib/ace/assign/version.rb +1 -1
  38. metadata +23 -9
@@ -0,0 +1,118 @@
1
+ # frozen_string_literal: true
2
+
3
+ require_relative "service_evidence"
4
+ require_relative "../molecules/ordinary_merge_contract"
5
+
6
+ module Ace
7
+ module Assign
8
+ module Authority
9
+ # Completion reads the existing effect owners at the held CAS prefix.
10
+ # A prompt, acknowledgment or failed dispatch never settles an effect.
11
+ class OrdinaryEffectSettlement
12
+ def initialize(journal:)
13
+ @journal = journal
14
+ end
15
+
16
+ def verify!(project:, assignment:, attempt:, events:, commit:)
17
+ retained = @journal.read_events(assignment, commit: commit).select { |event| event["attempt_id"] == attempt }
18
+ unless retained == events && Models::EvidenceEvent.chain_valid?(events)
19
+ unavailable!("ordinary settlement needs its exact canonical attempt")
20
+ end
21
+ selection = {"project_id" => project, "assignment_id" => assignment, "attempt_id" => attempt}
22
+ services = records!(@journal.service_request_records(commit: commit), events, selection, %w[service_claim service_transition])
23
+ merges = records!(@journal.fixed_merge_request_records(commit: commit), events, selection, %w[merge_admission merge_transition])
24
+ services.each { |record| service!(record, commit) }
25
+ merges.each { |record| merge!(record, commit) }
26
+ assignment_events = @journal.read_events(assignment, commit: commit)
27
+ assignment_events.group_by { |event| event.fetch("attempt_id") }.each_value do |chain|
28
+ unavailable!("ordinary delivery history is corrupt") unless Models::EvidenceEvent.chain_valid?(chain)
29
+ end
30
+ delivery!(events, merges, assignment_events)
31
+ {"journal_commit" => commit, "service_requests" => services.map { |record| record.fetch("request_id") }.sort,
32
+ "merge_requests" => merges.map { |record| record.fetch("request_id") }.sort}
33
+ end
34
+
35
+ private
36
+
37
+ def records!(inventory, events, selection, types)
38
+ records = inventory.select { |record| record.values_at("assignment_id", "attempt_id") == selection.values_at("assignment_id", "attempt_id") }
39
+ references = events.select { |event| types.include?(event["type"]) }
40
+ ids = references.map { |event| event.fetch("payload").fetch("request_id") }.uniq.sort
41
+ unless records.map { |record| record.fetch("request_id") }.sort == ids
42
+ unavailable!("ordinary effect inventory and event set differ")
43
+ end
44
+ records.each do |record|
45
+ latest = references.reverse.find { |event| event.dig("payload", "request_id") == record.fetch("request_id") }
46
+ unless record.slice(*selection.keys) == selection && latest.dig("payload", "state") == record["state"] &&
47
+ latest.dig("payload", "record_digest") == Atoms::EvidenceDigest.digest(record) &&
48
+ latest.dig("payload", "receipt_digest") == (record["receipt"] && Atoms::EvidenceDigest.digest(record["receipt"]))
49
+ unavailable!("ordinary effect record differs from its accepted owner event")
50
+ end
51
+ end
52
+ records
53
+ end
54
+
55
+ def service!(record, commit)
56
+ return if record["state"] == "rejected" && record["consumed"] == false
57
+ pending!("ordinary service effect is unresolved") unless %w[succeeded failed-settled].include?(record["state"])
58
+ receipt = record.fetch("receipt")
59
+ @journal.validate_terminal_receipt_structure!(record, record.fetch("state"), receipt)
60
+ # The installed owner consumes the existing canonical import, never
61
+ # the old local receipt file. This is independent of journal mode.
