nibrunnerGuides
Sleep and wake
When an app sleeps, and what wakes it.
An app nobody is visiting is snapshotted and suspended. The next request wakes it. That is how a host holds more apps than it runs at once. Which apps sleep, and after how long, is the document's to say, per instance.
Desired state
desiredState says whether an instance should be up:
desiredState | What it means |
|---|---|
running | Keeps the microVM up. |
on-request | Brings it up for the first deploy and for every request that finds it asleep, and lets it sleep in between. |
stopped | Takes it down, and leaves the app reachable enough to say so. |
When it sleeps
activation.sleepWhen says what puts an on-request instance back to sleep:
{
"desiredState": "on-request",
"activation": {
"sleepWhen": { "kind": "traffic-idle", "timeoutMs": 900000 }
}
}sleepWhen | The microVM is suspended when |
|---|---|
{ "kind": "traffic-idle", "timeoutMs": N } | Nothing has been sent to it for N ms. |
{ "kind": "max-lifetime", "ttlMs": N } | It has been up for N ms, however busy it is. |
{ "kind": "never" } | Never. Only the document takes it down. |
An instance that names no sleepWhen sleeps after five minutes idle if it is on-request, and
never otherwise.
Good to know
- Only an
on-requestinstance may name asleepWhenother thannever: nothing would wake anything else, and the next reconcile pass would bring it straight back up. A document that says so is refused whole. idleTimeoutMsis the older spelling ofsleepWhen: traffic-idle. A document that names both is refused.
What wakes it
Any arrival: an HTTP request on one of its hostnames, or anything on a raw port. The connection is accepted and held, and spliced through once the app's health check passes, so a client sees a slow first response rather than a closed socket.