High Volume Printing Fails

[Greg Carlo]

Hello,
The business recently created an SAP job that generates ~1400 spool requests. The requests range in size from 5k to 40k. SAP spools the jobs to the local HPUX server which then sends via LPD to a Windows print server running Print Services for UNIX. Windows then does it's thing sending off to the printer over the network.

All 1400 jobs from SAP make it into the HPUX queue without problem. HPUX then sends to Windows and between 0 and 5 minutes a request hangs and printing pauses. It may resume on it's own or might need a disable/enable of the HPUX queue to continue.

We spent the last 3 days troubleshooting and narrowed the problem to HPUX. We performed a test where the first HPUX system with SAP installed sent the print requests to a second HPUX system via LPD. The same problem occurs. We narrowed further by creating a printer that spools to a file locally on the HPUX server. Generating the 1400 jobs goes fast but still within 5 minutes it hangs.

This has been tested on two HPUX 11.31 systems in the environment. We have a ticket now open with HP and curious to see what they come back with.

While we wait for their response, any thoughts?

Thanks,
Greg

[Bill Hassell]

> HPUX then sends to Windows...

How do you know that this is an HP-UX problem? I have had nothing but nasty hangs using Windows (NT, 2003, etc) for years. Windows LPD service is notorious for unexplained hangs, especially if you have an overloaded network and/or Windows server. I have never had any problems with thousands of jobs sent to hundreds of printers using HP's JetDirect LAN cards (built-in to HP printers) and JetDirect external print servers. The LPR/LPD protocol is archaic as a protocol (in SAM, this is called 'remote' printing) and quite difficult to troubleshoot.

JetDirect protocol is an HP proprietary protocol using port 9100 (rather than 515 for LPD/LPR) and has a well defined handshake and status checking. Jobs can be traced down to the network handshake level.

As you have seen, HP-UX has no problems creating 1400 jobs (and I have seen 2000 to 4000 jobs queued with no issues). If you migrate all the printers to JetDirect connections, you'll be able to share the printers with Windows, Solaris, Macs and Linux as well as HP-UX.

[Greg Carlo]

Bill,
Thanks for the comments. I have seen the issues you mention about Print Services for UNIX on Windows. But it didn't apply in this case as our testing with Windows removed from the equation also yielded the same results.

HP answered our case and came back with the recommendation of applying patch PHCO_40128. We applied the patch and the errors went away. Looking at the patch notes, they discuss an FIFO queuing problem where the OS has trouble keeping up with the flood of requests being sent from the application. Apparently, some jobs would get stuck in the queue and would not be serviced by the spooler until a scheduler restart was done.

HPUX is great and I love working with it. But, in this case, I'm happy it was an HPUX problem. If it was a Windows problem I doubt we would ever find a solution to the issue.

Regards,
Greg

[VK2COT]

Hello Greg,

We are missing lot of debugging details so it is not easy to give a definite answer yet.

Last year I was involved in a similar
problem.

It turned out that QoS (Quality of
Service) settings on the VLAN was the
culprit. Nothing to do with HP-UX :)

In other words, make sure to test all
elements in the chain (network, DNS, disk
IO, and so on).

Cheers,

VK2COT

[Bill Hassell · Accepted solution]

Wow, this has been a problem in 11.23 (PHCO_39894) as well as 11.31 (PHCO_40128). Technically, all HP-UX releases would have the BSD limitation of a 3 digit print queue ID (0-999) or no more than 1000 active BSD jobs at the same time. This won't come up with HP-UX to HP-UX remote printing or JetDirect printing where the queue ID is a 4-digit number (10,000 jobs at the same time).

[Greg Carlo]

VK2COT,

Thanks for the reply. I agree that there are a lot of variables that could contribute to the problem. That's why we narrowed our testing to the fewest number of components we could. We took Windows and network out of the equation by performing all our tests locally.

We created a FOR loop that created 1400 print jobs 8k in size and sent it to a locally defined "print to file" printer on the HPUX box. The FOR loop created all the jobs in about 30 seconds. The spooler processed a large percentage of the jobs but stopped. This was all local to the machine. No outside components were involved.

We applied the patch from HP and all was fixed. The spooler processed all the jobs in the queue without hanging. The ideal setup would be to send fewer larger jobs and the business is looking into adjusting the SAP program to do this.

In the meantime, we got it working. I suppose our high volume printing scenario here is not typical in most businesses as I would have thought HP to have fixed something like this years ago. Regardless, it is fixed now. Thank you all for participating in this thread.

Regards,
Greg