search

GLYCAM-Web / website

A bare-bones repo to contain public website stuff and issues related to the GLYCAM Web Django apps.
4 stars 0 forks source link

Inconsistent behavior when the downloads page provides previously built structures, on actual. #171

Closed danwentworthart closed 6 months ago

danwentworthart commented 9 months ago

Tool: Carbohydrate Builder

Bug Description: Inconsistent behavior when the downloads page provides previously built structures, on actual.

To Reproduce:

  1. On Actual, (glycam.org, envy-eden) enter the following sequence into Build Via Txt: DXylpb1-2DManpa1-3[DXylpb1-2]DManp[6A]a1-3[DGlcpAb1-2]DManp[6A]a1-3[DXylpb1-2]DManpa1-3[DXylpb1-2]DManp[6A]a1-OH
  2. On the Options page, accept the default selected rotamers. It should have 8 structures requested by default.
  3. Observe the behavior. This has been built before, and all conformers should just load immediately. In fact, I see 5 of them do this correctly, but 3 consistently are queued for minimization, even though they have been built multiple times now.

Expected behavior Once conformers have been built, they should always be reused. This should lead to the following pattern: The second submission of any sequence/conformer should load almost instantaneously. Assuming you allow completion of the first request, the second one is supposed to just reuse that output.

Additional context I am still chasing logs to better understand the details here. I did fail to replicate in my dev env, but that isn't terribly revealing due to the number of variables involved when contrasting devEnv and Actual. If I find relevant issues, I will share them here.

danwentworthart commented 7 months ago

Thought maybe this had been fixed. Certainly has been improved. However, something is still needing checking on here.

In test.glycam.org, building the same structure the first time goes mostly as expected except:

1ogt_4ogt_6ogg gets stuck on minimizing for what seems like it might be forever. Requires more digging to see what is happening in production that doesn't occur in dev.

Second request loads all previous builds lightning fast, but 1ogt_4ogt_6ogg remains stuck.

danwentworthart commented 6 months ago

This issue was not present when I tested actual after it was updated to gratitude-giacometti, Thursday, March 21, 2024 12:04:26 PM,