I have Centos 6 machine that I use for a cups printer server. (Yes, I know, but it's working fine for what it's for at the moment.)
My ISP lost its connection to the network that printer server is on. While the outside Internet service was unavailable I sent a print job to the printer on that server and nothing printed even though the print server and the machine I sent the job from are on the same lan and are actually connected to the same router. The job appeared to be sent from the originating machine (but timed out, as shown below), and running lpq and lpq -a on the server showed "no entries" (because it timed out for some reason).
I did this a few times and even rebooted the server and always the same thing. Job sent, "no entries" on the server.
The my ISP's connection came back and blammo! All of a sudden I got a ton of sheets printed, everything that I had been trying to print for the previous ten minutes.
Here is what appears to be a relevant portion of the log on the machine that was sending the job to the print server:
Mar 30 13:05:06 mutt cupsd[1443]: Started backend /usr/lib/cups/backend/implicit class (PID 951459) Mar 30 13:05:06 mutt cupsd[1443]: REQUEST localhost - root "POST /admin/ HTTP/1. 1" 200 245 CUPS-Add-Modify-Printer successful-ok Mar 30 13:05:07 mutt cupsd[1443]: Job completed. Mar 30 13:06:02 mutt cupsd[1443]: Job submission timed out.
So what happened here? Why would cups care if it has a connection to the outside world when it's just printing a job sent to it from the local lan? The connection between the machine originating the print job and the print server timed out even though all of the relevant traffic was going over the lan.