-
Notifications
You must be signed in to change notification settings - Fork 13
Description
We have deployed the dashboard server internally, and folks are very happy with the initial results. However, we have run into an issue when uploading the outputs.
We are using a custom database implementation and are feeding A LOT of data into it. It takes a little bit to query all the results needed to generate the dashboard. Granted, our implementation could use some optimization.
When uploading a result via the API, the call is blocked until the result is parsed, the data inserted, a get all query is executed, and the new dashboard is generated. This is leading to very long apparent upload times, when in reality it's blocked on the get all query. This is compounded by multiple outputs being uploaded in series by our tooling.
Our workaround for now has been to implement an in-memory cache of the data (appended on each upload), but this is less than ideal.
It would be ideal if the API call only blocks on the parsing and insert of the new data, so that it could return a success or not. Then, perhaps a background task is triggered which will update the dashboard some configurable timeout (such that with x seconds of no uploads, the new dashboard is created, where x >= 0).
What do you think of this improvement?
Unfortunately, we are not authorized to make external contributions, but this may change in the future.