Sort-term strategy to improve the launch success rate

[arnim]

Hi 👋

Background: Just had a sort discussion with Markus Stocker from the ORKG team. They are doing (meta-) data management for various research output artefacts and are evaluating ways to make some of them online reproducible & executable. For this they have a binder specific example. https://orkg.org/comparison/R44930/ > "Visualizations with Jupyter"

In testing their example we produced many launch failures before successfully launching it. The example redirects to OVH.

In general OVH and GESIS are currently configured with the same weights of 100. I would naively assume that everything else being equal we would at average see similar numbers for launches / running pods. OVH, however, reports most of the time fare fewer running pods than GESIS. Is it possible that the difference are launch failures?

Would it help if we try to reduce slightly the relative weight of OVH? OVH is currently configured with a weight of 100.

What do you think?

CC @rgaiacs

[manics]

Changing the weight to bias launches towards Gesis sounds fine to me! We need to spend some time figuring out where the problems are occurring on OVH

[arnim]

Jup, Sort-term strategy only :)

I would also ask @rgaiacs if he feels comfortable with it.

[rgaiacs]

@rgaiacs works for me.

[arnim]

Should I open a pull request to set the OVH2 weight to 60, so we can test if this approach is effective?

[manics]

Sounds fine to me!

[rgaiacs]

This has been addressed in arnim@6433003.