2022-03-15 QC Meeting notes

Updates from MolSSI

• BP – Since deleting the data before the last meeting, the server seems to be doing fine.

• BP – Making the last big push on the refactor. This is on reaction datasets, which I don’t think that OFF uses. Once I move over ReactionServices then I think that’s the last big change that I expect to make.

• BP – After that, I’ll make a migrated server available for testing. This will contain all of your data up to last september. While that happens, I’ll mostly be taking feedback and making smaller interface changes.

• BP – So, I expect to get to that at the end of this week. Then I could use a week for the smaller tasks. So we might be looking at the QCF release at the beginning of April.

• BP – I could either spin up two separate servers, or spin up the old and new archive simultaneously at one address.

  • JW – No strong preference

  • BP – I think I’ll do the “two physics server” approach

• JW: While migration is going on should we shut down compute on our end?

  • BP - I would make sure the database migration doesn’t happen on a live server

  • DD - We can sit together and plan for it, I would be happy to help

•  

Infrastructure needs/advances

• DD – I’m pushing a QCEngine release to support requested features in QCSubmit.

• PB – We still have to compute pubchem set 2 and 3, so even if the QCSubmit release is delayed, our compute will be saturated for a while.

• DD – The QCEngine and QCSubmit releases are already made,and I’ve updated the managers.

Throughput status

• OpenFF ESP Industry Benchmark Set v1.1: last 3 jobs struggling to get workers(?), added compute label

  • PB – I added the compute-openff tag to see if this would get picked up

  • DD – It’s possible that they’re already running, they’d still show as incomplete. Unfortunately It’s hard to track their status in a detailed way.

  • SB – Is there a way to know whether they’re really incomplete vs. taking a long time to run? I know sometimes jobs get desynchronized and the server and manager are not in agreement to their status.

  • BP – In rare cases, a job can be complete, but the server thinks it’s incomplete.

  • SB (later in meeting) – I just checked and the 3 remaining jobs are now complete. I’ll report back if there are any remaining issues.

• Openff dipeptides torsiondrives v2.1: 24/26 TD complete

  • deadslow - (CC: Out of two one may error out so if we reach 25/26 we can move this to end-of-life)

• OpenFF Protein Capped 1-mer Sidechains v1.0: 5/46 TD

  • 2500+ errors, should we be concerned?

  • DD, CC, and PB will meet later today to have a working session on this - Will try to understand how to keep this morning forward

• SPICE PubChem Set 1 Single Points Dataset v1.2: Errorcyling not working on this now (looks like running out of memory, kill code 137 which is for SIGKILL by OS unless manually stopped), last 16K or so calculations

  • DD – At this point, we may need to use AWS. JW, what’s the status of that?

    • JW – Spoke with Karmen, made progress but no username/password yet.

  • DD – Will error cycle pubchem set 1 using my personal computer, since it’ll be large enough.

• SPICE PubChem Set 2,3 Single Points Dataset v1.2: waiting for workers

•  

User questions/issues

Science support needs