Dong H. Ahn
Dong H. Ahn
Per today's discussion (https://github.com/flux-framework/flux-core/issues/2608) , we will piggyback @chu11's job listing tool to present basic scheduler information such as estimated start-time and reason for pending etc. For other advanced resource...
https://github.com/flux-framework/flux-sched/pull/677#issuecomment-648994915 When the underlying graph hierarchy doesn't match with a jobspec, our graph infrastructure code returns an match error: ``` 2020-06-24T18:30:21.552622Z sched-fluxion-resource.err[0]: match_request_cb: match failed due to match error (id=0):...
This will be likely to be broken into multiple issues but I wanted to open this to remember this important items as we will work on "stabilization" tasks towards a...
Just a quick note, if a graph is such that you can get to a resource vertex through different paths, find breaks. `cluster0->rack0->storage_unit0` and `cluster0->gunit0->storage_unit0` ``` resource-query> set-status cluster0->rack0->storage_unit0 down...
Problem: Scoring API is designed for simple match policies and I find that it is pretty limiting to expose more sophisticated match policies. For example, this [line](https://github.com/flux-framework/flux-sched/blob/master/resource/policies/dfu_match_high_id_first.cpp#L70) implements our very...
From https://github.com/flux-framework/flux-sched/pull/693#discussion_r457793152, adding support for `expr_eval_err_t` or similar so that the upper layer can print more information about expression parsing error.
As part of PR #460, I generated other match formats. `simple`, `pretty_simple`, `jgf`, `pretty_jgf` all work okay. But somehow `rlite` wasn't working.
Our main traverser is depth first and up visit or DFU. This is a superset of depth first visit (DFV). It begins its DFV from the root of the selected...
This came out from a review on PR #614. The traverser currently uses `== -1` comparison to check an error condition for all of the writer emit methods. But this...