Christopher Schultz wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Rainer,
So, I dumped my configuration entirely and tried to get something served
from somewhere other than the default virtual host.
It turns out that the default virtual host does not set a ServerName,
which apparently really confuses Apache httpd 2.2. Either this was not a
problem for Apache httpd 2.0, or I configured it so long ago that I
don't remember doing it.
Everything started working again when I simply set the ServerName in the
default virtual host to:
ServerName my.ip.address:80
What a PITA.
Good to know.
I still have a few questions, though:
1. I still get the JkShmFile error (didn't specify such a directive;
should I? If I don't need it, can I disable the feature to avoid
the error message?)
I'm afraid we didn't prepare for that case :(
See my other mai, why it might be interesting even without real load
balancing to use the shared memory. If you don't like it, I'm afraid at
the moment you will have to live with the sartup error.
2. Is is okay to have connection_pool_size=1 for my workers? The docs
suggest to me that it /must be/ set to 1 if I'm using prefork MPM.
OK, I see, that our wording is still not clear. What we try to tell the
users:
- this attribute is automatically configured for Apache httpd and all
MPMs. The default is the number of threads per process, so that each
request handling thread should be able to use a backend connection
whenever it needs one.
- for other web servers the default is either 1 or 250 (IIS)
- You might choose non-default values. Using bigger values than the
default for Apache httpd doesn't make sense, it will simply waste
ressources. Especially it does not make sense to use a bigger value than
the default "1" for httpd 1.3 and 2.x/prefork MPM.
If you find a better wording for the workers.properties docs page than
the existing, please privide us with it :)
3. How much of this crap can I get rid of:
worker.worker3.type=ajp13
Default, but I would stick with it to make it more readable.
worker.worker3.host=localhost
Default, but again I would keep it, because it's not unlikely, that you
will move the backend away one day.
worker.worker3.port=8385
Needed.
worker.worker3.lbfactor=50
Only needed if used in combination with an lb worker *and* you want to
use different values for the members of the lb.
worker.worker3.connection_pool_size=1
Default and better to drop the explicit configuration. Once you switch
to a multi-threaded MPM, this value (1) is most likely not what you
want. The default gets automatically adapted to the value of threads per
process.
worker.worker3.connection_pool_timeout=60
OK. Maybe a little short. 600 (10 minutes) should be OK as well. It
would be good to adjust connectionTimeout in the Tomcat AJP connector to
the same value (multiplied by 1000, because it's milliseconds there). We
proudly present the new docs page
http://tomcat.apache.org/connectors-doc/generic_howto/timeouts.html.
worker.worker3.socket_keepalive=1
OK, in case there is a firewall between your httpd and the backend. As
long as the backend is localhost, it's unlikely you need it.
worker.worker3.socket_timeout=300
Difficult to judge. Socket timeouts help in some situations and make
things more difficult to understand in oter situations. You could keep it.
Clearly, I can eliminate the lb_factor as I am not using load balancing
of any kind. The pool and socket config are just grandfathered-in, so
I'd appreciate any thoughts you have on that, Rainer.
Once again, thanks for taking the time to read through my posts. You
always seem to be willing to look at anyone's debugging output to help
prove that mod_jk isn't the problem ;) Vielen dank!
Gern geschehen.
Unfortunately somtimes still mod_jk is the problem. But it looks like
such cases are few these days.
- -chris
Regards,
Rainer
---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
