Q:
Topic thread:
Printer Systems (PRINT - NA/ATS)
Printing Systems Manager (PSM)
My user frequently needs to reprint portions of a print file that is
over 1 gigabyte in size. They go to the appropriate tab under the
notebook icon for the print job and specify the range of the pages to
reprint. They notice that at this point the printer whose logical queue
the job is in becomes unusable for other work. Since some of their
"print range" requests take up to four hours, this is a problem.
Does what they are reporting to me make sense? Is there a way around
this problem that I can suggest? I have already asked that they cut
down the size of each print job.
Thanks.
A:
Researching... The responder who is best able to address this question
is out until Monday, 3/1. In the meantime, I will try to get an answer
for you.
Q:
A:
Based on the information in this item and in 00057,HNX,000, the print
range requests on the customer's 1 Gigabyte job can take up to four
hours. Development would like to know approximately how many pages are
in this 1GB job.
A:
One more question: How many queue devices does the customer currently
have defined? If there is only one defined, then nothing else can happen
while the print range request is being processed. However, if four are
defined, for instance, you can tie up one queue device, leaving the
other three to still do work. (Note that, conceptually speaking, the
number of queue devices in PSF/AIX equates to the maximum concurrent job
value in Infoprint Manager.)
Q:
Q:
The 1 Gb job they were testing the other day had 106,000 logical pages.
When I broached the subject of queues, their immediate response was that
when they tried queues, they found that they couldn't get jobs to
dispatch to other printers, they'd all get clogged up on one printer
when there were other printers sitting idle.
Thanks for your help on this.
A:
First, your last response doesn't make any sense, so I will try to
provide enough detailed explanation to clear up any misunderstanding.
In PSF/AIX, when you define a PSF printer, you get one queue (psf60) and
by default, four queue devices (dev1, dev2, dev3, dev4). If you sent 5
jobs to this printer, 4 of them will be concurrently processed by PSF,
but only one job will be sent to the printer at a time. The 5th job
remains in a pending state until at least one job has completed
printing.
We haven't talked about how many and what type of printers are defined
on the customers system, but it would be helpful to understand that
information. I still need to know how many queue-devices are defined for
a given printer, and in order to find that out, you could have the
customer do the following:
* Enter the qchk command: qchk -P
* Similarly issue the lpstat command
* Provide you with the /etc/qconfig file.
Also, if you can clarify what you meant in your last response, that
would be appreciated.
Finally, development is testing a print range request on a large job and
the simultaneous submission of a small job to see what happens.
Q:
Q:
My user has three or four duplex systems that they want to drive. Any
of the systems can print any job, class, etc. If they have four queues
on each printer and five jobs. Printer 1, queue 1 gets a job. Printer
1, queue 2 gets a job. Printer 1, queue 3 gets a job. Printer 1, queue 4
gets a job. And Printer 2, queue gets a job. Once those first four
jobs get on the queues for Printer 1, then they are no longer eligible
to go to one of the other, open, printers. I think that represents my
user's concern.
I will try to gather the info you ask for and I will also try to
understand their PSM configuration as it pertains to these printers and
queues as I partly wonder if we have a problem with terms here.
Thanks.
A:
Thanks for the update, but I think we're still not clear on terminology.
PSM with PSF/AIX can be confusing, and based on what you described, I am
myself confused. I will update this item if I have more information.
When you talk to the customer, please verify that what you described is
indeed exactly what is happening.
Q:
Sending LPSTAT and /etc/qconfig via Lotus Notes to you. QCHK basically
looks like listing any of the logical printers in the LPSTAT.
A:
With PSM involved, the /etc/qconfig file probably won't show us what
we need to see since PSM doesn't use the /etc/qconfig file or the
AIX spooling subsystem. I didn't realize PSM was involved before when
I made the suggestion to Amy re /etc/qconfig and qchk. We will
probably need to take a different tack, but let's look at the files
first.
Q:
A:
Q:
A:
Q:
A:
Testing - problems updating item.
Q:
A:
(Okay, let's try this again. For a while, it wouldn't include the text
I was trying to add to the item˘)
Jeff, we all apologize. Even though this was in the PSM queue, none
of us realized that PSM was involved since the body of the question
only mentioned PSF/AIX (even though you mentioned "notebooks" and
"logical queues". Part of the problem is that the PSM responder was
unavailable last week. No wonder we're talking at cross purposes
here˘ :-( But our PSM expert is available this week, so if we can,
let's start over.
