GCP: Instance group

Managed instance group contains identical instnaces, created from an instance template. Supports auto scaling, auto healing, rolling updates, load balancing. VM instances are stateless and disks are deleted on VM recreation. It is possible for load balancer to send traffic to instance group through a named port.

Configure autoscaling on and autoscaling policy. Auto scaling policy directs when to auto-scale, based on CPU utlization, HTTP load balancing utilization, Stack driver metric, or combination of the above. It is possible to specify maximum number of instances that can be in instance group. Enable auto healing through healhcheck configuration.

Instance group while running, you can add/remove labels. You can do modifications just as you can do with an instance running.

With instance group you can have rolling update, rolling restart/replace actions. When performing rolling update, perform canary deployment, and make sure that no rollback is needed. If rollback is needed, turn the traffic back to older template.

It is also possible to rolling restart or replace with Instance group.
Note that Instance template cannot be edited

Sample question:

A recent software update to an e-commerce website running on Google Cloud has caused the website to crash for several hours. The CTO decides that all critical changes must now have a back-out/roll-back plan. The website is deployed on hundreds of virtual machines (VMs), and critical changes are frequent. Which two actions should you take to implement the back-out/roll-back plan? (Choose two)

- Turn on the object versioning of GCS, let the previous version of static assets be maintained.
- Use managed instance groups with the “update-instances” command when starting a rolling update

Note:
Unmanaged instance group is for load balancing dissimilar instances that can be added/removed arbitrarily. Auto scaling, auto healing, rolling updates are not supported for unmanaged instance group.