Redirect spamming on backend failure
Open
- Dominant language
- TypeScript
- Stars
- 329
- Forks
- 67
- PR merge metrics
- No merged PRs in 30d
Description
If you've got Lab behind Hub, and a Lab page loaded, and the user Lab behind it goes away, the dask dashboard poller keeps polling at high frequency, but just gets redirected (by the hub) to the Hub login. Is there any way we could detect this condition (the problem being, I think we'd have to specifically parse the 302 redirect content looking for /login to determine when it's happened?) and back off on the refresh rate?
It certainly fills the Hub logs with spam, and I think it may be part of what's to blame for 504s happening at the nginx-ingress controller layer, if, as I suspect, the Hub's TCP accept() queues are filling up.
Contributor guide
Assessment
This issue has not been assessed yet.