First, let's talk some terminology. In PSM, there are different objects
that can be defined -- logical printer objects, queue objects, physical
printer objects, default jobs (IVJs), default documents (IVDs), and so
on.
You said that your customer has three-four duplex IPDS systems that
they are trying to drive.
1) Are they running them in duplex, or are they running them in
dual simplex?
2) How have they configured PSM? That is, can you tell us for each
IPDS printer how many logical printer, queue and physical
printer objects are defined? What we're trying to understand is
if they've configured PSM with a 1:1:1 relationship (1 LP, 1 Q,
1 PP per IPDS printer), if they're doing printer pooling, or if
they're using the hourglass model as described in the Administrating
PSM book (S544-3964), Chapter 1.
Are they using default jobs (IVJs) or default documents (IVDs)?
You can find out by issuing the following command:
pdls -c i-v-d (or i-v-j) -r brief servername:
If so, we may need to see the attributes of those default objects
to see if there are any attributes that could be affecting their
operation.
3) How are they reprinting the job and how are they specifying the
the range of pages they want to reprint? What state is the job
in when they try to reprint the job?
Again, we apologize for the misunderstanding on our part. We want to
help you resolve this as quickly as possible.
Returning to you for further clarification of their configuration.
Q:
Here's responses. In case that doesn't format well, I'll send a Word
document to Doug & Kristin.
1) Are they running them in duplex, or are they running them in
dual simplex?
They mostly run in duplex, but do sometimes run dual simplex and have
the appropriate resources defined.
2) How have they configured PSM? That is, can you tell us for each
IPDS printer how many logical printer, queue and physical
printer objects are defined? What we're trying to understand is
if they've configured PSM with a 1:1:1 relationship (1 LP, 1 Q,
1 PP per IPDS printer), if they're doing printer pooling, or if
they're using the hourglass model as described in the Administrating
PSM book (S544-3964), Chapter 1.
Are they using default jobs (IVJs) or default documents (IVDs)?
You can find out by issuing the following command:
pdls -c i-v-d (or i-v-j) -r brief servername:
If so, we may need to see the attributes of those default objects
to see if there are any attributes that could be affecting their
operation.
They are using the hourglass model. For example, for duplex statement
printing, they are using the following logical printer:
op-esp-statements
op-statements-duplex
op-statements-duplex-T
op-statements-duplex-8
One Queue called "oReports", and then the following physical printers:
opstmtab
opstmtcd
opstmtef
opstmtgh
Output from the pdls -c i-v-d -r brief servername:
Initial Logical
Value Document Server Printers Ready Copies
Sides Format
------------------------- ------ ----------------------- ------
----- ---------
op-esp-statements-ivd opspl1 op-esp-statements
op-mailstream-reports-ivd opspl1 op-mailstream-reports
op-simplex-statements-ivd opspl1 op-statements-simplex-B
op-statements-simplex
op-statements-simplex-N
op-refund-checks-ivd opspl1 op-refund-checks
op-refund-tags-ivd opspl1 op-refund-tags
modca-p
op-simplex-reports-ivd opspl1 op-reports-A 1
line-data
op-reports-Q
op-reports-D
op-reports-E
op-reports-F
op-reports-G
op-reports-I
op-reports-J
op-reports-M
op-reports-O
op-reports-P
op-reports-R
op-reports-S
op-reports-U
op-reports-V
op-reports-W
op-reports-C
op-duplex-statements-ivd opspl1 op-statements-duplex-T
op-statements-duplex
op-statements-duplex-8
op-barcoded-traytags-ivd opspl1 op-barcoded-traytags
modca-p
op-tags-3
op-refund-reports-ivd opspl1 op-refund-reports
line-data
op-od-3916-ivd opspl1 op-od-3916
modca-p
op-od-4000-ivd opspl1 op-od-4000
modca-p
op-refund-rpts-ivd opspl1 op-refund-rpts
ascii
op-printmail-tags-ivd opspl1 op-printmail-tags
modca-p
Output from the pdls -c i-v-j -r brief servername:
Initial Logical Printer
Printer
