[Seaside] Server Identification in Session Keys

[Sven Van Caekenberghe]

Thanks for the quick response, Lukas!

On 17 Aug 2010, at 17:32, Lukas Renggli wrote:

> On 17 August 2010 17:28, Sven Van Caekenberghe <sven at beta9.be> wrote:
>> Hi,
>> 
>> I am trying to configure Apache 2's mod_proxy_balancer with sticky sessions towards multiple Seaside 3 instances.
>> 
>> What I need, apart from starting multiple VM's with the same image running on different ports (I am working on one machine), is that each VM adds a server identification to the session key (_s), so that the balancer can route everything in the same session to the same instance.
>> 
>> How can this be done ?
> 
> The simplest thing is to use the port number as the identifier.
> 
>> As far as I understand, it is WAApplication (WARegistry>>#handleFiltered:) that hands out session ID's in the _s parameter (#keyField), but I cannot immediately find where the session ID's are generated (is it WACache>>#store using WAKeyGenerator ?) and how I could interfere in a clean way.
>> 
>> Has anyone done this before ?
> 
> Sure, that's standard practice. Override #updateUrl: either in your
> custom session class or in the root component.

Well, I got this now:

updateUrl: anUrl
	| port |
	super updateUrl: anUrl.
	port := self requestContext request host copyAfterLast: $:.
	anUrl addField: '_x' value: self session key, '.', port	

Is that a good way to get at the port ? I guess, when the port is not 80 and HTTP 1.1 is used.
Where does Seaside itself figure out the port when it is not hardcoded ?

I have to find out if Apache mod_proxy_balancer needs only a host id, of a full session id plus host id like I did above.
In the latter case, the overlap with _s is a bit much, I'll see.

Thx,

Sven