Hey, A Feature-Request for the `MF dashboard-UI` ...
# dev-metaflow
g
Hey, A Feature-Request for the
MF dashboard-UI
(is this the correct channel?) : Add another column for the
current_step
of each run, and allow also group-by by that or at last filter. If you push tons of jobs at once it can be much nicer to see it in step-aggregated-view. (i assume it might also be noisy to refresh, if steps are quickly changing. but this can be a sticky and changeable only at refresh) Also, i’m assuming you’re aware of a glitch that running-jobs appear as
Failed
when load is high with lots of jobs running (probably it’s just that time passed since job reported back). hope i’m not repeating past-msgs i had missed, or that we’re not using an older-UI version that has already added/fixed this (if so, feel free to delete this msg)
1
b
Hi @great-intern-74117. Thanks for the feedback. The latest version of MFGUI is v1.1.4 if you don't already have it. Can you sketch out how you think the
current_step
request would look? You can add more information by creating a feature request here https://github.com/Netflix/metaflow-ui/issues/new?assignees=&labels=&template=feature_request.md
The issue with running-jobs showing as
Failed
is probably known. It happens when the heartbeat is not received within a timeout. Do you see those tasks succeeding later?
g
Yes. sometimes we are spraying hundreds of jobs (that compete over resources in maximal capacity) and they all end after a few hours (sometimes even days) - they do produce logs and moving-bewteen-steps along the run (some steps are quite chubby though). They switch between as Failed and Running and end up as Completed. As for the current-step - i’ll think of something and submit 👍
b
@great-intern-74117 I have posted couple of mocks in the issue for clarification
👍 1