Most of our EKS clusters leverage Spot instances as a cost-efficient way to provide compute nodes. Historically we’ve been defaulting to the “lowest price” allocation strategy to maximize possible cost savings. However this can lead to quite some more interuptions than we want to tolerate and often a big inbalance between AZ spread if price pressure increases. Therefore we’ve updated our default to use a “capacity optimized” strategy instead for increased stability with (possibly) a marginal higher cost.
This change is in effect for all clusters with spot pools where no specific allocation strategy was defined. Most of our setups had already been specifically updated for capacity optimization in the past. If you instead wish to opt for a different strategy for your workload’s needs, please get in touch with us.