Unpausing index job not resuming

[ryanwinchester]

Check for other issues

• I've checked for other issues, and none exist.

Issue description

When I added a Local Folder storage, and then pause it, and unpause it, it doesn't actually resume.
It is stuck at 37% indefinitely.

I have quit and opened spacedrive and it is still not progressing.

Steps to reproduce

1. Click on "Add Storage"
2. Click on "Local Folder"
3. Select any folder and "Content" index mode
4. Wait for some progress.
5. Click "Pause job" button
6. Click "Resume job" button

Platform and versions

• OS: macOS
• Spacedrive: v2.0.0-alpha.1

Stack trace

Additional context

No response

Code of Conduct

• I agree to follow the Code of Conduct.

[tembo]

Issue Validation

This is a confirmed bug in the job system's resume logic. When you pause and resume an indexing job that's still in memory (hasn't been serialized to disk yet), the task system never actually receives the resume signal, so the job remains stuck in a paused state despite the UI showing it as running.

Relevant Code

• core/src/infra/job/manager.rs:2026-2061 - The in-memory resume path that's missing the task system resume call
• core/src/infra/job/manager.rs:1745-2025 - The full resume_job function with both paths
• crates/task-system/src/task.rs:527-540 - TaskRemoteController::resume() that should be called
• crates/task-system/src/worker/runner.rs:210-232 - Runner::resume_task() in the task system

What's Happening

The resume_job() function in core/src/infra/job/manager.rs has two code paths:

1. Job NOT in memory (lines 1783-2025): Deserializes the job from database, creates a new executor, and dispatches it to the task system. This works correctly.

2. Job still in memory (lines 2026-2061): Only updates the status_tx channel to JobStatus::Running and updates the database. It never calls task_handle.resume() on the task system.

When you pause a job, the task system moves the task from the active runner to paused_tasks HashMap (see runner.rs:876). To actually resume it, TaskRemoteController::resume() must be called, which sends a message to the task system to move the task back to the active queue.

The in-memory path updates the job's logical status but the task system still considers the task paused, so the job executor's run() method is never called again.

Potential Approaches

1. Add missing task system resume call - In the "job is in memory" branch (around line 2034), after updating the status channel, call running_job.task_handle.resume().await to actually signal the task system:

   // After line 2051 (database update), add:
   if let Err(e) = running_job.task_handle.resume().await {
       warn!("Failed to resume task for job {}: {}", job_id, e);
   }

2. Handle persistence-complete race condition - When a job is paused, the executor saves state and signals persistence_complete_tx. If resume is called before this completes, there could be a race. May need to await persistence_complete_rx before resuming.

3. Unify resume paths - Consider restructuring to always go through the database path (deserialize and re-dispatch), which would be more robust but potentially slower for quick pause/resume cycles.

Investigation Notes

• The branch name tembo/fix-index-job-resume-stuck suggests this issue is already being investigated
• The indexer job correctly implements is_resuming() and on_resume() hooks, so the job itself handles state restoration properly
• The bug is in the job manager layer, not the indexer job implementation
• This affects all job types, not just indexing - any job paused and resumed while still in memory would hit this bug

The fix should be straightforward: add the missing task_handle.resume().await call in the in-memory resume path.