[cups.bugs] Cups server with _lots_ of printer definitions

John E. Vincent cups-bugs at lusis.org
Fri Jan 7 18:46:09 PST 2005


Our installation is currently at almost 600 printers and that will 
double in the next 4 months. We're doing okay right now but management 
is a serious headache of all those printers.

We load balance our two CUPS servers behind a hardware load balancer. 
Originally, we were going to use a third cups server with no local 
printers to act as our load balancer but that died as we added more 
printers. The printers were being purged and readded too often for any 
jobs to actually go through. The polling mechanism doesn't really scale 
to that high of a volume right now.

My suggestion if you want to go to CUPS with that many servers is to 
implement something similar to what we're planning:

A hardware load balancer that can route based on URI. In our case we 
have three types of printers per store - check, contract and receipt. 
This will be scaling to over 1000 retail locations in the next few 
months. Our contract and receipt printers are currently pointing to same 
physical printer. We're probably going to end up breaking each CUPS 
server into dedicated printer types and then directing based on printer 
name to the appropriate server. You can still keep the same single 
hostname and minimize requirements for client reconfiguration.

Be aware that the CUPS lpd support is not the best. I'm not sure how you 
plan on using these printers (samba?) or what not but you may run into 
an issue when printing from older lpd clients with port allocation and 
such for capacity.

FYI, our CUPS servers are dual P4 2.4Ghz Xeons and 2GB of RAM. They each 
run Gentoo.

Patrick Spinler wrote:
> Hey:
> I'm testing cups for our internal use.  As one test, I loaded our cups 
> test server platform (a 2cpu 3Ghz P4 with 4gb of ram running redhat 
> enterprise 3) with the printcap from one of our solaris LPRng machines.
> That's about 5300 printers, btw.
> That is, for each entry, I did an "lpadmin -p ..." command, from a script.
> I then stopped and restarted the cups daemon.
> Now, I attempt a "lpstat -a" command, and the cups daemon process pegs 
> the cpu, and appears to hang for several minutes while it processes. 
> 4:59, on my system.
> It's even worse if I try to "manage printers" in the web interface, 
> after waiting 15 minutes, I stopped my browser, and had to kill the 
> 'printers.cgi' process by hand on the server.
> How large an installation is practical ?  What, if any, tuning can I do ?
> Thanks,
> -- Pat
> _______________________________________________
> cups-bugs mailing list
> cups-bugs at easysw.com
> http://lists.easysw.com/mailman/listinfo/cups-bugs

More information about the cups mailing list