Graceful Replacement of Preemptive VMs in GCP

elastigroup gcp

We’re thrilled to announce that Elastigroup now gracefully replaces Preemptive VMs in GCP. Preemptible VMs on GCP provide a discount of up to 80% but are limited to up to 24 hours and may be preempted by GCP at any time. Elastigroup predicts the preemption and automatically replaces the VM, keeping your workload up and running. The new graceful replacement process eliminates the risk posed by mass VM replacement by spreading the process across scheduled batches, avoiding potential bottlenecks at the 24-hour Preemptible VM limit.

Here’s how the process works. Elastigroup continuously predicts VM preemptions and automatically replaces any VM that GCP has scheduled for preemption. To replace a Preemptible VM, Elastigroup launches a new Preemptible VM and validates its health, before draining and terminating the old VM. This seamless process guarantees that your cluster will be running at your desired capacity at all times, and has been extensively battle-tested by Spotinst customers across a wide variety of workloads.

In addition to the prediction of any preemptions, Elastigroup monitors each Preemptible VM to ensure that they’re replaced before GCP’s 24-hour time limit. Elastigroup designates any Preemptible VM running for more than 18 hours as eligible for replacement. These eligible VMs are scheduled for replacement in batches of no more than 10% of the Elastigroup, and a grace period between batches is maintained to provide adequate time for the replacement process to complete. The scheduled replacement batches protect your cluster from the possible interruptions and downtime that would otherwise occur when attempting to replace an entire cluster just before GCP’s 24 hour limit on Preemptible VMs.

Graceful replacement of Preemptible VMs is now available for all Elastigroups running on GCP and requires no action on your part. Head to the Spotinst Console to get started with Elastigroup on GCP today!