On 2/13/13 8:11 AM, Olaf Krische wrote:
> Hello again,
> app-0$ ./resinctl disable -server app-0
This is a case where Resin's load balancer is more capable than
mod_caucho (we have some open bugs to upgrade it.)
mod_caucho primarily handles connection failures. It's not aware of 503's.
> app-0$ ./resinctl web-app-stop myapp
Currently, neither load balancer is precise enough to handle per-web-app
disables (again, an open bug, although it might require Resin 7 for that
enhancement because we're revamping the load balancing.)
> app-0$ ./resinctl stop
This should work with mod_caucho.
> Finally a change:
> - apache's /caucho-status: app-0 is getting red. Juhu.
> But still:
> - apache's access.log is full with "503" (why oh why its not going to app-1,
> - app-1 almost gets almost gets no traffic.
We'll need to check on this. With the backend stopped, mod_caucho should
failover to the existing servers.
If you're using the pre-fork model, though, it may take a number of
requests before the Apache processes become aware of the closed server,
because each process is independent.
> app-0$ ./resinctl start
> - app-0/1 slowly start get request again according to resin's access.log
> - in between i even start to see "404" in apache's log. Hm, this could be my
> fault, though resin should not start to deliver, when the webapp is not fully
> In any way, i am pretty lost.
> The only properties i have overwritten:
> dev_mode : false
> app_servers : 172.17.32.4:6800 172.17.32.5:6800
> session_store : false
> remote_admin_enable : true
> web_admin_enable : true
> web_admin_external : true
> rest_admin_enable : true
> admin_user : admin
> admin_password : ...
> And i have put those two ResinConfigServer ("config triad") into the
> ResinConfigServer 172.17.32.4 6800
> ResinConfigServer 172.17.32.5 6800
> Is anyone using mod_caucho in production and has no problems with it? Am i
> missing something here? :-)
> resin-interest mailing list
resin-interest mailing list