On Jan 8, 2014, at 1:40 AM, Rakesh G K <[email protected]> wrote:

> Thanks Leif. 
> But what i was trying to convey was a bit different.
> I have 'n' ats nodes, 'm' tomcat instances and 'p' services offered by each 
> of the tomcat servers. So instead of having 'n' X 'm' X 'p' different entries 
> in the remap.config, I was wondering if it is possible to have just 'p' 
> entries for each of the services, by somehow having a single identifier for 
> the ats and a single identifier for any of the tomcat.
> 
> Is this realizable? Or should all of these entries be mentioned?.

Would the regex_map mapping type work for you?

https://trafficserver.readthedocs.org/en/latest/reference/configuration/remap.config.en.html#regular-expression-regex-remap-support

regex_map http://www.ats[0-9]+.com/origin http://www.example.com.origin

> 
> -Rakesh
> 
> 
> On Tue, Jan 7, 2014 at 10:40 PM, Leif Hedstrom <[email protected]> wrote:
> 
> On Jan 7, 2014, at 5:10 AM, Rakesh G K <[email protected]> wrote:
> 
>> Hello,
>> 
>> When in a clustered mode deployment, I need to be able to map certain 
>> requests that an ats node receives to the origin server. 
>> When I have a load balancer fronting the cluster, there is no way of knowing 
>> which server which get what request. So assuming that there are 3 ats nodes, 
>> and a origin server URL, say http://www.example.com/origin, is there a 
>> better way of remapping this than what is written below?
>> 
>> map http://www.ats1.com/origin http://www.example.com/origin
>> map http://www.ats2.com/origin http://www.example.com/origin
>> map http://www.ats3.com/origin http://www.example.com/origin
>> 
>> I
> 
> 
> This sounds like a case where you want to keep pristine Host headers across 
> the board. That would imply making sure your origins can handle those 
> “client” submitted Host: header, but would allow for a single, unified remap 
> config across the board. Just make sure you have
> 
>       CONFIG proxy.config.url_remap.pristine_host_hdr INT 1
> 
> 
> That is the default as well. Now, I’m not a clustering expert, so maybe 
> cluster doesn’t honor this. If it doesn’t, that smells like a bug to me.
> 
> — Leif
> 
> 

Reply via email to