Container v3.0.0-rc4 breaks TLS forwarding about traefik HOT 6 CLOSED

Hello @troykelly,

Thanks for your interest in Traefik!

The newly introduced lingeringTimeout value has a very narrowed value by default: https://doc.traefik.io/traefik/migration/v2/#transportrespondingtimeoutstcplingeringtimeout

In the case of "server first" protocol handling, increasing the value must solve the issue you are facing.

from traefik.

Thank you @rtribotte - has it not been implemented until RC 4?
If it was implemented in 2.3, wouldn't I have seen this issue a lot earlier?

from traefik.

I would also like to know if there was a change between rc3 and rc4 which is related to this setting: lingeringTimeout

from traefik.

Hello @troykelly,

Thanks for your interest in Traefik!

The newly introduced lingeringTimeout value has a very narrowed value by default: https://doc.traefik.io/traefik/migration/v2/#transportrespondingtimeoutstcplingeringtimeout

In the case of "server first" protocol handling, increasing the value must solve the issue you are facing.

Would you consider making the default timeout value something sane, or better yet reverting the change? This seems like a regression in that things were working in the previous release candidate.

from traefik.

My personal feelings aside, what do other similar proxys do? What's best practice?

It seems like this is a helpful feature, perhaps implemented in a way that's doing more damage than good.

• What was the "default" before the change?
• Is it unsafe to set this as zero?
• Perhaps we should add more to the documentation - ie for http, best practice is n seconds, for redis x seconds, postgres y and so on.

from traefik.

On behalf of the maintainer team, sorry for the inconvenience.
We opened a global issue to cover this bug: #10598
I'm closing this one, as we will update the global issue while we make progress on this.

from traefik.