Dosu Logo
Home
Documents
Task_Management_Actions
Task_Management_Actions
Type
External
Status
Published
Created
Feb 27, 2026
Updated
Feb 27, 2026
Updated by
Dosu Bot

Task Management Actions#

Overview#

Implement task management capabilities allowing users to cancel, retry, reassign tasks, view logs, and download results. This ticket delivers the task detail investigation flow from spec:50650885-e043-4e99-960b-672342fc4139/c565e255-83e7-4d16-a4ec-d45011fa5cad.

Scope#

Included:

  • TasksController with actions: show, cancel, retry, reassign, logs, download_results
  • Task detail page with task information and action buttons
  • Agent compatibility validation for reassign
  • CanCanCan authorization rules for task actions
  • TaskActionsComponent for action buttons
  • Routes for task actions
  • Turbo Stream responses for inline updates

Excluded:

  • Task creation (handled by Agent API)
  • Task assignment logic (already exists in Agent#new_task)
  • Bulk task operations (out of scope)

Acceptance Criteria#

TasksController#

Show Action:

  • Route: GET /tasks/:id
  • Displays task details: state, agent, attack, campaign, progress, timestamps
  • Shows current status (from latest HashcatStatus)
  • Shows action buttons (cancel, retry, reassign based on state)
  • Authorization: project-based (can only view tasks in user's projects)

Cancel Action:

  • Route: POST /tasks/:id/cancel
  • Uses state machine event: task.cancel
  • Only available for pending/running tasks
  • Returns Turbo Stream response (update task + show toast)
  • Authorization: project-based

Retry Action:

  • Route: POST /tasks/:id/retry
  • Uses state machine event: task.retry (failed → pending)
  • Only available for failed tasks
  • Increments retry_count, clears last_error
  • Returns Turbo Stream response (update task + show toast)
  • Authorization: project-based

Reassign Action:

  • Route: POST /tasks/:id/reassign
  • Validates agent compatibility (hash type, performance, project access)
  • Updates task agent and resets to pending state
  • When reassigning a paused task from a paused attack, automatically resumes the attack
  • Returns error if agent incompatible
  • Returns Turbo Stream response (update task + show toast)
  • Authorization: project-based

Logs Action:

  • Route: GET /tasks/:id/logs
  • Displays HashcatStatus history (last 100 records)
  • Shows: timestamp, status, progress, hash_rate, temperature
  • Paginated if > 50 records
  • Authorization: project-based

Download Results Action:

  • Route: GET /tasks/:id/download_results
  • Generates CSV of cracked hashes for this task
  • CSV columns: Hash, Plaintext, Cracked At
  • Filename: task_{id}_results_{timestamp}.csv
  • Authorization: project-based

Agent Compatibility Validation#

  • Method agent_compatible_with_task?(agent, task) validates:

    • Agent supports task's hash type (check allowed_hash_types)
    • Agent meets performance threshold (check meets_performance_threshold?)
    • Agent has access to task's project (check project_ids)

  • Reassign UI only shows compatible agents (filtered dropdown/select)

  • Backend validation provides defense-in-depth

Authorization Rules#

  • CanCanCan abilities added to app/models/ability.rb:
    • can :read, Task, attack: { campaign: { project_id: user.project_ids } }
    • can :cancel, Task, attack: { campaign: { project_id: user.project_ids } }
    • can :retry, Task, attack: { campaign: { project_id: user.project_ids } }
    • can :reassign, Task, attack: { campaign: { project_id: user.project_ids } }
    • can :download_results, Task, attack: { campaign: { project_id: user.project_ids } }
    • can :manage, Task if user.admin?

Task Detail Page#

  • Task header shows: ID, state, agent, attack, campaign
  • Progress section shows: percentage, ETA, status text
  • Status history table shows recent HashcatStatus records
  • Error section shows task errors (if any)
  • Action buttons shown based on task state and permissions
  • Links to related resources (agent, attack, campaign)

Components#

  • TaskActionsComponent created with conditional button display
  • Component shows cancel/retry/reassign/logs/download based on task state
  • Component respects authorization (only show allowed actions)

Testing#

  • Request specs for all TasksController actions
  • System test: Cancel task, verify state change and toast
  • System test: Retry failed task, verify state change
  • System test: Reassign task to compatible agent
  • System test: Reassign to incompatible agent, verify error
  • System test: Download results CSV, verify content
  • Authorization specs for task actions
  • Task pause/resume actions respect the grace period for reassignment
  • Reassigning a paused task from a paused attack triggers attack resume
  • Paused tasks can be reclaimed by their original agent without waiting for grace period
  • Task state transitions properly set/clear the paused_at timestamp

