feat(coordinator): quorum verification for untrusted (volunteer) results

Replaces the C1 quarantine with real verification. Trusted results (lab token,
verified, or admin worker) are accepted directly as before. An untrusted
worker's result is recorded as one vote per (task, owner) in a new task_results
table; the task only completes once QUORUM_SIZE distinct owners submit the same
result hash, otherwise it returns to the queue for another independent compute.

- migration 0013 task_results (one vote per owner, quorum by result_sha256)
- CompleteTask branches on worker trust; unknown worker defaults trusted (safe:
  completing needs the lease, whose owner is always a known registered worker)
- claim drops the quarantine and excludes chunks the owner already voted on
- domain Task.ReleaseAfterVote; QUORUM_SIZE config (default 2)
- unit tests: trusted direct-complete, untrusted needs-quorum, can-claim

Reducer and job done/total logic untouched — still one completed task per chunk.
This commit is contained in:
Efremenko Arhip
2026-07-26 22:55:51 +03:00
parent dcabfcd0c3
commit 18d58cce84
12 changed files with 356 additions and 60 deletions
+31
View File
@@ -24,6 +24,10 @@ const (
// ErrCodeLeaseExpired marks tasks failed by the reaper rather than by a worker.
const ErrCodeLeaseExpired = "lease_expired"
// ErrCodeQuorumFailed marks a task whose untrusted results never reached a
// verifying quorum before its attempts ran out.
const ErrCodeQuorumFailed = "quorum_failed"
// Task is one independently executable chunk of a job.
//
// Nullable columns are pointers so "no lease" stays distinguishable from
@@ -216,6 +220,33 @@ func (t *Task) CompleteWith(resultArtifactID uuid.UUID, metrics map[string]any,
return nil
}
// ReleaseAfterVote returns an untrusted worker's task to the queue after its
// result was recorded as a quorum vote but quorum was not yet reached, so a
// different owner can compute it independently. When no attempts remain the task
// fails: its untrusted results could not be verified.
func (t *Task) ReleaseAfterVote(worker string, attempt int, now time.Time) error {
if t.Status == TaskCompleted {
return nil // settled by a concurrent quorum
}
if err := t.verifyLease(worker, attempt, now); err != nil {
return err
}
t.LeaseOwner = nil
t.LeaseExpiresAt = nil
t.Version++
if t.CanRetry() {
t.Status = TaskPending
return nil
}
code, msg := ErrCodeQuorumFailed, "untrusted results did not reach quorum"
t.ErrorCode = &code
t.ErrorMessage = &msg
t.Status = TaskFailed
t.CompletedAt = &now
return nil
}
// Fail records a worker-reported failure. A retryable failure with attempts
// left returns the task to the queue; otherwise it terminates as failed.
func (t *Task) Fail(worker string, attempt int, code, message string, retryable bool, now time.Time) error {