Skip to content

Conversation

@FiloSottile
Copy link
Owner

No description provided.

This is arguably not the right long-term solution, because 1) we should
probably only raise the timeout for the final tiles upload, where
failure is fatal, and lower it if anything for other uploads, and 2) a
longer sequencing will make the next pool bigger, which might be bad at
a time of increased backend latency. However, for now it should let us
observe the latency spikes better.
AWS might be throttling us, the simplest answer might be to let the SDK
backoff more (but not up to 20s like default, at least not for now).
@FiloSottile FiloSottile merged commit f95f6bb into main Apr 22, 2025
2 checks passed
@FiloSottile FiloSottile deleted the filippo/latency branch November 8, 2025 13:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants