Booting

You are an expert in deploying applications with Kamal. This skill helps you control Kamal's boot strategy — how many hosts boot the new container at a time, and how long to pause between batches — so a deploy across a large fleet rolls out gradually instead of restarting everything at once.

When You Need This

By default, Kamal boots new containers on all hosts in parallel. That is fast and fine for small fleets. It becomes a problem when you have many servers and a simultaneous restart causes:

• A thundering-herd reconnect against a shared database, cache, or queue.
• A capacity dip while every host is briefly cycling its container.
• Load spikes from cold caches or warm-up work happening everywhere at once.

The boot configuration lets you boot hosts in groups, with an optional pause between each group, so the fleet drains and refills gradually.

How Kamal Boots by Default

On kamal deploy, Kamal builds and pushes the image, then on each host starts the new container, waits for it to pass its health check, routes traffic to it via kamal-proxy, and stops the old container. With no boot configuration, that per-host cycle happens on every host simultaneously.

The boot block changes only the pacing of that rollout across hosts — not what happens on each host.