Timeout for single request breaks the whole page

[mnonnenmacher]

Loading the statistics for a very large organization can take some time:

When the request (api/v1/organizations/1/statistics/runs) times out with 504 (gateway timeout), it is reattempted a few times. When it timed out for four times, the whole page goes into an error state:

Instead, only the affected parts of the page should show an error, because a problem with loading the statistics should not completely break the frontend.

[sschuberth]

@lamppu do we already have dedicated endpoints for the organization level? If not yet, is there a change that these would make the timeout not happen in the first place?

[lamppu]

@lamppu do we already have dedicated endpoints for the organization level? If not yet, is there a change that these would make the timeout not happen in the first place?

Yes, it already uses a dedicated endpoint. But perhaps that endpoint should be split into multiple endpoints. But in any case, an error should be caught and handled in the frontend too.

[mnonnenmacher]

@lamppu do we already have dedicated endpoints for the organization level? If not yet, is there a change that these would make the timeout not happen in the first place?

This happens for an organization with more than one thousand repositories and a firewall that enforces a timeout after 30 seconds, I think fixing this would require to not calculate the statistics on the fly which is a bigger effort. But even then the UI should handle an error of the endpoint more gracefully.

[sschuberth]

an organization with more than one thousand repositories

Ouch 😮

[Etsija]

Maybe CatchBoundary component could help here.

[Etsija]

TanStack Query has external error handling, and there's also an example of error handling here, which could be tried.