Flexera logo
Image: Reduce Cloud Costs: Watch Out for Zombie Instances

It happens over and over again: You receive a cloud bill that is higher than expected. After some investigation, you find there are a bunch of zombie instances — unused servers left running because no one turned them off or an instance that failed in booting. Multiply the number of zombies by each cloud you use and you get a zombie apocalypse that is costing you a lot of money. Wouldn’t it be great if you could find all of the zombie instances across all the clouds you use so that you can reduce cloud costs?

As more enterprise IT teams broker cloud services to cloud users, they are providing a multi-cloud self-service portal to development teams, testing teams, sales teams for demos, training teams who need software environments for classes, and any cloud user that needs infrastructure on demand. As they provide cloud access to diverse teams across their organizations, it’s easy for cloud bills to go up as users neglect to shut down instances that were needed for temporary use or instances that failed to properly boot. RightScale helps you to reduce cloud costs by finding and killing unused or unresponsive instances across all the clouds you use.

How to Reduce Cloud Costs

1. Kill Instances That Are Stranded In Booting

When launching cloud instances in RightScale Self-Service or RightScale Cloud Management, you may have instances that are successfully provisioned from the cloud provider (and therefore are incurring charges) but fail later in the launch process and therefore are not fully operational. This most commonly happens if there are errors in configurations scripts. Using RightScale Cloud Analytics, you can find these instances and kill them.

RightScale tracks the state of each instance across all of your clouds. From the Analyze page in RightScale Cloud Analytics, first click the checkbox to filter by the “stranded in booting” state.

Then click the Full instance details tab to see the details of all the stranded instances including instance ID. From here, you can click the instance ID, the deployment name, or the cloud, which will open this instance in Cloud Management, and you can take action, such as terminating the server, to optimize your costs.

2. Kill Instances That Are Unused or Unresponsive

Identifying unused or unresponsive instances can be trickier. For example, you might have an instance that shows 0 percent CPU utilization and appears unused, but that instance might be part of a warm failover scenario and shouldn’t be shut down. There are several approaches you can use for handling these situations.

The first approach to shut down unused or unresponsive instances leverages RightScale Cloud Management to set up monitoring alerts based on any relevant criteria (CPU utilization, memory utilization, etc). These alerts can be included in standard RightScale ServerTemplates™ that apply to all instances that use those ServerTemplates.

Once the alerts are defined, you can set up escalation actions that run a script to terminate the instance if it meets your criteria for unused instances or relaunch the instance if it meets your criteria for unresponsive instances. If you prefer manual intervention, then the escalation can generate an automatic email notification to the owner to let him know that his instance is unused.

A second approach to shut down unused or unresponsive instances is to use RightScale Self-Service to define a scheduled workflow that runs a script to look for these particular instances and take action. At RightScale, for example, we run a script that looks across our Services account for instances that have been running for more than 24 hours, are not locked, and don’t have “Save” anywhere in the name. Your specific criteria may be different.

A third approach is to prevent the problem in the first place. If you are launching instances through RightScale Self-Service, use end dates and schedules within Self-Service to automatically shut down the instances once you are done with them or turn them off during hours or days when they are not used.

By using the monitoring and detection features available in RightScale, you can reduce cloud costs across your entire organization.