Scott, thanks for your quick response. 

> On 2/13/13 9:31 AM, Olaf Krische wrote:
> > Or in short:
> >
> > Whenever i disable or even stop a resin in a cluster, then i see a lot
> of 503 in apache's access.log. This should not happen. Why is that so? Why
> the other servers do not take over?
> For stopping, is this transient or permanent?


And i know why. Because when i stop app-1, mod_caucho also stops to deliver to 
app-0, even caucho-status shows green for it. Also the other way around, if i 
stop app-0, it also stops delivering to app-1.

Interesting is also, what caucho-status shows:

 Default Virtual Host
 Error: host unavailable/busy

 Source: Cache File (/tmp/, Thu Jan 1 01:00:00 1970 )
 Last-Update: Thu Jan 1 01:00:00 1970 

And shortly later without "host unavailable/busy":

 Default Virtual Host

 Source: Cache File (/tmp/, Thu Jan 1 01:00:00 1970 )
 Last-Update: Thu Jan 1 01:00:00 1970 

And then again later:

 Default Virtual Host

 Source: Resin-ETag (Wed Feb 13 19:26:21 2013 )
 Last-Update: Wed Feb 13 19:26:21 2013 

And hopping between all that.

> >
> > And second: when i disable (not stop) a resin in a cluster, why
> caucho-status still shows it in green as "ok"?
> Because mod_caucho can still connect to the backend Resin. When Resin is 
> disabled, it's still accepting connections; it's just returning 503s.


It would be totally awesome to have a way to take the server out of load 
balancing for mod_caucho, without having to stop it. The server should be happy 
and alive for whatever, maintenaince, calming down, resin-admin and so on.

In the old days, there i would just remove the server from apache's local 
resin.conf and mod_caucho was happy with it.

And Resin as Loadbalancer: this would open more issues for me right now and is 
a tough decision: perl, mod_proxy, lots of rewrite rules to convert. 

First i need to sell resin4 anyways. :)

resin-interest mailing list

Reply via email to