Cups scaling

[Anonymous]

> > - printer configuration: syncing the needed config files?  Or
> > configuring the printers on both nodes?
>
> Right now you have to manually manage this - it is usually just a
> matter of mirroring /etc/cups and restarting the slave nodes.
Ok.

> > - print jobs: is it enough to mount the directory that holds the
> > print jobs on the second node?  Will the second node pick up the jobs
> > and reprint / resume them?
>
> No, that's not how CUPS works.  I recommend reading the CUPS overview
> document, but basically CUPS failsafe/load-balancing is provided
> through implicit classes created by the clients - the clients hold
> on to the print jobs until a server says they have been printed
> successfully, and the clients determine which server in the class
> to print to.
I did, but I have different situtation.  I have 1 server and that's the cups server + the place where print jobs are generated.  So I can not use implicit classes because I have only 1 cups server.

I have 2 options:
- like described above: 1 cups server and try to make sure it can faileover to the other node (and if it's possible don't loose the jobs in the queue)
- Use 2 other servers as cups server and use implicit classes on the client.
Just wondering, what if you have an implicit classes with 2 remote cups server and you send a job to server A.  The job is queued because the printer is busy.  Server A crashes.  Does the client know that his job will never be printed?  Will it resend the job to the other server?

> > To maintain the 600 printers, is it possible to store the
> > configuration in a database / flat file?  So with 1 click on the
> > buttom, the printers are generated.  I probably can script this
> > myself, but maybe there is already a solution.
>
> We have no built-in database support, but that is an oft-requested
> feature.  We'll likely add support for this in CUPS 1.3, but don't
> expect that any time soon (our focus is on getting 1.2 out now)
Ok, good to know.


Stef