This doesn't have to do anything with Docker and/or multi-stage builds. It has to do something with the way sniffing works (see details in the Wiki). The node IPs are probably not accessible from your application.

actually, my app can access the node I'm trying by exec in the container and curl to elastic directly

but when logging this error still happen

What i'm doing wrong cause it's work normal when i'm put build file on top image not second images

Okay. When you enable sniffing, it will retrieve the node IPs from the http.publish_address. And that seems to be routeable, you're right.

Where does that msg=health check timeout: ... come from? Is that a prefix you use, or does it come from Elastic? Because that might actually mean that sniffing works, but health checks don't.

Can you try to disable health checks temporarily to see if that makes a difference?

Any news here?

Oh, my bad sorry for no response. After your suggestion, I'm disabled health checks its work fine

@Navapon I have been trying to reproduce this with https://github.com/olivere/elastic-with-docker (now updated to work with ES 6.x), but I couldn't. Maybe you can try it with that repo, and I'll re-open here if you can get the bug there. Thank you.