Value Job Server Printers Ready Locations
Requested Models Requested
------------------------- ------ -----------------------
-------------------------------- ----------------
op-esp-statements-ivj opspl1 op-esp-statements "Omaha SPC
Prod Print Room" 3900
4000
1NO3C-ivj opspl1 1NO3C 1NO3C
op-ondemand-reprints-ivj opspl1 "Omaha SPC
PAWS" 3916
op-mailstream-reports-ivj opspl1 op-mailstream-reports qcroom
3NE7J-ivj opspl1 3NE7J 3NE7J
2NO4G-ivj opspl1 2NO4G 2NO4G
op-simplex-reports-ivj opspl1 op-reports-A "Omaha SPC
Prod Print Room" 3900
op-reports-Q
4000
op-reports-D
op-reports-E
op-reports-F
op-reports-G
op-reports-I
op-reports-J
op-reports-M
op-reports-O
op-reports-P
op-reports-R
op-reports-S
op-reports-U
op-reports-V
op-reports-W
op-reports-C
op-simplex-statements-ivj opspl1 op-statements-simplex-B "Omaha SPC
Prod Print Room" 3900
op-statements-simplex
4000
op-statements-simplex-N
op-duplex-statements-ivj opspl1 op-statements-duplex-T "Omaha SPC
Prod Print Room" 3900
op-statements-duplex
4000
op-statements-duplex-8
op-refund-checks-ivj opspl1 op-refund-checks "Omaha SPC
Prod Print Room" 3900
4000
op-barcoded-traytags-ivj opspl1 op-barcoded-traytags "Omaha SPC
Prod Print Room" 3900
op-tags-3
4000
op-od-reprints-3916-ivj opspl1 op-od-3916 "Omaha SPC
PAWS" 3916
op-od-reprints-4000-ivj opspl1 op-od-4000 "Omaha SPC
Prod Print Room" 3900
4000
op-refund-reports-ivj opspl1 op-refund-reports "Omaha SPC
Prod Print Room" 3900
4000
op-refund-tags-ivj opspl1 op-refund-tags "Omaha SPC
Prod Print Room" 3900
4000
op-refund-rpts-ivj opspl1 op-refund-rpts "Omaha SPC
Quality Control Room" ibm4202-3
op-printmail-tags-ivj opspl1 op-printmail-tags "Omaha SPC
Prod Print Room" 3900
4000
3) How are they reprinting the job and how are they specifying the
the range of pages they want to reprint? What state is the job
in when they try to reprint the job?
They are using the PSM tab that allows them to specify a page range. It
is my understanding that in PSM, this process results in execution of
the AFPSPLIT batch utility.
A:
While I am researching this I want to verify with you where the
"logjam" is happening. As I understand it from your very first
description of the problem, once you do the "range of pages" reprint
you can no longer submit jobs to the logical printer? I want to be
sure the problem starts with the logical printer and not the
physical printer (physical printer not able to process any jobs while
the range of pages is processing? Is this correct? I will update with
more research answers or questions on Monday. Leaving open.
Q:
Q:
I am not in a position to be able to get specifics on your last
question right now, nor will I be able to for at least a week. What I
just heard in a phone conversation with my customer seems to contradict
what I wrote at the very beginning. Today they said that the way they
do the split is to open the properties notebook for the physical
printer, select the appropriate tab and indicate the range to print. I
don't recall and don't have access to PSM and so can't verify with my
own eyes where the range is indicated.
A:
Awaiting information for development.
Q:
A:
After talking to development, this appears to be a bug. The only
explanation would be if the max concurrent jobs attribute was set to
one. So please verify this with the following command:
pdls -cp -r "max-concurrent-jobs" PhyPrt_name
If this value is set to 1, then this is your problem. You would need
to disable the physical printer and do a pdset to increase the value.
If this value is set to a value MORE than 1 (like 4), then this is
would appear to be a bug and a PMR would need to be opened. Level 2
can then run traces to determine what the problem is and get it fixed.
In PSM, when the max-concurrent-jobs attribute is set to more than one,
then it will start processing as many jobs as the value is set to, and
the first job done processing, will be the first to print. So the
fastest wins. The afpsplit that is happening under the covers when
printing a range of pages should NOT cause the other jobs to not print.