62
+ owner = ServiceEvidence.new(journal: @journal)
63
+ references = receipt.fetch("evidence")
64
+ if %w[publish prune-preserved-workspace].include?(record["operation"])
65
+ contents = owner.call(references, record, record.fetch("state"), {commit: commit})
66
+ unless contents.is_a?(Array) && contents.size == references.size
67
+ unavailable!("ordinary service evidence collection differs")
68
+ end
69
+ else
70
+ contents = references.map { |reference| owner.call(reference, record, record.fetch("state"), {commit: commit}) }
71
+ end
72
+ references.zip(contents).each do |reference, bytes|
73
+ unavailable!("ordinary service evidence digest differs") unless bytes.is_a?(String) && Digest::SHA256.hexdigest(bytes.b) == reference.fetch("sha256")
74
+ @journal.verify_service_attestation!(bytes, record, record.fetch("state")) unless %w[publish prune-preserved-workspace].include?(record["operation"])
75
+ end
76
+ end
77
+
78
+ def merge!(record, commit)
79
+ unless @journal.fixed_merge_settlement!(record.fetch("request_id"), commit: commit) == record
80
+ unavailable!("ordinary merge settlement differs from its selected record")
81
+ end
82
+ end
83
+
84
+ def delivery!(events, merges, assignment_events)
85
+ deliveries = assignment_events.select { |event| event["type"] == "delivery" }
86
+ events.select { |event| event["type"] == "delivery" }.each do |intent|
87
+ payload = intent.fetch("payload")
88
+ case payload["stage"]
89
+ when "intent"
90
+ settled = deliveries.any? { |event| event.dig("payload", "stage") == "result" &&
91
+ event.dig("payload", "intent_digest") == intent.fetch("digest") &&
92
+ event.dig("payload", "intent_attempt_id") == intent.fetch("attempt_id") &&
93
+ event.dig("payload", "operation") == payload.fetch("operation") &&
94
+ event.dig("payload", "head") == payload.fetch("head") && event.dig("payload", "outcome") == "succeeded" }
95
+ pending!("ordinary delivery intent is unresolved") unless settled
96
+ when "ordinary-merge-intent"
97
+ result = deliveries.reverse.find { |event| event.dig("payload", "stage") == "ordinary-merge-result" &&
98
+ event.dig("payload", "request_id") == payload.fetch("request_id") && event.dig("payload", "input_digest") == payload.fetch("input_digest") }
99
+ pending!("ordinary merge delivery is unresolved") unless result
100
+ accepted = result.fetch("payload")
101
+ record = merges.find { |item| item["request_id"] == payload.fetch("request_id") }
102
+ if accepted["outcome"] == "refused_no_effect" && record.nil?
103
+ refusal = accepted.slice("code", "message", "request_id", "input_digest").merge("schema" => "ace.lab.service.ordinary-merge-refusal/v1")
104
+ Molecules::OrdinaryMergeContract.validate_refusal!(refusal)
105
+ pending!("ordinary merge receiver availability is unresolved") if refusal["code"] == Molecules::OrdinaryMergeContract::UNAVAILABLE_CODE
106
+ elsif !record || record["input_digest"] != payload.fetch("input_digest") || record["receipt"] != accepted["receipt"]
107
+ unavailable!("ordinary delivery result differs from the merge owner")
108
+ end
109
+ end
110
+ end
111
+ end
112
+
113
+ def unavailable!(message) = raise AttemptErrors::EvidenceUnavailable, message
114
+ def pending!(message) = raise AttemptErrors::ServiceSettlementPending, message
115
+ end
116
+ end
117
+ end
118
+ end
@@ -0,0 +1,191 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "ace/herdr"
4
+ require_relative "prepared_work"
5
+ require_relative "assignment_definition"
6
+ require_relative "ordinary_queue"
7
+ require_relative "ordinary_attempt_start"
8
+ require_relative "ordinary_progress"
9
+ require_relative "ordinary_results"
10
+ require_relative "ordinary_completion"
11
+ require_relative "../molecules/lifecycle_exclusion"
12
+
13
+ module Ace
14
+ module Assign
15
+ module Authority
16
+ # Ordinary lifecycle uses the existing canonical execution ref and CAS.
17
+ # The installed Lab composition supplies one journal and exclusion owner
18
+ # shared with HITL and merge. No requester cache supplies accepted state.
19
+ class OrdinaryLifecycle
20
+ include OrdinaryAttemptStart
21
+ include OrdinaryProgress
22
+ include OrdinaryResults
23
+ include OrdinaryCompletion
24
+ OPERATIONS = (%w[register_assignment registration_status start_attempt finish] + OrdinaryProgress::PROGRESS_OPERATIONS + OrdinaryResults::RESULT_OPERATIONS).freeze
25
+ TRANSFER_OPERATIONS = {"register_assignment" => {direction: :upload, purpose: :candidate, roles: [:supervisor]}}
26
+ .merge(OrdinaryProgress::PROGRESS_TRANSFERS).merge(OrdinaryResults::RESULT_TRANSFERS).freeze
27
+ REGISTER_KEYS = %w[assignment_id definition_digest prepared_head prepared_tree manifest_sha256 expected_generation pane transfer].freeze
28
+
29
+ def initialize(project:, admission:, journal:, exclusion:, input_root:, inbox: nil)
30
+ @project, @admission, @journal, @exclusion = project, admission, journal, exclusion
31
+ @input_root = PrivateDirectory.verify!(input_root)
32
+ unless inbox.nil? || inbox.is_a?(Ace::Herdr::Organisms::Inbox)
33
+ raise ArgumentError, "ordinary completion requires the installed Inbox owner"
34
+ end
35
+ @inbox = inbox
36
+ end
37
+
38
+ def authorize_transfer!(request:, peer:, role:, deadline:)
39
+ authorize!(request, peer, role, deadline)
40
+ if OrdinaryResults::RESULT_TRANSFERS.key?(request.fetch("operation"))
41
+ validate_result!(request)
42
+ authorize_start_replay!(request, peer, role, deadline) if request.fetch("operation") == "submit_result"
43
+ return
44
+ end
45
+ if OrdinaryProgress::PROGRESS_TRANSFERS.key?(request.fetch("operation"))
46
+ validate_progress!(request)
47
+ return
48
+ end
49
+ unless request.fetch("operation") == "register_assignment" && role == :supervisor
50
+ raise AttemptErrors::UnauthorizedIdentity, "ordinary registration requires the installed supervisor"
51
+ end
52
+ validate_registration!(request)
53
+ end
54
+
55
+ def dispatch(request:, peer:, role:, deadline:, transfer: nil)
56
+ authorize!(request, peer, role, deadline)
57
+ case request.fetch("operation")
58
+ when "registration_status"
59
+ strict!(request.fetch("params"), %w[assignment_id])
60
+ raise ArgumentError, "status mutation must be null" unless request.fetch("mutation_id").nil?
