jargon

Comparison

Spot instancevsSpot interruption

Spot instance

the machine costs a fraction of the normal price and can be taken away from you with two minutes' notice.

Buying spare capacity at a steep discount on the condition that the provider may reclaim it. It is excellent for anything interruptible and replayable — batch, builds, stateless workers — and hostile to anything that cannot be moved quickly. Using it well means diversifying across instance types and zones so a reclamation in one pool does not take your whole fleet.

Full entry →

Spot interruption

a two-minute warning arrives and everything on that node has to drain, checkpoint and be gone before it is killed.

The provider reclaiming spot capacity, preceded by a short notice. Handling it well is a workload property rather than a platform setting: the node must cordon and drain on the signal, and the workload must shut down cleanly within the window. It is also useful as a permanent, free chaos experiment, since a fleet that handles interruptions gracefully handles node failure gracefully too.

Full entry →

Related comparisons