issues with Amazon load balancer

[michaelgursky]

(Just noting for the record, not necessarily requesting any changes.)

I was recently testing roxy with an AWS cluster set up using the ELB (elastic load balancer). I ran into a couple of issues:

1. Roxy sends a version name of "XDBC/1.0" in the HTTP request line. This causes the load balancer to reject the request with 400: Bad request. Changing this (in write_header in xcc.rb) to "HTTP/1.1" fixes the problem.
2. "deploy modules" with the batch setting on (the default), results in 504: Gateway timeout. The fix is to use "deploy modules --batch=false" so that each module is committed individually.

[michaelgursky]

After further testing, it seems that changing the version string to HTTP/1.1 is a bad idea for actual XDBC traffic. With this change, each module copied by "deploy modules" ends up with the file size as the first line of the file followed by the actual code.

[grtjn]

I think this requires tunneling. Not sure we got docs or examples on that, but have heard about people trying to do that..

[bakkenh]

In case someone runs across this on a google search, the best solution is covered here:

https://github.com/marklogic/roxy/wiki/Using-Roxy-Deployer-with-Amazon

I had this same issue- the fix was to change the listener to accept TCP instead of HTTP on the XCC port.

[dmcassel]

Thanks @bakkenh. If you've got more details that we can put on that wiki, I'd be happy to add them.

[grtjn]

Added details about 504 timeout to the wiki..