61
+ assignment = request.fetch("params").fetch("assignment_id")
62
+ token!(assignment)
63
+ commit = @journal.ref_value
64
+ data = registration(assignment, commit: commit) || {"assignment_id" => assignment, "generation" => 0}
65
+ {data: data.merge("journal_commit" => commit), replayed: false}
66
+ when "register_assignment"
67
+ authorize_transfer!(request: request, peer: peer, role: role, deadline: deadline)
68
+ register(request, peer, deadline, transfer)
69
+ when "start_attempt"
70
+ raise ArgumentError, "ordinary start forbids transfer" if transfer
71
+ start_attempt(request, peer, role, deadline)
72
+ when "assignment_snapshot"
73
+ raise ArgumentError, "ordinary snapshot forbids upload" if transfer
74
+ assignment_snapshot(request, peer, role, deadline)
75
+ when "queue_start", "queue_finish", "queue_fail"
76
+ progress(request, peer, role, deadline, transfer)
77
+ when "submit_result", "result_status", "result_snapshot"
78
+ ordinary_result(request, peer, role, deadline, transfer)
79
+ when "finish"
80
+ raise ArgumentError, "ordinary completion forbids transfer" if transfer
81
+ finish_result(request, peer, role, deadline)
82
+ else
83
+ raise ArgumentError, "unknown ordinary lifecycle operation"
84
+ end
85
+ end
86
+
87
+ private
88
+
89
+ def register(request, peer, deadline, transfer)
90
+ params = request.fetch("params")
91
+ # Native observation and candidate subprocesses happen before CAS.
92
+ identity = @admission.identity!(peer: peer, pane: params.fetch("pane"), deadline: deadline)
93
+ unless identity.role == "coordinator"
94
+ raise AttemptErrors::UnauthorizedIdentity, "ordinary registration native owner differs"
95
+ end
96
+ raise AttemptErrors::MalformedTransfer, "prepared input requires one bundle" unless transfer && transfer.count == 1
97
+ bytes = transfer.bytes.freeze
98
+ descriptor = params.fetch("transfer")
99
+ work = PreparedWork.admit(bytes: bytes, head: params.fetch("prepared_head"), tree: params.fetch("prepared_tree"),
100
+ sha256: descriptor.fetch("sha256"), size: descriptor.fetch("bytes"), root: @input_root, deadline: deadline)
101
+ unless work.manifest.values_at("assignment_id", "project_id") == [params.fetch("assignment_id"), @project] &&
102
+ work.manifest_sha256 == params.fetch("manifest_sha256")
103
+ raise ArgumentError, "ordinary prepared assignment association differs"
104
+ end
105
+ definition = work.definition_bytes(head: params.fetch("prepared_head"), tree: params.fetch("prepared_tree"))
106
+ assignment = AssignmentDefinition.parse(bytes: definition, digest: params.fetch("definition_digest"),
107
+ assignment_id: params.fetch("assignment_id"), project_id: @project)
108
+ # Campaign registration needs its existing canonical campaign owner;
109
+ # it cannot be manufactured by attaching unchecked parent metadata.
110
+ if assignment.review_campaign || assignment.campaign_execution
111
+ raise AttemptErrors::EvidenceUnavailable, "ordinary campaign registration requires its campaign owner"
112
+ end
113
+ principal = {"uid" => peer.fetch("uid"), "actor" => identity.actor}
114
+ queue = OrdinaryQueue.new(work: work, root: @input_root)
115
+ queue_sha = Digest::SHA256.hexdigest(queue.bytes)
116
+ digest = Atoms::EvidenceDigest.digest({"params" => params, "project_id" => @project, "principal" => principal})
117
+ keys = [@exclusion.task_key(assignment.task_id), @exclusion.assignment_key(assignment.id)]
118
+ @exclusion.with_shared_multi(keys, deadline: deadline) do
119
+ # Replay bypasses the CAS callback; current authority is mandatory
120
+ # before that lookup as well as on every fresh CAS attempt.
