@Reshiram110 was using the new rbxthumb API to display avatars in UI, specifically the following rbxthumb url:
rbxthumb://type=Avatar&id=73223877&w=720&h=720
which failed to load with the following error:
Image "rbxthumb://type=Avatar&id=73223877&w=720&h=720" failed to load : Request failed
I was able to reproduce this 20 minutes later, but it seems to vary by userId (times in PST):
| UserId | Attempt 1 (Resh - 00:03) | Attempt 2 (Me - 00:21) | Attempt 3 (Me - 00:23) |
|---|---|---|---|
| 73223876 | N/A | Success | N/A |
| 73223877 | Failure | Failure | Success |
| 73223878 | N/A | Failure | Failure |
| 73223879 | N/A | Success | N/A |
Upon inspecting the network requests in Fiddler, there weren’t any backend errors or image corruption. The backend was just returning “pending” for all four attempts to load the thumbnail. Each retry after the previous request returned pending were sent with no delay.
This seems like two separate issues.
Backend:
It seems like an oddity that thumbnails for 73223876 / 73223879 were generated instantly but not for 73223878 and took 20 minutes for 73223877 to get out of the pending state. There doesn’t seem to be any correlation between the accounts. 73223878 and 73223879 have almost the same appearance. Both are not very popular accounts and should be in relatively the same state WRT existing thumbnails generated
Client:
The error could probably be surfaced to users better. “Request failed” isn’t entirely accurate since the request successfully returned pending. “Thumbnail is still pending”/or an N/A image may be more descriptive of what’s going wrong. Also, if, oddities aside, thumbnail generation takes longer than the second or so it takes all four requests to complete to generate thumbnails, it may be worth having some backoff factor if “pending” keeps getting returned
