World O2: Counters · Boss challenge

Count the orders

You will learn: to answer five real questions about Kedai Kancil with PromQL.

Checked with Prometheus 3 on 1 Nov 2026

The problem

Boss: the owner has five questions, and it is 28m. Question 1: which target is down?

Show only the targets that are down.

PromQL query

Metrics in this data (tap to add):

Eval time: 28m

Time 0 is when the data starts. The query looks back from the eval time.

Predict

shop-2 is down. Who can still tell you whether customers reach the shop?

Try it

Question 2: orders per minute, one line per stall.

Use rate, sum by (stall), and turn seconds into minutes.

PromQL query

Metrics in this data (tap to add):

Eval time: 28m

Time 0 is when the data starts. The query looks back from the eval time.

Check

Does it answer “orders per minute”? Swipe right for yes.

Swipe right for good, left for needs work. Or use the buttons.

1 / 4rate(kedai_orders_total[5m]) * 60

Try it

Question 3: what share of web requests fail with a 5xx code?

Divide the 5xx rate by the rate of all requests. One number.

PromQL query

Metrics in this data (tap to add):

Eval time: 28m

Time 0 is when the data starts. The query looks back from the eval time.

Check

Next comes p95. Put the recipe in order, from the inside out.

Inside at the top. Put these in order. Use the arrows, or drag.

kedai_checkout_duration_seconds_bucket
histogram_quantile(0.95, … )
sum by (le) ( … )
rate( … [5m])

Try it

Question 4: p95 checkout time, both shops together.

Build the p95 recipe and Run.

PromQL query

Metrics in this data (tap to add):

Eval time: 28m

Time 0 is when the data starts. The query looks back from the eval time.

Predict

p95 is 1.86 s. The SLO says 95% of checkouts finish under 1 s. Is it met?

Try it

Last question: how many orders came in the last 10 minutes, all stalls together?

Use increase over 10 minutes, then add everything up.

PromQL query

Metrics in this data (tap to add):

Eval time: 28m

Time 0 is when the data starts. The query looks back from the eval time.

You did it!

You can now answer real questions about a shop with PromQL.

Go deeper (optional)

Real rep: ask your own kedai-watch stack the same kinds of questions.

Real mission

In your kedai-watch folder, start the stack.
docker compose up -d
Which targets are down? Run up == 0. Stop node-exporter to see one appear.
docker compose stop node-exporter
Requests per second, per handler: sum by (handler) (rate(prometheus_http_requests_total[1m]))
p90 of Prometheus’s own pages, with histogram_quantile.
Stop everything.
docker compose down

0 of 5 steps done · Your ticks are saved.