121
+ authorize!(request, peer, :supervisor, deadline)
122
+ @journal.mutate(assignment_id: assignment.id, attempt_id: "definition-#{assignment.id}",
123
+ mutation_id: request.fetch("mutation_id"), operation: "register_assignment", parameters_digest: digest,
124
+ expected_generation: params.fetch("expected_generation"), with_replay: true, deadline: deadline,
125
+ replay_authorizer: -> { authorize!(request, peer, :supervisor, deadline) }) do |events, commit, _generation|
126
+ authorize!(request, peer, :supervisor, deadline)
127
+ unless events.empty? && registration(assignment.id, commit: commit).nil?
128
+ raise AttemptErrors::Conflict, "ordinary assignment is already registered; retain its original definition"
129
+ end
130
+ bundle_sha = Digest::SHA256.hexdigest(bytes)
131
+ definition_path = "execution/definitions/#{assignment.id}-#{params.fetch('definition_digest')}.json"
132
+ bundle_path = "execution/prepared/#{assignment.id}-#{bundle_sha}.bundle"
133
+ queue_path = "execution/definitions/#{assignment.id}-queue-#{queue_sha}.json"
134
+ {events: [], blobs: {definition_path => definition, bundle_path => bytes, queue_path => queue.bytes}, data: {
135
+ "authority" => "ordinary", "assignment_id" => assignment.id, "project_id" => @project,
136
+ "task_id" => assignment.task_id, "registered_by" => principal, "phase" => "registered",
137
+ "definition_ref" => definition_path, "definition_digest" => params.fetch("definition_digest"),
138
+ "queue_ref" => queue_path, "queue_sha256" => queue_sha,
139
+ "prepared_work" => work.reference(head: params.fetch("prepared_head"), tree: params.fetch("prepared_tree")),
140
+ "prepared_bundle_ref" => bundle_path, "prepared_bundle_sha256" => bundle_sha,
141
+ "prepared_bundle_bytes" => bytes.bytesize}}
142
+ end
143
+ end
144
+ end
145
+
146
+ def registration(assignment, commit:)
147
+ return nil unless commit
148
+ events = @journal.read_events(assignment, commit: commit)
149
+ events.group_by { |event| event.fetch("attempt_id") }.each_value do |chain|
150
+ raise AttemptErrors::EvidenceUnavailable, "ordinary assignment evidence is corrupt" unless Models::EvidenceEvent.chain_valid?(chain)
151
+ end
152
+ registrations = events.select { |event| event["type"] == "authority_mutation" && event.dig("payload", "operation") == "register_assignment" }
153
+ return nil if registrations.empty?
154
+ data = registrations.one? && registrations.first.dig("payload", "data")
155
+ unless data && data.values_at("authority", "assignment_id", "project_id") == ["ordinary", assignment, @project]
156
+ raise AttemptErrors::EvidenceUnavailable, "ordinary assignment registration owner differs"
157
+ end
158
+ data
159
+ end
160
+
161
+ def authorize!(request, peer, role, deadline)
162
+ unless deadline.is_a?(Numeric) && deadline.finite? && deadline > Process.clock_gettime(Process::CLOCK_MONOTONIC)
163
+ raise AttemptErrors::EvidenceUnavailable, "ordinary lifecycle deadline elapsed"
164
+ end
165
+ unless request.fetch("project_id") == @project && @admission.authorize!(request: request, peer: peer, deadline: deadline) == role
166
+ raise AttemptErrors::UnauthorizedIdentity, "ordinary lifecycle principal differs"
167
+ end
168
+ end
169
+
170
+ def validate_registration!(request)
171
+ params = request.fetch("params")
172
+ strict!(params, REGISTER_KEYS)
173
+ token!(request.fetch("mutation_id"))
174
+ token!(params.fetch("assignment_id"))
175
+ token!("definition-#{params.fetch('assignment_id')}")
176
+ unless params.fetch("expected_generation").is_a?(Integer) && params.fetch("expected_generation") >= 0
177
+ raise ArgumentError, "ordinary generation is invalid"
178
+ end
179
+ end
180
+
181
+ def strict!(value, keys)
182
+ raise ArgumentError, "ordinary authority fields differ" unless value.is_a?(Hash) && value.keys.sort == keys.sort
183
+ end
184
+
185
+ def token!(value)
186
+ raise ArgumentError, "ordinary identity is invalid" unless value.is_a?(String) && Molecules::JournalMutation::ID.match?(value)
187
+ end
188
+ end
189
+ end
190
+ end
191
+ end
@@ -0,0 +1,54 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "socket"
4
+ require "ace/runtime/molecules/process_identity"
5
+ require "ace/herdr/molecules/bounded_observation"
6
+
7
+ module Ace
8
+ module Assign
9
+ module Authority
10
+ # Ordinary UNIX peer attribution. This is not a protected launch observer:
11
+ # it confers no Yama, capability, cgroup or descendant-closure guarantee.
