Sandbox Execution¶

AsyncMQ can execute handlers in a subprocess sandbox.

Enable It¶

from asyncmq.conf.global_settings import Settings


class AppSettings(Settings):
    sandbox_enabled = True
    sandbox_default_timeout = 30.0
    sandbox_ctx = "spawn"  # or "fork" depending platform/runtime needs

When enabled, worker runtime calls sandbox.run_handler(...) instead of executing the handler directly.

Behavior¶

run_handler(task_id, args, kwargs, timeout, fallback=False):

• executes task in child process
• waits up to timeout
• on timeout, terminates child
• raises TimeoutError by default after the child is terminated
• if fallback=True, retries execution in current process after timeout

Tradeoffs¶

Pros:

• isolates crashes from main worker process
• hard execution timeout boundary

Cons:

• extra process startup overhead
• task args/results must be process-serializable
• fallback=True weakens timeout isolation because timed-out handler code is executed in the current worker process

Practical Notes¶

• Prefer spawn in environments where fork semantics are problematic.
• Keep timeout realistic for your workload.
• Leave fallback disabled for production queues unless the handler is trusted, idempotent, and safe to run after already timing out in the child process.
• For performance-critical short jobs, benchmark sandbox overhead before enabling globally.