[ERROR] Failed handshake: read tcp ...

[kingdonb]

Accepted connection.
---> [ERROR] Failed handshake: read tcp 10.244.0.91:2223->10.244.0.1:49157: read: connection reset by peer
---> [ERROR] Failed handshake: read tcp 10.244.0.91:2223->10.240.0.4:51527: read: connection reset by peer
Accepted connection.
---> [ERROR] Failed handshake: read tcp 10.244.0.91:2223->10.240.0.5:62983: read: connection reset by peer
Accepted connection.

In deis/deis#4431 (V1 PaaS) this message was supposedly demoted to a debug warning, but in the current release of builder it appears to be pretty clearly labeled as a scary [ERROR]

I had done a complete wipe of my deis namespace/helm release and rotate of all credentials and keys, up and running with a restored database from bucket storage, and an unrelated problem with DNS/hosts was preventing me from reaching builder. Well this error led me down a rabbit hole of trying to figure out whether there was a database record of the public key somewhere, and some controller component that needed to communicate with builder wasn't able.

It would be nice to clarify somehow in these messages that they are really just health checks hitting the builder pod. I'm not sure if this is a regression from V1 PaaS or if it looks exactly how it did after deis/deis#4431 had been merged, or what, but it seems like the behavior is in an undesirable state.