12
+ class OrdinaryPeer
13
+ def initialize(identity: nil)
14
+ @identity = identity
15
+ end
16
+
17
+ def capture(socket, deadline: default_deadline)
18
+ identity = identity_observer(deadline)
19
+ uid, gid = socket.getpeereid
20
+ pid = if RUBY_PLATFORM.include?("darwin")
21
+ socket.getsockopt(0, 2).int # SOL_LOCAL / LOCAL_PEERPID
22
+ elsif Socket.const_defined?(:SO_PEERCRED)
23
+ observed_pid, observed_uid, observed_gid = socket.getsockopt(Socket::SOL_SOCKET, Socket::SO_PEERCRED).data.unpack("i!3")
24
+ unavailable! unless [observed_uid, observed_gid] == [uid, gid]
25
+ observed_pid
26
+ end
27
+ value = identity.capture(pid)
28
+ unavailable! unless value && value["uid"] == uid && socket.getpeereid == [uid, gid] && identity.capture(pid) == value
29
+ value.merge("gid" => gid).freeze
30
+ rescue SystemCallError, IOError, SocketError, NotImplementedError
31
+ unavailable!
32
+ end
33
+
34
+ def live!(peer, deadline: default_deadline)
35
+ unavailable! unless peer.is_a?(Hash) && identity_observer(deadline).capture(peer["pid"]) == peer.slice("pid", "uid", "started_at", "host")
36
+ peer
37
+ end
38
+
39
+ private
40
+
41
+ def default_deadline = Process.clock_gettime(Process::CLOCK_MONOTONIC) + 5
42
+
43
+ def identity_observer(deadline)
44
+ @identity || Ace::Runtime::Molecules::ProcessIdentity.new(
45
+ executor: Ace::Herdr::Molecules::BoundedObservation.new(deadline: deadline))
46
+ end
47
+
48
+ def unavailable!
49
+ raise AttemptErrors::UnauthorizedIdentity, "ordinary kernel peer is unavailable or changed"
50
+ end
51
+ end
52
+ end
53
+ end
54
+ end
@@ -0,0 +1,162 @@
1
+ # frozen_string_literal: true
2
+
3
+ module Ace
4
+ module Assign
5
+ module Authority
6
+ # Source-owned progress for the already accepted work. Queue completion
7
+ # remains a worker report, never a successful execution result or merge.
8
+ module OrdinaryProgress
9
+ PROGRESS_OPERATIONS = %w[queue_start queue_finish queue_fail assignment_snapshot].freeze
10
+ PROGRESS_TRANSFERS = {
11
+ "queue_finish" => {direction: :upload, purpose: :artifacts, roles: %i[worker supervisor]},
12
+ "assignment_snapshot" => {direction: :download, purpose: :assignment_snapshot, roles: %i[worker supervisor reviewer]}
13
+ }.freeze
14
+ PROGRESS_COMMON = %w[assignment_id attempt_id expected_generation pane].freeze
15
+
16
+ private
17
+
18
+ def validate_progress!(request)
19
+ params = request.fetch("params")
20
+ operation = request.fetch("operation")
21
+ if operation == "assignment_snapshot"
22
+ strict!(params, %w[assignment_id attempt_id kind])
23
+ token!(params.fetch("attempt_id")) unless params.fetch("attempt_id").nil?
24
+ unless request.fetch("mutation_id").nil? && %w[prepared_work queue].include?(params.fetch("kind"))
25
+ raise ArgumentError, "ordinary snapshot must be a read of accepted work or queue"
26
+ end
27
+ else
28
+ fields = case operation
29
+ when "queue_start" then %w[step]
30
+ when "queue_finish" then %w[step transfer]
31
+ when "queue_fail" then %w[message]
32
+ else raise ArgumentError, "unknown ordinary progress operation"
33
+ end
34
+ strict!(params, PROGRESS_COMMON + fields)
35
+ token!(params.fetch("attempt_id")); token!(request.fetch("mutation_id"))
36
+ unless params.fetch("expected_generation").is_a?(Integer) && params.fetch("expected_generation").positive?
