feat(ui): Split fetching of releases into two network requests #18795
Conversation
| [ | ||
| 'releasesWithHealth', | ||
| `/organizations/${organization.slug}/releases/`, | ||
| {query: {...query, health: 1}}, |
There was a problem hiding this comment.
Have we considered having another endpoint that fetches only the health data without having to redo the work around getting a list of releases? Perhaps a bulk GET endpoint that takes a list of versions and returns the health data?
There was a problem hiding this comment.
Yeah, and this might be something we want to try later. Generally it's a lot more work because it's conditional not just on the version but also each individual project and environment selection and we would need to pass a lot of data as GET parameters which might be way beyond the max payload size for URLs.
| @@ -333,4 +346,11 @@ const ChartWrapper = styled('div')` | |||
| } | |||
| `; | |||
|
|
|||
| const StyledPlaceholder = styled(Placeholder)` | |||
| height: 20px; | |||
There was a problem hiding this comment.
Do you need this height if each component also sets the height to 25px?
There was a problem hiding this comment.
I removed the height props (it was a leftover from the previous iteration)
|
@markstory one ironic thing about your proposal is that it turns out to be slower by my quick testing. Since we now fetch both requests at the same time (with and without stats data) the stats come in quicker than if you wait for the first request to finish to then load the second one. I did a quick experiment and the current — kinda hacky — version is likely to outperform the other one in most cases at the expense of transferring more data and keeping the server busier. |
|
@mitsuhiko I didn't consider that we'd have to make 2 requests in serial. Having to do that would make the overall flow slower. If the extra work we're putting on the server becomes a problem we can revisit the problem then 😄 |
Releases endpoint can be called with
health: 1parameter which will return releases with their health data.This request can take significantly longer (especially for big orgs) compared to
health: 0.We decided to call
health: 0on the releases page and then fetchhealth: 1lazily, showing placeholders in the meantime.We will skip the two-phased load for sorting other than date created (default) since in those cases we already need to go to the session store anyways (which takes here the long time).
#sync-getsentry