trafficserver-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Leif Hedstrom <zw...@apache.org>
Subject Re: Newbie questions
Date Tue, 22 Feb 2011 23:08:09 GMT
On 02/22/2011 03:53 PM, Jorn Argelo wrote:
> Dear all,
>
> We are investigating a potential migration from Squid to ATS. On the 
> first look most of the stuff we use Squid for is available in ATS, but 
> I have a couple of questions about the remaps for a reversed proxy ATS:
>
> 1) The documentation states that the port in both URLs must be the 
> same. If I understand it well, this means I cannot connect to port 80 
> on ATS and have, for example, port 50888 listening on my origin server 
> and use ATS to connect to the origin server. Is this correct? If it's 
> not correct, how would I put it in the remap.config?

The documentation is wrong, blame Igor :). Seriously though, we should 
fix that doc bug, this has caused confusion before. You are most 
certainly able (and encouraged) to map from one port to the other. I do 
this all the time, on my "production" box, to hide Apache HTTPD from 
outside access. The map rule would look exactly as you expect, e.g.

     map http://www.example.com   http://localhost:8888



>
> 2) Is there any way of splitting the remap.config in multiple files? 
> This for manageability of remaps and redirects.

Not right now. At least not inside any of the "core" code. What I'd 
suggest (and has been done before) is to do this yourself, e.g. simple do

       cat /some/path/*.config > 
/usr/local/etc/trafficserver/remap.config && traffic_line -r


>
> 3) Squid's cache peer construction allows direct connection to an 
> origin server using the host header provided by the client. And, based 
> on the acl, you can let Squid decide to what cache peer it needs to 
> connect to based on the Host header. How does this work with ATS? I 
> understood from the documentation that the remap.config is also used 
> for reversed proxy mode, but I still need to provide a target 
> (internal) URL. In our current setup the internal and public URL are 
> the same so I don't think this will work then? Since we have nothing 
> to fill in as target URL.

Not sure I understand this. You will need to map something like

     map http://www.example.com/  http://localhost:8888


The default configuration as of current trunk (might be in v2.1.5 as 
well) is to preserve the Host: header as sent to the origin server. you 
can override that in records.config (look for the configuration for 
"pristine host header").

>
> 4) I am still not sure what the use of the reversed_map is and when I 
> am supposed to use it, even though I read the documentation several 
> times. Can someone please clarify this?

It's intended to be used when a server return redirects / location 
information back to the user, to make sure that the user goes through 
the proxy and not directly to the origin server. If you do not produce 
any such information (which would be the case if you have the backend 
respond / use the same URLs as the proxy, and use pristine host 
headers), then it's not an issue. If however the HTTPD daemon could e.g. 
return a header with a 302 response and Location: localhost:8888, then 
you will need the reverse_map to handle it).


Cheers,

-- leif


Mime
View raw message