37
+ raise ArgumentError, "ordinary progress generation is invalid"
38
+ end
39
+ if params.key?("step") && !params["step"].nil? &&
40
+ !(params["step"].is_a?(String) && params["step"].bytesize.between?(1, 128) && PreparedWork::SCOPE.match?(params["step"]))
41
+ raise ArgumentError, "ordinary progress step is invalid"
42
+ end
43
+ if params.key?("message") && !(params["message"].is_a?(String) && params["message"].bytesize.between?(1, 1024) && !params["message"].include?("\0"))
44
+ raise ArgumentError, "ordinary failure message is invalid"
45
+ end
46
+ end
47
+ token!(params.fetch("assignment_id"))
48
+ end
49
+
50
+ def assignment_snapshot(request, peer, role, deadline)
51
+ validate_progress!(request)
52
+ params = request.fetch("params")
53
+ commit = @journal.ref_value
54
+ registered = registration(params.fetch("assignment_id"), commit: commit)
55
+ raise AttemptErrors::NotFound, "ordinary assignment is not registered" unless registered
56
+ queue_record = current_queue_record(registered, commit)
57
+ selected_events = if params.fetch("attempt_id")
58
+ result_events(params, commit)
59
+ else
60
+ []
61
+ end
62
+ kind = params.fetch("kind")
63
+ source = kind == "prepared_work" ? registered : queue_record
64
+ path_key, digest_key, size = if kind == "prepared_work"
65
+ ["prepared_bundle_ref", "prepared_bundle_sha256", registered.fetch("prepared_bundle_bytes")]
66
+ else
67
+ ["queue_ref", "queue_sha256", PreparedWork::MAX_TOTAL]
68
+ end
69
+ bytes = @journal.bounded_blob(source.fetch(path_key), commit: commit, max_bytes: size)
70
+ unless Digest::SHA256.hexdigest(bytes) == source.fetch(digest_key)
71
+ raise AttemptErrors::EvidenceUnavailable, "ordinary snapshot bytes differ from accepted selection"
72
+ end
73
+ authorize!(request, peer, role, deadline)
74
+ {data: {"project_id" => @project, "assignment_id" => params.fetch("assignment_id"), "kind" => kind,
75
+ "attempt_id" => params.fetch("attempt_id"), "generation" => @journal.authority_generation(selected_events),
76
+ "scope" => selected_events.find { |event| event["type"] == "intent" }&.dig("payload", "scope"),
77
+ "journal_commit" => commit, "sha256" => source.fetch(digest_key), "bytes" => bytes.bytesize,
78
+ "prepared_work" => registered.fetch("prepared_work"), "definition_digest" => registered.fetch("definition_digest")},
79
+ transfer_parts: [bytes], replayed: false}
80
+ end
81
+
82
+ def progress(request, peer, role, deadline, transfer)
83
+ validate_progress!(request)
84
+ unless %i[supervisor worker].include?(role)
85
+ raise AttemptErrors::UnauthorizedIdentity, "ordinary progress requires its original worker"
86
+ end
87
+ params, operation = request.fetch("params"), request.fetch("operation")
88
+ report = if operation == "queue_finish"
89
+ raise ArgumentError, "ordinary progress report needs one bounded part" unless transfer && transfer.count == 1
90
+ bytes = transfer.bytes.dup.force_encoding(Encoding::UTF_8)
91
+ unless bytes.bytesize.between?(1, 65_536) && bytes.valid_encoding? && !bytes.include?("\0")
92
+ raise ArgumentError, "ordinary progress report must be bounded UTF-8"
93
+ end
94
+ bytes
95
+ elsif transfer
96
+ raise ArgumentError, "ordinary progress operation forbids transfer"
97
+ end
98
+ commit = @journal.ref_value
99
+ registered = registration(params.fetch("assignment_id"), commit: commit)
100
+ raise AttemptErrors::NotFound, "ordinary work is not registered" unless registered
101
+ work = registered_work(registered, commit, deadline)
102
+ digest = Atoms::EvidenceDigest.digest({"params" => params, "project_id" => @project, "peer_uid" => peer.fetch("uid")})
103
+ keys = [@exclusion.task_key(registered.fetch("task_id")), @exclusion.assignment_key(params.fetch("assignment_id"))]
104
+ @exclusion.with_shared_multi(keys, deadline: deadline) do
105
+ authorize!(request, peer, role, deadline)
106
+ @journal.mutate(assignment_id: params.fetch("assignment_id"), attempt_id: params.fetch("attempt_id"),
107
+ mutation_id: request.fetch("mutation_id"), operation: operation, parameters_digest: digest,
108
+ expected_generation: params.fetch("expected_generation"), with_replay: true, deadline: deadline,
109
+ replay_authorizer: -> { authorize_start_replay!(request, peer, role, deadline) }) do |events, prefix, _generation|
