Explain like I'm five · SRE

What is a Health Check?

A health check, told as a story about asking every ride 'are you okay?' every few seconds.

RUNNINGhuh, it's not moving?the ride is stalled, but the sign still says 'running'

The ride is stalled, but the sign still says 'running.'Guests wait in line only to find out it's not running.

all looks fine to me?from a distance it looks fine, but inside it may have already stopped

From a distance it looks fine, but inside it may have already stopped.Looking at the screen alone can't tell you what's inside.

OKOKXno answer → removed from the listevery few seconds, ask each ride 'click — are you okay?' — no answer, and it's pulled from the listget well again, and it comes back on the list on its own

Every few seconds, ask each ride 'click — are you okay?' — no answer, and it's pulled from the list.Get well again, and it comes back on the list on its own.

Ask it, again and againevery few seconds
No answer? Off the listso no guest gets sent there
Ask two different thingsalive? and ready?
Recovers? Back automaticallyno one has to do it by hand
STATUS BOARDNo.1: alive ✓ ready ✓No.2: alive ✓ not ready ✗No.3: no answer → removedtwo different questions — are you alive? are you ready for guests?

Two different questions — are you alive? are you ready for guests?A ride can be alive but still not ready yet.

asking this often wears us out toochecking too often becomes its own burden — you need the right interval

Checking too often becomes its own burden.You need to find the right interval.

In one breath

A health check = asking every ride 'are you okay?' every few seconds, and pulling it from the list when there's no answer — then letting it back in once it's well again.

A method of pinging a service at regular intervals to see if it's still alive. Whether it's alive (liveness) and whether it's ready for traffic (readiness) are two different questions, and the result feeds straight into a load balancer, updating the list of who's in service.

When grown-ups say it

Health check
Asking 'are you okay?' every few seconds. No answer means it's pulled from the list.
Liveness
The question 'are you alive at all?' Passing this alone doesn't mean everything's fine.
Readiness
The question 'are you ready for guests?' A ride can be alive but still not ready yet.
Ping / Heartbeat
The short click-signal sent out. The reply to this signal is what tells you the status.
Automatic removal and return
Dropping out on no answer, and returning once well again. No one has to manage this by hand.
Load balancer integration
Where the health-check result goes next. It only sends guests to rides that passed. → the usher who spreads guests across booths
Failover
Switching to another ride instead of one that's not answering. A failover can only start once the health check has already noticed.
Service discovery
The sign that shows which rides are currently open. Health-check results are what keep this sign always up to date.
Next story
The Sign That Shows Which Booths Are Open →
Want another term as a picture book? Request →