[4.0.0] socket hang up

[abraxxa]

Since the move to node.js with rc1 we get the following error:

{"@timestamp":"2015-02-20T10:11:45.695Z","level":"error","message":"socket hang up","node_env":"production","error":{"message":"socket hang up","name":"Error","stack":"Error: socket hang up\n at createHangUpError (http.js:1477:15)\n at Socket.socketOnEnd as onend\n at Socket.g (events.js:180:16)\n at Socket.emit (events.js:117:20)\n at _stream_readable.js:944:16\n at process._tickCallback (node.js:442:13)\n","code":"ECONNRESET"}}

Kibana is started as a regular user on a Debian 7 64bit box behind a nginx reverse proxy acting as ssl server:

server {
    listen   443 ssl spdy;
    server_name  kibana.our.tld;

    error_log    /var/log/nginx/kibana.our.tld.ssl.error.log error;
    access_log   /var/log/nginx/kibana.our.tld.ssl.access.log;

    ssl_certificate         ssl/kibana.our.tld.pem;
    ssl_certificate_key     ssl/kibana.our.tld.key;

    include includes/ssl.conf;

    root    /home/nac/kibana-4.0.0-linux-x64;
    location / {
        proxy_pass http://localhost:5601;
        #proxy_set_header X-Requested-With XMLHttpRequest;
        #chunked_transfer_encoding on;
        #proxy_read_timeout 90;
        #proxy_http_version 1.1;
    }
}

[rashidkpc]

Does this issue occur without the proxy in place? The proxy really isn't necessary with Kibana 4, as it supports SSL out of the box.

[abraxxa]

I discussed this already on IRC: we have multiple different apps running on the same up address so nginx is required.
The ciphers supported by nginx are stronger and configurable and I'm sure the tls performance is better as well.
How can I debug this further? Would a tcpdump of the loopback interface help?

[rashidkpc]

I understand now, but does this occur without the proxy?

[rashidkpc]

Closing due to lack of feedback

[abraxxa]

Just because I'm busy doesn't mean that the problem doesn't exist.
I'm not able to access any but the already used ports on the production server and need some time to replicate the setup on a dev box.

[abraxxa]

I've deleted the .kibana index completely and started from scratch with my visualizations and dashboard which fixed the issue. It seems something from beta 3 caused this issue although I tried to redo the dashboard and its visualizations before.

[abraxxa]

Too soon ;(
The issue occurs in Firefox but doesn't in Chromium which is used to redo everything because of its vast better performance.

[abraxxa]

I got a separate IP address for Kibana on the prod server and configured it to listen on 443 on this port.
So far everything works.
I'd prefer to have it behind nginx again mainly to have the TLS settings in a single place.

[stackoverflower]

I am getting exact error
*Courier Fetch Error: unhandled courier request error: socket hang up
I am on Version: 4.2.2, Build: 9177. *
I get this error only when I use Proxy (IIS) in front of Kibana which I need to restrict access to Kibana. I am not sure what is causing this or how to fix it.

Error: unhandled courier request error: socket hang up
at handleError (http://kibana-server/bundles/kibana.bundle.js:70047:23)
at DocRequest.AbstractReqProvider.AbstractReq.handleFailure (http://kibana-server/bundles/kibana.bundle.js:69967:15)
at http://kibana-server/bundles/kibana.bundle.js:69861:18
at Array.forEach (native)
at http://kibana-server/bundles/kibana.bundle.js:69859:19
at wrappedErrback (http://kibana-server/bundles/commons.bundle.js:39286:79)
at http://kibana-server/bundles/commons.bundle.js:39419:77
at Scope.$eval (http://kibana-server/bundles/commons.bundle.js:40406:29)
at Scope.$digest (http://kibana-server/bundles/commons.bundle.js:40218:32)
at Scope.$apply (http://kibana-server/bundles/commons.bundle.js:40510:25)

Since this issue has been marked as Closed, posting this as Issue#6443

[jpgil]

Try removing http.cors and http.compression as noted in issue 6719