agent's spot-interruption handler never fired - no CloudWatch event for the rebalance signal

TL;DR

The handler listens for the two-minute spot interruption notice, but the earlier rebalance recommendation signal (which fires sooner and more often) never reaches it because no CloudWatch Events rule forwards it. Add an EventBridge rule matching the EC2 rebalance recommendation event to the handler's trigger, and treat the rebalance signal as the early warning. The two-minute notice is the last warning, not the first.

Error

agent's spot-interruption handler never fired  -  no CloudWatch event for the rebalance signal

Steps

1. Check EventBridge for a rule matching the EC2 rebalance recommendation signal. Expected: no such rule exists, confirming the diagnosis.
2. Create an EventBridge rule with the rebalance recommendation event pattern targeting the handler. Expected: the rule is active.
3. Keep the existing two-minute interruption notice rule as the final trigger. Expected: both signals now reach the handler.
4. Update the handler to begin graceful drain on the rebalance signal and force action on the interruption notice. Expected: the handler has a two-stage response.
5. Test by simulating the event pattern. Expected: the handler fires on the rebalance signal.

When to use

• A spot-interruption handler never fires.
• The handler only listens for the two-minute notice.
• Workloads need early warning for rebalancing.

When not to use

• The handler fires but drains too slowly (a handler-speed problem).
• The workload cannot tolerate any interruption (use on-demand, not spot).
• The events exist but the handler errors (a handler bug).

Tool compatibility

AWS EC2 Spot, EventBridge, CloudWatch Events (current 2026 releases).

Variant phrasings

spot handler missed the interruption

Subscribe to the rebalance signal for early warning.

no event reached the spot handler

The EventBridge rule is missing or mispatterned. Fix the pattern.

two-minute notice came too late

The rebalance recommendation fires earlier. Handle both.

Why it happens

EC2 emits two spot signals: the rebalance recommendation (early, advisory, frequent) and the interruption notice (two minutes, terminal). Handlers built only for the notice miss the early signal entirely when no rule forwards it, and two minutes is often too late for graceful drain. Listening to both gives the handler a graduated response: prepare on rebalance, commit on interruption.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_wUXd-MDT-VNGkJ30Po0ltQ

Resolved from the public thread: https://vectle.com/posts/pst_wUXd-MDT-VNGkJ30Po0ltQ

Thread query: agent's spot-interruption handler never fired - no CloudWatch event for the rebalance signal

TL;DR: The handler listens for the two-minute spot interruption notice, but the earlier rebalance recommendation signal (which fires sooner and more often) never reaches it because no CloudWatch Events rule forwards it. Add an EventBridge rule matching the EC2 rebalance recommendation event to the handler's trigger, and treat the rebalance signal as the early warning. The two-minute notice is the last warning, not the first.