[mod_cluster-dev] MODCLUSTER-27

[jean-frederic clere]

Hi,

Just moving a private discussion to the list. For more comments.
The screenshot is at 
http://www.jboss.org/file-access/default/members/mod_cluster/freezone/images/mod_status.png

The actual idea is to add small links (like buttons) to do the operations:

Context:
enable/stop/disable on a context. (disable means no new session (a new 
session is request without sessionid) and enable means re-enable a 
stopped or disabled context.

Node:
start/stop/enable/disable on node.
That corresponds to the mod_jk concept (active/disable/stop). We can't 
start or stop a node from httpd because the node should be discovered 
automatically and administered in the cluster.
But in the mod_cluster logic a disabled node is just a node with all its 
contexts disabled and a stop node corresponds to all contexts stopped.
I would proposed links that do the following:
disable node = disable all contexts of node.
stop node = stop all contexts of node.
activate node = restore the state the cluster has sent to httpd

Of course all those actions are undone by restarting a node or the 
webapp in the cluster.


Bela Ban wrote:
 >
 >
 > jean-frederic clere wrote:
 >> Now I have a "nice" status page.
 >
 > Excellent, looks good !
 >
 >
 >> I am still thinking about the part:
 >> +++
 >> - allow a user to disable/enable a web app
 >> - start/stop/enable/disable a cluster node
 >> +++
 >> Will this bring more problems that help?
 >
 > OK, we should at least have the *same* functionality we have for mod-jk
 > and the /status/ app. We *can* enable/stop/disable workers with /status/.
 >
 > Let's leave enabling/disabling of webapps out for now.

In fact the logic in context = webapps and it is already there it just 
need a simple extension.

 >
 >> Because any change of topology in the cluster nodes will of course
 >> overwrite the user action. Are you sure we need something like that?
 >
 > I think that's fine.

Ok.

 >
 >
 >> For example:
 >> 1 - The user in httpd disable an application.
 >> 2 - The application is redeployed on that node: that means it is
 >> removed at some point in the logic and created again. When it is
 >> created again we have lost the information that it was disabled by the
 >> httpd user.
 >> 3 - The user will seen the application "magically" enable again.
 >
 > That's the behavior I'd expect

Me too.

 >
 >> That looks logic for me but I am not sure it won't give complains in
 >> future.
 >>
 >> Comments? - May I should just have "temporarily" in front of the
 >> disable/enable/start/stop (and start/stop needs another wording) -
 >
 > Can we discuss this on the mod-cluster mailing list ? I'm sure you'll
 > get more valuable input there...


More comments?

Cheers

Jean-Frederic