fix(tasks): hydrate runner progress after single-node restart

[cursor]

Bug and impact

1. Remote tasks stuck after single-node restart (critical)

When Semaphore restarts while remote runner tasks are still executing, TaskPool.GetTask only hydrated from the database in HA mode. On a single-node deployment the in-memory pool is empty after restart, so UpdateRunner silently dropped all progress reports. Tasks remained stuck in running in the database even after the runner completed successfully.

Trigger: Restart the Semaphore server (upgrade, crash, deploy) while one or more remote runner tasks are actively executing.

2. History page crash on websocket events (high)

History.vue called this.items.some() without a null guard when handling websocket updates. If an update arrived before the first paginated load finished (or after a failed load), the History view threw TypeError: Cannot read properties of null.

Trigger: Open project History while tasks are actively updating; websocket event arrives before /tasks/last returns.

Root cause

• HA work added HydrateTaskRunnerFromDB behind an util.HAEnabled() guard in GetTask, leaving single-node restarts without the same recovery path already used across HA nodes.
• Task pagination added websocket reload throttling to History.vue but omitted the null guards already present in TaskList.vue.

Fix and validation

• Always hydrate from DB in GetTask when a task is not in the in-memory pool (same behavior HA already relied on).
• Mirror TaskList.vue null guards and throttle state initialization in History.vue.
• Added TestUpdateRunner_RunningTaskInDBButNotInPoolAccepted covering post-restart runner completion.

go test ./api/runners/... ./services/tasks/... -count=1

  

[cursor]

Stale comment

Security review

Outcome: No medium, high, or critical vulnerabilities found.

Reviewed the added/modified code in services/tasks/TaskPool.go, api/runners/runners_test.go, and web/src/views/project/History.vue.

TaskPool.GetTask hydration change

The PR removes the util.HAEnabled() guard so HydrateTaskRunnerFromDB runs whenever a task is absent from the in-memory pool (e.g. after a single-node restart). This expands an existing HA code path; it does not add a new externally reachable entry point.

Traced the primary sink (UpdateRunner → taskPool.GetTask → status/log updates / FinalizeRemoteTask). Existing controls remain in place:

• Authentication: RunnerMiddleware requires a valid X-Runner-Token before UpdateRunner runs.
• Authorization: tsk.Task.RunnerID must match the authenticated runner; mismatches are added to TerminatedJobs and skipped.
• State guards: finished tasks are rejected; invalid statuses return 400; TryFinalize deduplicates terminal finalization.

A runner cannot use hydration to update tasks assigned to other runners or bypass runner authentication.

History.vue null-safety

The websocket handler changes only add null guards (this.items != null, this.items || []). They prevent runtime errors when the list is uninitialized and do not introduce new rendering of attacker-controlled data.

Prior automation threads

No unresolved security-review threads were present on this PR.

  

_{Sent by Cursor Automation: Find vulnerabilities}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 299970ada8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep GetTask scoped to active tasks

In single-node mode this changes GetTask from a queue/running lookup into “load any DB row”. RemoveTask treats any nonnil result from taskPool(r).GetTask as an active task and returns 400 before deleting (checked api/projects/tasks.go:423-434), so completed tasks that are no longer queued or running now hydrate here and cannot be deleted. Consider keeping this fallback on a runner-progress-specific path, or filtering it to unfinished remote-runner tasks so inactive historical tasks still return nil.

Useful? React with 👍 / 👎.

[cursor]

Security review

Outcome: No medium, high, or critical vulnerabilities found.

Re-reviewed the PR after the latest synchronize (head 9ba363ae), including services/tasks/TaskPool.go, api/runners/runners_test.go, and web/src/views/project/History.vue.

TaskPool.GetTask hydration change

Removing the util.HAEnabled() guard expands an existing HA code path so HydrateTaskRunnerFromDB runs whenever a task is absent from the in-memory pool (e.g. after a single-node restart). This does not introduce a new externally reachable entry point.

Traced the primary sink (UpdateRunner → taskPool.GetTask → log/status updates / FinalizeRemoteTask). Existing controls remain effective:

• Authentication: RunnerMiddleware requires a valid X-Runner-Token before UpdateRunner runs.
• Authorization: tsk.Task.RunnerID must match the authenticated runner; mismatches are added to TerminatedJobs and skipped (api/runners/runners.go).
• State guards: finished tasks are rejected; invalid statuses return 400; TryFinalize deduplicates terminal finalization.

A runner cannot use hydration to update tasks assigned to other runners or bypass runner authentication.

History.vue null-safety

Websocket handler changes only add null guards (this.items != null, this.items || []). They prevent runtime errors when the list is uninitialized and do not introduce new rendering of attacker-controlled data.

Prior automation threads

No unresolved security-review threads from earlier automation runs. Prior assessment validated on the updated head commit.

  

_{Sent by Cursor Automation: Find vulnerabilities}