saving image and deployment features

[Lukas Renggli]

Hi Martial,

> I am real interested in seaside/magritte/pier and as a newbie (in  
> squeak but  not in smalltalk) I would like to do a little Pier/ 
> seaside website. I chose to run my squeak process in a linux server  
> with a framebuffer X server to administrate my image with VNC  
> access (and not by the seaside inspector available with halos). I  
> did it with a special user (not root) and my firewall connection  
> configuration permits me to access externally to my webserver on  
> standard port 80.

welcome to Seaside/Magritte/Pier ;-)

So far your configuration looks pretty standard for such a setup.

> But I am a bit unsecure with a deployment of a "classical" pier
> configuration. If my machine crash, I lost all my data. So I need  
> to do
> backup. I can run a:
> SmalltalkImage current snapshot: true andQuit: false.
> but if a user does something on my website, it's rejected. So I read
> somewhere (maybe thru this mailing-list archive) about a way to do
> backup. Actually, I would like to save the image everytime it's
> possible. I heard about the use of:
> UnixProcess saveImageInBackgroundNicely but I don't know how to use  
> it.
> With forkHeadlessSqueakAndDoThenQuit message, it seems to be  
> possible to
> save my image in another until I do the permanent and tested weekly
> backup. So if I crash or there is a glitch, my image or the  
> periodically
> save image can be broken but I can restore the most recent backup.
> If someone can explain me how to do (or find how to do) this, I'll be
> glad. I don't want to build a wiki/blog like site and lost in a crash
> all the readers'comments/posts.

This is what is implemented in PRImagePersistency. If you have  
OSProcess loaded it will do it nicely in the background, else it will  
just do a snapshot and block the VM for a few seconds while saving. I  
am successfully using this method for all my Pier images. Have a look  
at the implementation, there are a lot of things that you can tweak  
and change there.

Keith Hodges did a persistency implementation using the object  
database Magma. This looks very promising and I think this will be  
the way to go in the future. I am sure he is willing to tell you more  
about it ...

There are a couple of other persistency strategies floating around,  
check out the mailing list archive.

> I also noticed some errors in the deployment of  
> squeakfoundation.org. By
> clicking on 'Edit' or 'Login', there's error. But instead of  
> permitting
> me to go back to the previous page with a short error message (and
> advertising the webadmin) like on the other classic php/jsp website, I
> go to debug message. It's really bad. I guess that when I click on
> 'Debug' I open a morph Debugger window on the squeak image. It's not
> what I (and probably the squeakfoundation.org webmaster) want. I am  
> sure
> there is a way to shortcut the doesNotUnderstand message in seaside/ 
> pier
> so that in a deployment mode, the user has just a friendly error  
> message
> and the webmaster received an email from squeak with the debug message
> infos and the context.

I am not responsible for this server, but I know the problem: the  
versions of Pier and Seaside do not match.

To answer your question: in the Seaside configuration interface there  
is a setting to choose among different error handlers or implement  
your very own one ...

Cheers,
Lukas

-- 
Lukas Renggli
http://www.lukas-renggli.ch