Task Pause/Resume Mechanism#

Tasks can be paused during agent shutdown or disconnection. The pause/resume mechanism tracks task state across agent lifecycle events:

Pause Behavior:

  • The pause event sets paused_at: Time.current using update_columns to bypass optimistic locking
  • Claim fields (claimed_by_agent_id, claimed_at, expires_at) are cleared during agent shutdown
  • The agent_id column (NOT NULL) is retained, pointing to the original owning agent
  • This prevents StaleObjectError during cascaded operations when multiple tasks are paused simultaneously

Resume Behavior:

  • The resume event clears paused_at: nil and sets stale: true, also using update_columns
  • stale: true ensures the resuming agent re-downloads crack data
  • This bypasses optimistic locking to avoid concurrent update conflicts
  • See Task State Machine Transitions documentation for full state machine details

Grace Period for Reassignment:

  • After pausing, tasks use the paused_at timestamp to determine reassignment eligibility
  • Within the grace period (agent_considered_offline_time, default 30 minutes), only the original agent can reclaim the task
  • After the grace period expires, any compatible agent can claim the orphaned task
  • Tasks from offline/stopped agents are available for reassignment immediately, bypassing the grace period
  • This two-stage reclamation prevents tasks from being stolen during brief agent disconnections

Task Assignment Priority:

TaskAssignmentService uses two-stage task reclamation:

  1. Own paused tasks first: Agents reclaim their own paused tasks to leverage restore files and minimize redundant work
  2. Orphaned tasks second: After the grace period, agents can claim paused tasks from other agents

When reassigning a paused task from a paused attack (e.g., during shutdown cascade), the attack is automatically resumed to maintain consistent state.

Agent Shutdown Impact#

When an agent shuts down (via agent.shutdown! or agent disconnection), it triggers a cascade of state transitions:

Task Pausing:

  • All running tasks are paused and their claim fields cleared
  • The paused_at timestamp is set to track when the pause occurred
  • The agent_id remains set to the original agent for reclamation

Attack Pausing:

  • Attacks with no remaining active (non-paused) tasks are automatically paused during shutdown
  • This updates the Activity page to reflect that work has stopped
  • Paused attacks must be resumed when their tasks are reclaimed by other agents

Reassignment Considerations:

  • Paused tasks from a paused attack require the attack to be resumed during reassignment
  • TaskAssignmentService automatically resumes paused attacks when claiming their tasks
  • This prevents inconsistent state where tasks are active but their attack remains paused
  • Cross-reference the Agent Shutdown Cascade documentation for complete shutdown behavior

State Machine Integration#

Task state machine transitions use update_columns to avoid optimistic locking issues:

Why update_columns:

  • The pause/resume events bypass ActiveRecord callbacks and validations
  • This prevents StaleObjectError during cascaded operations (e.g., attack shutdown pausing multiple tasks)
  • The lock_version counter is not incremented, avoiding version conflicts
  • See StaleObjectError From Cascading Resume documentation for handling concurrent updates

State Transition Sequence:

pause event: task.update_columns(paused_at: Time.current)
resume event: task.update_columns(stale: true, paused_at: nil)

This approach ensures consistent state management across the task lifecycle without introducing race conditions during bulk operations.

Technical References#

  • Core Flows: spec:50650885-e043-4e99-960b-672342fc4139/c565e255-83e7-4d16-a4ec-d45011fa5cad (Flow 3: Task Detail Investigation)
  • Tech Plan: spec:50650885-e043-4e99-960b-672342fc4139/f3c30678-d7af-45ab-a95b-0d0714906b9e (TasksController, State Machine Extensions, Authorization)
  • Task State Machine Transitions: For detailed pause/resume behavior and paused_at timestamp management
  • Agent Shutdown Cascade: For agent shutdown impact on task and attack state
  • StaleObjectError From Cascading Resume: For handling concurrent updates during attack resume

Dependencies#

Requires:

  • ticket:50650885-e043-4e99-960b-672342fc4139/[Database Schema & Model Extensions] - Needs Task retry event
  • ticket:50650885-e043-4e99-960b-672342fc4139/[UI Components & Loading States] - Needs TaskActionsComponent, toast notifications

Blocks:

  • None (other features can be implemented in parallel)

Implementation Notes#

  • Create new controller: app/controllers/tasks_controller.rb
  • Create new views directory: app/views/tasks/
  • Use existing Task state machine events where possible
  • Test agent compatibility validation with various scenarios
  • Ensure CSV download works in air-gapped environment (no external dependencies)
  • Use Turbo Stream responses for all actions (progressive enhancement)

Estimated Effort#

2-3 days (controller + views + validation + tests)