Hopefully the problem is simply the max-concurrent-jobs attribute and
this will be a simple, easy and fast fix. But if it is not the case,
then please open a PMR with Level 2 to address the problem.
Thank you for using WWQ&A.
Q:
Max-concurrent-users is set at 1. Will suggest they set up a test.
THANKS.
A:
I'm hoping your meaning "max-concurrent-jobs" instead of
"max-concurrent-users". If you really meant to say max-concurrent-jobs
and it is set to 1, then this is your problem.
I hope this is the case. Thanks.
Q:
Doug, the customer is still reluctant to change the max-concurrent-jobs
to something more than 1. In JES2, the similar parameter could be
changed because their print selection criteria, ie. form type, would
restrict what else could be dispatched on a particular printer. In
their PSM world, they haven't yet made such a distinction. Thus they
still fear that they'll end up with jobs already dispatched on the first
printer when their second and third printers are sitting idle. Things
are hectic there right now, but when they calm down, I'll encourage that
we do some experiementing anyway. It makes sense that if they want to
allow PSM and PSF/AIX to put a second job on a printer, during an
AFPSPLIT scenarion, then they will have to increase the
max-concurrent-jobs parameter.
Another nuance on this problem with the length of time to do an AFPSPLIT
has come up though and also I am not sure if I ever heard whether four
hours is a reasonable time to do an AFPSPLIT on 1Gb file with and
106,000 pages. So,
1. Does four hours sound reasonable for a J40 with sufficient disk when
doing an AFPSPLIT on a 1Gb/106,000 page file?
2. Is there a way to skip all the way to the end of a file then use the
backspace command to get to the first page we want to print?
Thanks.
A:
1) It is hard to estimate how long that could take, because it really
depends on a lot of different factors (content of file, i.e. lots of
graphics, how much text, size of system, memory, etc). 4 hours
doesn't really surprise me. If the customer really feels strongly that
this is excessive, then I can TRY and have our performance group look
at it.
2) Unfortunately, there isn't any way in PSM to advance to then of the
file, the count backwards. If and when they migrate to IPM V3, they
can use the forward spacing feature instead of having to do the
afpsplit. This MAY be better performance wise, but I don't know for
sure.
Let me know how their testing goes.
Thanks,
Q:
Doug, thanks for your additional thoughts & I am sorry I didn't get this
entered before you responded this last time because as a result of
testing this afternoon, we have one more question.
Since they don't know how many pages are in the print file, they usually
enter all 9's for the end of a range. The test file they were working
with today is substantially smaller than their production print files,
but when they tried to put a realistic number, instead of all 9's in the
end-of-range field, it *SEEMED* to take less time than they would have
expected. Does that make sense or are we grasping at straws? I can
imagine ways that the code could have been written where all 9's would
be alot different than an accurate number, on the other hand it could
have been written to simply stop processing at EOF, whether the range
number was hit or not.
Thanks for you thoughts.
A:
I would tend to suggest that you don't use the all 9's, because I would
think that it will try and process all of those pages (although I
haven't confirmed with development). Page 173 of the PSM Reference
manual explains that if you only put the starting number and leave
the ending number blank, the job will start on the page specified and
print to the end of the file. If you leave the starting page blank and
specify an ending page number, it will start at the beginning and end
on the page specified. HOWEVER, I tried this in the GUI and, the GUI
requires both fields to be filled in. But this DOES work with the pdset
command. I queued up a job and did a 'pdset' on the job and it accepted
the command with the one variable:
pdset -c job -x "page-select=100:" 21
Here I am having the job start on page 100 and printing to the end of
the file on job number 21. Please try this from the commandline as a
test to see if the job processes any faster. I realize this may not be
as convenient for the operators (that is, not being able to do this from
the GUI), but it is a quick command and may be worth their time.
Please let me know how this test works out.
Thanks˘
S e a r c h - k e y w o r d s:
page-select psm aix print range of pages afpsplit psf gui
WWQA: ITEM: RTA000156074 ITEM: RTA000156074
Dated: 04/1999 Category: PSMAIX
This HTML file was generated 99/06/24~12:43:43
Comments or suggestions?
Contact us