This is especially a very bad design for desktop/home users. They often only turn on their printers when they really want to print (to save energy). Then it happens easily that they send a job and forget to turn on the printer. CUPS disables the queue and the annoying re-enabling procedure has to be done. And many users even do not know about the problem. So the printer does not print and the queue gets nuked and re-created. Or the people cry on the help forums or at the installation support of their distributions.

This problem can be easily worked around by installing beh, the Backend Error Handler. This script makes the handling of such backend errors configurable, so that the problem can easily be worked around. The new possibilities are:

• Let queues simply not being disabled. Simple approach, but job gets lost.
• Repeat a given number of times.
• Repeat infinitely often, until the job gets finally through. This is the standard of LPRng, and it eliminates loss of the job.
• The interval between two attempts to run the backend can also be configured.
• Configuration is done independently for each print queue. So local printers and network printers can be treated differently.

Simply download the script, copy it into the CUPS backend directory (normally /usr/lib/cups/backend/), and make it executable with "chmod 755 beh. Then restart CUPS that it registers the new backend, usually with "killall -HUP cupsd" or "/etc/init.d/cups restart". If you did all correctly, there must be a line containing "beh" in the output of "lpinfo -v".

beh is a wrapper which is called by CUPS in place of the usual backend, which is called by beh now. This way beh can, depending on its configuration, repeat the call of the backend or simply hide the error status of the backend from being seen by the CUPS daemon.

To make a print queue usinf beh, get root and call a command as follows:

lpadmin -p <queue name> -E -v beh:/<dd>/<att>/<delay>/<originaluri>

with

<queue name>:

The name of your print queue

<dd>:

Don't Disable, if "1", beh always exits with zero status, so the queue gets never disabled when the original backend exits with an error. "0" carries the error status of the last call of the backend (after <att> retries) on to CUPS, so the queue usually gets disabled.

<att>:

Attempts, number of attempts to recall the backend in case of an error. "0" means infinite retries. In this case <dd> gets meaningless.

<delay>:

Delay between two attempts to call the beckend, to be given in seconds and as an integer number. Meaningless if <att> is one.

<originaluri>:

The original URI, which your queue had before (use "lpstat -v" to get your queue's URI shown).

All parameters, especially, <dd>, <att>, and <delay> have always to be specified, even if one of them is meaningless due to the setting of the others.

beh works with every backend except the hp backend from HPLIP. If beh is used with the hp backend, the HP Toolbox will not find the printers any more.

Example URIs:

beh:/1/3/5/socket://printer:9100

On the network printer with host name "printer" it is tried to access 3 times with 5 second delays between the attempts. If the job still fails, the queue is not disabled (and the job discarded).

beh:/0/10/60/socket://printer:9100

Retry 10 times in one minute intervals, disable the queue when still not succeeding.

beh:/1/0/60/usb://Brother/HL-5040%20series

On a Brother HL-5040 on the USB try infinitely often until the printer comes back, in intervals of one minute. This way the job does not get lost when the printer is turned off and one can intendedly delay printing by simply switching off the printer. The ideal configuration for desktop printers and/or home users.

Report bugs in linuxprinting.foomatic.devel.