Skip to content

known issues with flux #63

@lee218llnl

Description

@lee218llnl

from #51
Known issues to resolve:

test.attach_4_shutdownbe doesn't completely work. We use tools/flux/flux-helper.sh (bash script) to bulk launch the tool daemons and we need to kill this process sufficiently good enough (a.k.a. kill support) so that flux exec used in the script gets killed as well. rm_flux.conf is configured to send two consecutive SIGINTs to this process but bash wouldn't be killed. Here, the way the signal is sent which is essentially equivalent to kill -INT but this isn't good enough. If this were kill -INT -PID (note minus PID), I think this would kill both flux-helper.sh and flux exec processes sufficiently. But this can have other side effects since this will send SIGINT to the group that share the terminal. So I decide to revisit this later as part of hardening. The way we do bulk launch with Flux may change as well. Please see flux-framework/flux-core#3110 (comment).

No attempt was made to add support launch mode including porting test codes.

Had mild levels of testing to make sure this doesn't break other resource managers like SLURM. But we need whole lot more. (Really need CI).

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions