Set up health checks

An Auto Scaling Group replaces a VM that is unhealthy, so the group keeps its size. Every group checks its VMs and hosts. You can add a check of your own application over HTTP or TCP, which makes the group replace a VM whose application stops answering.

Before you begin

• An Auto Scaling Group, or the Create Auto Scaling Group form open. See Create an Auto Scaling Group.
• For an HTTP or TCP check, the port, and for HTTP the path, that your application answers on.

Choose a check

Health settings are in Advanced Settings, under Health & recovery, on the create form and on Edit Auto Scaling Group.

Infrastructure health checks are always on: the group detects VM and host failures and replaces an unhealthy VM.

Application health check has three choices:

• VM status only is the default. The group checks only the VM and host status, and a VM is healthy when it reports ready. To check your application itself, choose HTTP or TCP.
• HTTP: the group calls an HTTP endpoint on each VM and replaces any that fail.
• TCP: the group opens a TCP connection to a port on each VM and replaces any that fail.

Set up an HTTP or TCP check

Choose HTTP or TCP, then set the check. The defaults are a good place to start:

Field	Default	Rule
Protocol (HTTP only)	HTTP	HTTP or HTTPS
Port	80	1 to 65535
Path (HTTP only)	/	Starts with /
Interval (seconds)	30	At least 15
Timeout (seconds)	5	From 1 up to the interval
Unhealthy after (fails)	3	At least 1
Healthy after (successes)	2	At least 1

The check goes to the VM’s private IP. An HTTP check passes on any answer from 200 to 299, and an HTTPS check accepts a self-signed certificate. A TCP check passes when it opens a connection within the timeout. A VM must fail Unhealthy after checks in a row before it is replaced, and it is cleared after Healthy after passes.

Then tell the group how long your app takes to start: Under 30s, 1 minute, 3 minutes, 5 minutes (the default) or Custom in seconds. The group ignores health checks while a new VM is starting.

What happens to a VM that fails

The group builds a new VM from the source snapshot, so the new VM starts with only what the snapshot holds. Keep data you need somewhere else. Because a failed check replaces the VM, point the check at a port and path that only fail when your application is down. The default port 80 and path / are a starting point.

Checks run only while the group is Running.

See the result

Scaling Activity records “Replaced an unhealthy VM” with the reason, such as a timeout, a refused connection or an HTTP error, the time it first failed and the target. On VMs & Monitoring, the removed VM is marked “Replaced after failing health checks”, and the group’s Configuration on Overview names the check in force.

To see how scaling and replacement fit together, read How auto scaling works. To change capacity by usage, see Scale on usage.