Performance: Saving changes to a learning resource can take 15-20 seconds

[pdpinch]

On lore-ci right now, it's taking me 15-20 seconds to save changes in the learning resource panel.

There's also no user feedback while the save is in progress, so it's tempting to click the save button multiple times, probably leading to issues like #517

[carsongee]

So looking at this just with a browser, I see some crazy stuff happening on the panel:

It looks like a lot of the performance may be the javascript paging through the entire static asset set before rendering that panel. In the case of this screenshot, that appears to have caused a 503 when trying to save the resource, and a second attempt resulted in a successful save.

Looking at the SQL DB, the disk didn't even get hit, and we only used 25% of available DB resources:

The memory on the dyno did double up to 750MiB which is above our 512 max and may be why the 503 came in (from the Heroku LB).

I don't have ES perf stats, but since the assets are pure DB I don't think it is involved.

[carsongee]

So, looking at that screenshot, I think the bigger issue is that we are patching with 4.6 MiB of data (which is the size of the XML for that LR), instead of just sending the description and terms. I think we just need to strip out that item in the PATCH.

[pdpinch]

Three things to fix

• Decrease the number of workers
• POST should be smaller. It might include XML, which needn't be posted.
  2
• Performance: Static Assets tab slows loading of learning resource panel #536 Listing static assets makes multiple calls, which uses too many workers. Suggest going with an "infinite scroll" approach. Or if we can't find a react infinite scroll, could be a "more" button.
  3

[noisecapella]

This isn't fully solved, I can still reliably take 4 seconds to save a learning resource on my local machine. I investigated and the underlying code updates the entire row in the database, not just the terms and description, the bulk of which is the content xml. One solution is to implement a custom update method on LearningResourceSerializer which does a more focused update.