110
+ authorize_start_replay!(request, peer, role, deadline)
111
+ unless Molecules::CanonicalAttemptState.derive(events) == "running"
112
+ raise AttemptErrors::Conflict, "ordinary progress requires the original running attempt"
113
+ end
114
+ record = current_queue_record(registered, prefix)
115
+ scope = events.find { |event| event["type"] == "intent" }.fetch("payload").fetch("scope")
116
+ queue = OrdinaryQueue.new(work: work, root: @input_root, bytes: queue_bytes(record, prefix))
117
+ next_queue = case operation
118
+ when "queue_start" then queue.transition(operation: :start, scope: scope, step: params.fetch("step"))
119
+ when "queue_finish" then queue.transition(operation: :finish, scope: scope, step: params.fetch("step"), report: report)
120
+ when "queue_fail" then queue.transition(operation: :fail, scope: scope, message: params.fetch("message"))
121
+ end
122
+ sha = Digest::SHA256.hexdigest(next_queue.bytes)
123
+ path = "execution/definitions/#{params.fetch('assignment_id')}-queue-#{sha}.json"
124
+ {events: [], blobs: {path => next_queue.bytes}, data: {
125
+ "authority" => "ordinary", "assignment_id" => params.fetch("assignment_id"), "attempt_id" => params.fetch("attempt_id"),
126
+ "project_id" => @project, "scope" => scope, "queue_ref" => path, "queue_sha256" => sha, "state" => "running",
127
+ "queue_complete" => next_queue.status.subtree_complete?(scope)}}
128
+ end
129
+ end
130
+ end
131
+
132
+ def current_queue_record(registered, commit)
133
+ assignment = registered.fetch("assignment_id")
134
+ events = @journal.read_events(assignment, commit: commit)
135
+ events.group_by { |event| event.fetch("attempt_id") }.each_value do |chain|
136
+ raise AttemptErrors::EvidenceUnavailable, "ordinary queue history is corrupt" unless Models::EvidenceEvent.chain_valid?(chain)
137
+ end
138
+ updates = events.select do |entry|
139
+ entry["type"] == "authority_mutation" &&
140
+ %w[start_attempt queue_start queue_finish queue_fail].include?(entry.dig("payload", "operation"))
141
+ end
142
+ return registered if updates.empty?
143
+ event = @journal.latest_event!(assignment_id: assignment, event_digests: updates.map { |entry| entry.fetch("digest") }, commit: commit)
144
+ record = event.dig("payload", "data")
145
+ unless record && record.values_at("authority", "assignment_id", "attempt_id", "project_id") ==
146
+ ["ordinary", assignment, event.fetch("attempt_id"), @project]
147
+ raise AttemptErrors::EvidenceUnavailable, "ordinary canonical queue is unavailable"
148
+ end
149
+ record
150
+ end
151
+
152
+ def queue_bytes(record, commit)
153
+ bytes = @journal.bounded_blob(record.fetch("queue_ref"), commit: commit, max_bytes: PreparedWork::MAX_TOTAL)
154
+ unless Digest::SHA256.hexdigest(bytes) == record.fetch("queue_sha256")
155
+ raise AttemptErrors::EvidenceUnavailable, "ordinary canonical queue bytes differ"
156
+ end
157
+ bytes
158
+ end
159
+ end
160
+ end
161
+ end
162
+ end
@@ -0,0 +1,115 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "tmpdir"
4
+ require "fileutils"
5
+ require_relative "prepared_work"
6
+ require_relative "private_directory"
7
+ require_relative "../organisms/assignment_executor"
8
+
9
+ module Ace
10
+ module Assign
11
+ module Authority
12
+ # A disposable projection for the existing queue engine. Only bytes read
13
+ # from (or about to be committed to) the canonical execution ref enter it.
14
+ # No files left by an earlier coordinator are recovery input.
15
+ class OrdinaryQueue
16
+ attr_reader :bytes
17
+
18
+ def initialize(work:, root:, bytes: nil)
19
+ raise ArgumentError, "ordinary queue requires prepared work" unless work.is_a?(PreparedWork)
20
+ @work, @root = work, PrivateDirectory.verify!(root)
21
+ value = if bytes
22
+ invalid! unless bytes.is_a?(String) && bytes.bytesize.between?(1, PreparedWork::MAX_TOTAL)
23
+ JSON.parse(bytes)
24
+ else
25
+ {"version" => 1, "steps" => work.manifest.fetch("steps").to_h { |step|
26
+ name = step.fetch("filename"); [name, work.files.fetch("steps/#{name}")]
27
+ }, "reports" => {}}
28
+ end
29
+ validate!(value, initial: bytes.nil?)
30
+ @bytes = Atoms::EvidenceDigest.canonical_json(value).freeze
31
+ rescue JSON::ParserError
32
+ invalid!
