Shut down services gracefully

Prompt that defines a shutdown sequence so deployments do not abort requests or lose jobs. Paste it into an AI tool of your choice, which then defines the shutdown sequence.

Topic: Apps & prototypes
Type: Prompt
Level: Intermediate
Time: about 15 min

Results depend on the tool, the model and what you enter.

Result preview

Example: written by our editors, not real AI output.

Target sequence after SIGTERM: 0 s health check reports "not ready" 0–5 s load balancer removes the service 5–25 s finish running requests 25 s return open jobs to the queue 28 s close connections, process exits Platform deadline: 30 s (fits) Test: 200 requests/s, deployment, 0 aborted requests

Fill in

What you type stays in this browser window. We store and send nothing.

ServiceWhat kind of service?

EnvironmentWhere does it run?

SymptomWhat happens during a deployment?

Your prompt

Help me build a service that shuts down cleanly.
Service: [Service]
Environment: [Environment]
Symptom: [Symptom]
Steps: 1. Read the startup code, the signal handling and the runtime settings and describe what happens today on shutdown. 2. Define the order: stop accepting new requests, set the health check to not ready, finish running requests, complete or hand back jobs, close connections. 3. Check that the service's wait time is shorter than the deadline after which the environment kills it. 4. Find work that would be lost on a hard kill, such as unacknowledged messages, half-written files or buffers that are never flushed. 5. Propose code for the signal handling, with a time limit and a log line per step. 6. Describe a test that stops the service under load and counts aborted requests.
Output: current sequence, target sequence with timings, code proposal and test plan.

Take particular care to let running work finish cleanly before the process stops.

Before you answer, check: Are all instructions met? If key information is missing or a placeholder in square brackets is still in the text, ask instead of guessing. Mark assumptions as assumptions.

Help me build a service that shuts down cleanly.
Service: [Service]
Environment: [Environment]
Symptom: [Symptom]
Steps: 1. Read the startup code, the signal handling and the runtime settings and describe what happens today on shutdown. 2. Define the order: stop accepting new requests, set the health check to not ready, finish running requests, complete or hand back jobs, close connections. 3. Check that the service's wait time is shorter than the deadline after which the environment kills it. 4. Find work that would be lost on a hard kill, such as unacknowledged messages, half-written files or buffers that are never flushed. 5. Propose code for the signal handling, with a time limit and a log line per step. 6. Describe a test that stops the service under load and counts aborted requests.
Output: current sequence, target sequence with timings, code proposal and test plan.

Take particular care to let running work finish cleanly before the process stops.

Before you answer, check: Are all instructions met? If key information is missing or a placeholder in square brackets is still in the text, ask instead of guessing. Mark assumptions as assumptions.

This is an instruction for an AI, not the result itself.

Where this prompt works

These AI tools are a good fit for this prompt:

How to use it

Fill in the fields or tap a suggestion. Leave out names, addresses and other personal data.
Copy the finished prompt into an AI tool of your choice.
Paste the startup code, the signal handling, the health check and shutdown deadline settings and a log excerpt from a deployment.

Why does this work?

Every deployment stops processes. Without a fixed sequence, each release hits a few users, and jobs quietly lose data.

Tips

Test the sequence on a staging system under realistic load before you change deadlines in production.

You may use our prompts freely, including at work. Please do not republish them as your own collection. We give no warranty for AI results; check them yourself. Terms of use

Shut down services gracefully

Result preview

Fill in

Your prompt

Where this prompt works

How to use it

Why does this work?

Tips

Also a good fit

Choose pagination for an API list New

Write an API reference people can use right away New

Plan a throttled data backfill New