| |
|---|
Updates from MolSSI | BP – Mostly backend work. Nothing that should be too visible. BP – Need to cut a release of basis-set exchange. Looking to do that this week. BP – I DID see some cases where multiple managers pull down the same task. Will look at putting in a simple fix for that today.
|
Manager/Queue status | DD – Managers running on PRP, UCI HPC3, MSKCC Lilac. Doing protein fragment torsion drives. DD – A few datasets are almost ready for submission, most need my review. DD – I’m preparing a dataset for JM, will submit soon. It’s a reproduction of the biphenyl set. TG – No problems seen in UCI managers. DD – PRP set up horizontal pod autoscaler, so we may be able to self-regulate our numbers of managers. JW – Could be fun to post a weekly “leaderboard” of completed jobs of each type, separated by job type and cluster
|
Follow-ups from last time | |
User questions | JH – I have XTB set ready to go. Can our managers handle this? DD – Latest “kitchen sink” managers should be able to handle these. JH – Any blockers to merging this? Managers should be able to distinguish whether they can do these jobs. DD – We should be fine to merge this. It won’t block anything higher-priority. JH – PR needs review. It just adds compute, no dataset changes
TG – BP, when PB was looking at old data, we found that some torsiondrives were missing QC_VARS data field. I think it was due to psi4 or QCEngine version difference on managers. Would it be possible to wipe and re-run this job? BP – This is a hard thing for us to re-run. (General) – We can manually skip this in future dataset analysis. It’ll be really hard to change this because the records are pretty much immutable. This sort of problem should be far less frequent given infrastructure improvements. PB – JM was using QCA data to get WBOs, but now is using OpenEye to recompute them. So this particular problem may not affect us.
TG – Anticipated submission date for Sage datasets?
|