33
+ end
34
+
35
+ def status
36
+ value = JSON.parse(@bytes)
37
+ assignment = Models::Assignment.from_h(@work.definition, cache_dir: File.join(@root, @work.manifest.fetch("assignment_id")))
38
+ Molecules::QueueScanner.new.scan_captured(steps: value.fetch("steps"), reports: value.fetch("reports"), assignment: assignment)
39
+ end
40
+
41
+ # This returns proposed bytes only. The lifecycle owns authorization,
42
+ # generation, replay and publication together with its attempt events.
43
+ def transition(operation:, scope: @work.manifest.fetch("scope"), step: nil, report: nil, message: nil)
44
+ invalid! unless @work.manifest.fetch("steps").any? { |entry| entry.fetch("number") == scope }
45
+ value = JSON.parse(@bytes)
46
+ Dir.mktmpdir("ordinary-queue-", @root) do |scratch|
47
+ File.chmod(0o700, scratch)
48
+ directory = File.join(scratch, @work.manifest.fetch("assignment_id"))
49
+ %w[steps reports].each { |name| FileUtils.mkdir_p(File.join(directory, name), mode: 0700) }
50
+ assignment = Models::Assignment.from_h(@work.definition, cache_dir: directory)
51
+ write(File.join(directory, "assignment.yaml"), assignment.to_h.to_yaml)
52
+ write(File.join(directory, "job.yaml"), @work.files.fetch("job.yaml"))
53
+ %w[steps reports].each do |group|
54
+ value.fetch(group).each { |filename, content| write(File.join(directory, group, filename), content) }
55
+ end
56
+ executor = Organisms::AssignmentExecutor.new(cache_base: scratch, selected_assignment: assignment)
57
+ case operation
58
+ when :activate
59
+ state = executor.status.fetch(:state)
60
+ invalid! unless state.find_by_number(scope)&.status == :pending
61
+ executor.step_writer.mark_active(state.find_by_number(scope).file_path)
62
+ when :start
63
+ executor.start_step(step_number: step, fork_root: scope)
64
+ when :finish
65
+ executor.finish_step(report_content: report, step_number: step, fork_root: scope)
66
+ when :fail
67
+ executor.fail(message, fork_root: scope)
68
+ else
69
+ raise ArgumentError, "unknown ordinary queue transition"
70
+ end
71
+ result = {"version" => 1}
72
+ %w[steps reports].each do |group|
73
+ result[group] = Dir.children(File.join(directory, group)).to_h do |filename|
74
+ [filename, File.binread(File.join(directory, group, filename)).force_encoding(Encoding::UTF_8)]
75
+ end
76
+ end
77
+ self.class.new(work: @work, root: @root, bytes: Atoms::EvidenceDigest.canonical_json(result))
78
+ end
79
+ end
80
+
81
+ private
82
+
83
+ def validate!(value, initial:)
84
+ invalid! unless value.is_a?(Hash) && value.keys.sort == %w[reports steps version] && value["version"].is_a?(Integer) && value["version"] == 1 &&
85
+ value["steps"].is_a?(Hash) && value["reports"].is_a?(Hash)
86
+ expected = @work.manifest.fetch("steps").map { |step| step.fetch("filename") }
87
+ invalid! unless value.fetch("steps").keys.sort == expected.sort
88
+ reports = []
89
+ value.fetch("steps").each do |filename, content|
90
+ parsed = @work.parse_queue_step!(content)
91
+ progress = parsed.fetch(:frontmatter).keys & PreparedWork::PROGRESS
92
+ allowed = initial ? %w[status] : %w[status started_at completed_at error stall_reason]
93
+ invalid! unless (progress - allowed).empty?
94
+ original = @work.files.fetch("steps/#{filename}")
95
+ invalid! unless PreparedWork.work_projection(content) == PreparedWork.work_projection(original) &&
96
+ %w[pending active done failed].include?(parsed.fetch(:frontmatter).fetch("status", "pending"))
97
+ info = Atoms::StepFileParser.parse_filename(filename)
98
+ reports << Atoms::StepFileParser.generate_report_filename(info.fetch(:number), info.fetch(:name))
99
+ end
100
+ invalid! unless (value.fetch("reports").keys - reports).empty?
101
+ value.fetch("reports").each_value { |content| @work.parse_queue_step!(content) }
102
+ invalid! if Atoms::EvidenceDigest.canonical_json(value).bytesize > PreparedWork::MAX_TOTAL
103
+ end
104
+
105
+ def write(path, content)
106
+ File.open(path, File::WRONLY | File::CREAT | File::EXCL | File::NOFOLLOW, 0600) { |file| file.write(content) }
107
+ end
108
+
109
+ def invalid!
110
+ raise AttemptErrors::EvidenceUnavailable, "ordinary canonical queue differs from its accepted work"
111
+ end
112
+ end
113
+ end
114
+ end
115
+ end