Pre-flight checks to compare requested resources to available AWS Batch resources | Voters

Pre-flight checks to compare requested resources to available AWS Batch resources

acknowledged

Ken Brewer

It would be helpful to include Seqera Platform-based pre-flight checks to compare requested system resources versus the maximum capacity of EC2 instances available to the AWS Batch cluster. This could prevent jobs with excessive resource demands locking up a compute environment.

January 29, 2024

Sapphire Canidae

We ran into a similar issue where we had requested 1000 GB RAM for a process, but the largest instance type in the autoscaling group we had provisioned was an r6id.32 xlarge (128 CPU and 1024 GiB RAM). This resulted in a blockage in our queue, with lots of jobs stuck in "RUNNABLE". It was only when we lowered our process resource request to 900GB that we were able to unblock the queue. Having a preflight check or warning would have been an incredibly useful feature.

Rob Newman

marked this post as

acknowledged

Yellow sunshine Firefly

We have run into issues where tasks requesting more resources than the cluster could ever give cause other jobs to get stuck in line,