Answer the question
In order to leave comments, you need to log in
When starting 100 tasks (Job Array) on AWS Batch, about 25 will start immediately, and the rest are waiting in the RUNNABLE status. What is the limitation?
I am running Job Array on AWS Batch in a Fargate Spot environment. The main goal is to get some work done as quickly as possible. When I run 100 jobs, I expect all those jobs to run at the same time. But only about 25 of them are launched immediately, the rest are waiting for execution with the RUNNABLE status.
Jobs are executed in a computing environment with max. 256 CPUs. Each job uses 1 CPU.
I did not find any restrictions or quotas that can affect the process of completing jobs.
What could be the reason why not all of them start at once, but only some of them?
Answer the question
In order to leave comments, you need to log in
Contacted AWS Support about this issue. They confirmed the problem, but did not find a solution for it.
It was advised to use On-Demand instead of Fargate when you need to complete a large amount of tasks as quickly as possible.
After switching to the On-Demand provisioning model, the problem was solved.
Now the maximum number of concurrent tasks is launched, specified in the compute environment settings.
This is just what I needed.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question