Lab 6.L2: a health check that fits the app

You need csph pointed at your learn project and lab environment (lesson 5.1.1), room for one Flex spherelet, and curl.

1. Deploy the slow shop

   Copy

   cat > shop-slow.yaml <<'EOF'
   version: "2"
   services:
     - name: shop-slow
       type: web-service
       image: quay.io/computesphere/learn-shop-api:1.4.0
       port: 8080
       plan: flex
       env_vars:
         START_DELAY: "150"
         REQUIRE_SIGNING_KEY: "false"
   EOF
   csph deploy --file shop-slow.yaml
   

   With no health check yet, ComputeSphere only checks that the port answers, so this goes Running early.

   You should seecsph prints Running and the URL within a minute or so.

2. Measure its start-up

   Copy

   curl https://<your-shop-slow-url>/healthz
   

   Wait a minute after Running before the first call. If curl says it could not resolve host, wait and retry. Repeat until it says ok: that's the start time your check has to fit.

   You should see{"ready_in_seconds":…,"status":"starting"} at first; about 150 seconds after the start, {"status":"ok"}.

3. Add a health check that's too eager

   In the console, open shop-slow, Settings. Under Health check, choose Add health check and set Endpoint path /healthz, Port 8080, Initial delay 5, Check interval 10. Choose Save, then Redeploy now, and open Logs, Deploy.

   You should seeCreated and Pulling image in the deploy log: a new version is starting.

4. Watch the restart loop

   Open Logs, then Deploy, then Runtime. Give it five minutes. Your URL keeps answering: the old version serves until the new one passes. Left alone, the update ends Failed after about ten minutes.

   Check yourself

   Why does the new version never get ready?

   It's restarted before its 150-second start-up finishesIt crashes on start

   You should seeUnhealthy: Health checks are failing. in the deploy log, then Stopping, and the start-up lines repeating in the runtime log about every two minutes.

5. Give it time to start

   In Health check, choose Edit, set Initial delay to 170, then Save and Redeploy now.

   You should seeLive in the deploy log about three minutes later, and no more restarts.

6. Check it answers

   Copy

   curl https://<your-shop-slow-url>/healthz
   

   You should see{"status":"ok"}

csph services list
csph services delete <shop-slow-service-id>
rm shop-slow.yaml