-
Notifications
You must be signed in to change notification settings - Fork 51
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
FLUX_JOB_ID is set in batch scripts #4716
Milestone
Comments
garlick
added a commit
to garlick/flux-core
that referenced
this issue
Oct 25, 2022
Problem: FLUX_JOB_ID and other job environment variables remain set in the initial program if Flux was launched by Flux. This sends the wrong message to programs that need to detect if they're running in a job environment or nots, since they refer to the Flux broker's job environment, not the current process. Clear the job-related environment variables in the the broker rc script environment, where the inital program is considered "rc2". Drop unsetenv() calls for FLUX_JOB_ID, FLUX_JOB_SIZE, FLUX_JOB_NNODES in the broker mainline, now redundant. Fixes flux-framework#4716
garlick
added a commit
to garlick/flux-core
that referenced
this issue
Oct 25, 2022
Problem: FLUX_JOB_ID and other job environment variables remain set in the initial program if Flux was launched by Flux. This sends the wrong message to programs that need to detect if they're running in a job environment or nots, since they refer to the Flux broker's job environment, not the current process. Clear the job-related environment variables in the the broker rc script environment, where the inital program is considered "rc2". Drop unsetenv() calls for FLUX_JOB_ID, FLUX_JOB_SIZE, FLUX_JOB_NNODES in the broker mainline, now redundant. Fixes flux-framework#4716
garlick
added a commit
to garlick/flux-core
that referenced
this issue
Oct 25, 2022
Problem: FLUX_JOB_ID and other job environment variables remain set in the initial program if Flux was launched by Flux. This sends the wrong message to programs that need to detect if they're running in a job environment or not, since these variables refer to the Flux broker's job environment, not that of the current process. Clear the job-related environment variables in the the broker rc script environment, where the inital program is considered "rc2". Drop unsetenv() calls for FLUX_JOB_ID, FLUX_JOB_SIZE, FLUX_JOB_NNODES in the broker mainline, now redundant. Fixes flux-framework#4716
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Problem: the broker allows its FLUX_JOB_ID to leak through to the initial program, which makes it hard to distinguish whether a program is being run from a batch script or from a job. It looks like a job, but the job is really in the enclosing instance not the instance that started the initial program.
FLUX_JOB_ID confuses our OpenMPI schizo plugin, preventing singleton MPI jobs from being runnable from the batch script. See #4671 (comment)
It also confuses libyogrt as mentioned in #3817
The text was updated successfully, but these errors were encountered: