geneontology/go-site

In GO metadata, change all ORCIDs used in GO from HTTP to HTTPS

Closed this issue · 27 comments

kltm commented

Currently, there is a mix of HTTP and HTTPS used for the scheme in URIs for user ORCIDs. After some discussion we have decided to 1) convert all of our user metadata concerning ORCIDs over to HTTPS and 2) ensure at the schema test level that no further HTTP additions can be made.

(As part of this, there will be related tickets about upgrading ORCID information in the noctua-models.)

kltm commented

To make sure that everybody is aware, as there might be synchronization issues as we go through this (although it's unlikely to be much worse than it is now anyways) tagging @vanaukenk @ukemi @cmungall @tmushayahama @lpalbou @sierra-moxon @dustine32 .

kltm commented

And @balhoff (sorry!), I'm remembering that users.yaml and groups.yaml get ontologized; i.e.: https://github.com/geneontology/pipeline/blob/fb7da0ea388409e4cda2fd71aa90e0003a2ed840/Jenkinsfile#L794 . I'm assuming this gets absorbed into some larger GO ontology load at some point.

From my reading then, after we get this HTTPS update into go-site, we'll want to get a snapshot out and then proceed with geneontology/noctua-models#194, testing in dev and then to master.

@pgaudet I think we probably also should wait until this current release is done to just have less things moving around at the same time.

Should we go ahead with this now, since the release is done and hopefully there is a new one next week or so?

kltm commented

@pgaudet Good question. The last release was the 16th and we're at about a week away from the next release attempt. I think that if we assume that we'll be "late" or don't care about hitting the first attempt, we can go ahead and start give it a try.

kltm commented

@vanaukenk , thinking about how the upgraded data will roll around, if we want to do testing in noctua-dev, I think I would recommend either waiting to get started until after this Thursday's restart or getting started now and skipping this Thursday's restart.

Let me know when this passes through. I think it should be ok, but just to be on the safe side, I would like to test the GO-CAM API, Noctua Landing Page, ART and Noctua Form.

ukemi commented

Should MGI go ahead and change the upstream data?

kltm commented

For everybody following along, the data timeline would look like:

  • update go-site user.yaml metadata (from HTTP -> HTTPS for all ORCIDs)
  • allow this data to get out as part of a snapshot release (would appear in preassembled ontologies)
  • update noctua-models dev branch models to new orchids (and then restart, picking up the new ontologies)
  • update noctua-models master branch models to new orchids (and then restart, picking up the new ontologies)
kltm commented

@lpalbou Absolutely--sounds great. Which of the steps above are you most interested in?

kltm commented

@ukemi Probably not necessary, since we'll be doing everybody in bulk at the same time anyways.

ukemi commented

OK. So you are going to transform the upstream GPAD that @dustine32 imports from MGI?

kltm commented

@ukemi Sorry, I misinterpreted "upstream". As far as your MGI data goes, if you update it before we start, it will be out-of-sync until we make the changes. Practically speaking, since I believe the only place it expresses itself is in your export GPAD, and that is in production at no location, I believe it would be fine to do now. (Although I may have missed something--that's why this stuff is always so fun!)

@kltm I would vote for waiting on the https update until after this Thursday's Noctua outage, as some curators are waiting for an updated GO in Noctua and this https issue is important, but not as urgent as the ontology update IMHO.

ukemi commented

OK. We are forging ahead on the MGI end. I started a ticket. At this point the worst thing I see is that the curators' names won't resolve on the imports.

@lpalbou Absolutely--sounds great. Which of the steps above are you most interested in?

Steps 3 and 4:
update of dev models = test NLP, ART, NF
update of prod models = GO-CAM API & GO website

kltm commented

I was planning on starting this today, but would like to wait until we have a recent good snapshot run complete (we lost a couple of runs to worm anatomy ontology issues and then some upstream ftp issues) so we don't keep people waiting as we debug.

kltm commented

A release is freshly out and noctua just cycled; I'm going to go ahead and start on this #1708 (comment)

kltm commented

@vanaukenk Now merged into go-site master, we should see results after the next snapshot success #1708 (comment)

@kltm Just checking in with this one and confirming that the GO-CAM models metadata has not yet been updated on dev or production.

@hdrabkin asked about the display of curator orcids on the production Noctua landing page.

Any curators whose orcid was originally https has their name displayed, while those whose orcids were originally http currently only show the http orcid (i.e. the original entry).

image

I could send an email out explaining why this is, so no one worries about what happened to their orcid - name link.

kltm commented

@vanaukenk The current state is here: #1708 (comment)
While the new updated is merged into go-site master, are still waiting on a successful snapshot before it starts percolating out to more places.

kltm commented

@vanaukenk noctua-dev should (I believe) be ready to test for HTTP -> HTTPS orcid upgrade.

#1708 (comment)

@kltm
@ukemi made some changes to a model this morning and now both his name and http orcid are associated with the model and displaying on the Noctua landing page. He hadn't logged out since the update.
After logging out and back in, updates no longer associated the http orcid.

ukemi commented

Update. This problem seemed to go away after logging out and back in again. However, the two models that I modified before that still need to be fixed.

kltm commented

Okay, we'll need to coordinate a barista reset and an update for the models.
For the latter, we can possibly coordinate that with the update we'll need to be doing with the ZFIN models.

kltm commented

I believe this cleared w/o fanfare some time ago?
Please reopen if I'm mistaken.