Open qiukunlong opened 7 years ago
Hello @qkl871118
Thank you for raising this issue, I appreciate it.
cwltool
, the reference implementation of the CWL standards, doesn't currently implement this feature as it doesn't schedule jobs on remote resources, it just runs them locally.
This could be achieved for jobs that are run in a Docker container: https://docs.docker.com/engine/admin/resource_constraints/#memory PRs that achieve this are very welcome.
For local running without Docker containers then that is still possible, but more difficult and less portable.
In CWL CommandLineTool Document, it is described that "If a job attempts to exceed its "max" resource allocation, an implementation may deny additional resources, which may result in job failure".
I set ramMax: 2000 in my CWL, and use cwltool to run the CWL, the job exceed the max memory I set and achieve 6G memory. In my opinion, the job memory used should limited at a upbound of 2G.