The systems layer beneath the current work.

A service crashes. Restarting it is easy. Deciding whether to try again, how long to wait and when to stop is the part your operators shouldn't have to improvise.

Kestowv's hosted pilot puts health checks, restart limits and backoff into a defined service policy. A controlled shutdown can drain work before the service stops. The agreed policy defines the retry budget, delay and conditions for stopping.

• Check whether the service is healthy, not merely present.
• Limit retries and leave time between them.
• Stop or drain work through the declared recovery path.

A new version is more than a folder with a different name. The pilot stages versioned files, checks their hashes and tests the release against the health criteria you've supplied. If activation fails that gate, it restores the previous release.

That gives the operator two useful answers: which version was activated, and whether it met the agreed checks. Recovery still needs to account for your application's data and dependencies; restoring a release isn't a substitute for a database recovery plan.

One busy service shouldn't be allowed to consume every resource its neighbors need. On a supported Linux pilot host, Kestowv can apply declared CPU, memory, I/O and process limits through the host's controls.

Define the budget before the run. Check that enforcement is available on the machine. Then test the normal workload, a failed service and the recovery path against the same expectations.