Hidden Value September 2000

Hidden Value details commands and procedures in MPE that can improve your productivity with HP 3000 systems. Get a free NewsWire HP 3000 “3000 for 2000” cap — submit your MPE tip directly to us here at the NewsWire. Send your tips to editor@3000newswire.com, or fax them to 512.331.3807.

Edited by John Burke

Has anyone encountered a problem when running the Posix shell in a batch job stream? I have a job that does a DSLINE, several DSCOPYs, then invokes the Posix shell to compress some files. But occasionally the job stream skips the commands in the Posix shell. How do I avoid this?

Barry Lake replies:
I’ve encountered that, too. I worked around it by putting all of my Posix/shell commands in a command file (shell script), and then running the shell calling the script in the info string:
!run sh.hpbin.sys;info=”../CMD/make_tarball”

How do I turn HP’s Predictive Support off?

Gary Paveza replies:
Run PSCONFIG.PRED.SYS, then option 4, configuration menu, then option 5 schedule, then option 1 schedule (turn off).

We have this set this way on all our HP 3000s. Our operators manually STREAM PREDICTA.PRED.SYS. This allows the operator to watch exactly when it runs. This way they know they have to print out the report generated (and notify anyone if predictive dials out).

Will a “partial” backup store only the datasets that have been modified, or will it store the entire database? We’re working on some recovery plans and we’re wondering if restoring the latest partial over the latest full will cause the database to be inconsistent. Doesn’t seem likely even if only rootfile and datasets that have changed are stored. Everything should be consistent, right?

Doug Werth replies with the most complete and shortest response:
The (current) default is to store the entire database unless the ;PARTDB option is used.
From a recovery standpoint, if you restore your most recent partial backup first, and then restore from your full backup using the ;KEEP option it won’t really matter whether the entire database was on the partial or not.

While your method of restoring the partial over top of the full will work, restoring the partial first followed by the full using “;keep” is more efficient.

Is there any reason not to turn on the “enable database for dumping” flag? Overhead, performance, disk space, etc.? I infer that an I or J file is created only if there is a major error. Having just had the horse stolen (an error with the databases disabled for dumping), I think that we ought to bar the stable door. What do you think?

Steve Cooper replies:
My understanding is that there is absolutely no overhead until an IMAGE abort occurs. Then, if the flag is on, a PRIV dump file is created, so that one can tell what happened. I strongly recommend that everyone immediately enable this flag on all databases.

SIGIMAGE has discussed the fact that several flags have questionable default values, and it has recommended some changes to HP.

Ken Sletten (Chairman of SIGIMAGE) adds:
I confirmed Steve’s note above with the HP Response Center several years ago, when I enabled dumping on all of our IMAGE databases: no run-time overhead unless there is a TurboIMAGE ABORT. With DUMPING ENABLED, in the rare case of a TurboIMAGE ABORT you will (at least most of the time) get not just one but two PRIV mode dump files, one starting with “I” and the other starting with “J.”
HP has responded to the SIGIMAGE suggestions. There is now an item on the latest “Now / Soon Available in Image/SQL” list (to be presented at the SIGIMAGE meeting at HP World in Philadelphia) that says: “ONE-time make DBUTIL DUMPING and DSEM flags default = ENABLED.”

“One-time” means that a bit will be set and left set to indicate that DUMPING and DSEM flags were automatically turned on ONCE for a particular database. If during subsequent updates the DUMPING and DSEM flags are found in the DISABLED state (after having once been automatically ENABLED), they will NOT be automatically ENABLED again. (Of course users may continue to ENABLE them manually with DBUTIL at any time.)

I managed to get almost everything up and running on our new HP 3000/979 (we migrated from a HP 3000/957) over the weekend. Everything, that is, except the terminals that are connected to a DTC. I get the error “CAN’T FWRITE $STDLIST FROM “STARTLOGON” ON DEV ####”. What can I do?

John Burke replies:
Be sure to reset any DTCs. By changing systems, you changed the MAC address of the host. The DTCs carry a table relating name to MAC address. A trick you can often use to avoid resetting the DTC is have someone (at the DTC prompt) issue the connection command:
c hostname -d

The “-d” tells the DTC to flush and rebuild its host-to-MAC address table. This comes in handy when you have dozens of DTCs and resetting them all at once is impractical.

I had a problem today with connections to my HP 3000. After about 100 users, VT sessions would no longer get a prompt. It turned out that I had my TCP connection count set to 128. I had to increase that. Is there a way to see how many connections there actually are? Do I need to reboot in order to change the maximum number? (NETCONTROL;update=all does not do the trick.)

Ed Stouder, Doug Becker and Doug Werth reply:

:RUN NETTOOL.NET.SYS;INFO=”STATUS;TCPSTAT;TCPGLOBAL @;QUIT”

This will show you something like this:

TCP GLOBAL Statistics
Maximum number of connections allowed : 128
Maximum number of sockets allowed : 512
Number of connections currently open : 14
Maximum re-transmissions per packet : 1

As you can see, the “Maximum number of connections allowed” in this case is 128.
To enable the change you need only stop and the restart the network, not the entire machine. Unless you are running a 24x7/True-Online backup, then stopping/restarting the network should be a simple operation to perform at backup time without restarting the entire system, since you will have already logged off all of the users. This could save you a considerable amount of time.

I have several DDS-2 tapes that have been written to by PC-based backup software. When loaded into the HP’s tape drive, it tries to find the tape header and fails, telling me there is a parity error on the tape. How do I erase, reformat or write a new header to the tape so it can be used on the 3000?

John Clogg replies:
The error is simply informing you that AVR (automatic volume recognition) failed. It will treat the tape as unlabeled. You can go ahead and use the tape like any other, and it will write over the old label. No special handling is required.

How do you set the network port on an HP 3000/979 from the AUI port to the ThinLAN port — or does this have to be done by the CE internally?

Dennis Heidner replies:
You should have a CE do it; the setting is a jumper on the MIO card.

[Editor’s note: There was some controversy as to the default and whether it might now be auto-sensing. From the “Installation Guide: HP 3000 9x9KS Enterprise Servers”, March 1999, available at docs.hp.com, we have: “CAUTION The ThinLAN and AUI LAN can not be connected at the same time. Only one LAN connector can be active at a time. The AUI connector is selected as the default connector. If you need to have the ThinLAN connector active, contact your local Hewlett-Packard Service or Sales representative.”]

How can I reset the CPU and CONNECT values displayed in the REPORT command?

Larry Simonsen, Bob Sorenson, Jim Phillips and Bruce Toback reply:
The :RESETACCT command resets these counters.

Is there a Posix shell command to list the process/user who has a file open?

Michael Berkowitz and Mark Bixby reply:
In Unix-land that would be fuser(1), but in MPE-land you must do it the MPE way. Depending upon the level of detail you want, then from the shell: callci listfile /file,8 or, the more detailed view, callci listfile /file,9

How do I find information on how to use tar on the e3000?

Mark Bixby and Barry Lake reply:
Assuming you have your Posix environment set up correctly, any of the following will work to get information on tar:

:XEQ SH.HPBIN.SYS -L
$ man tar
$ help tar
$ tar -?

For tips on setting up your Posix environment properly, see Michael Hensley’s paper www.allegro.com/papers/posix/ pnpposix.html.

Keven Miller adds a neat trick for using “man” without having to go into the shell:
One of my logon scripts executes SETVAR TERM “hp2392a” so that from the MPE CI prompt, assuming “hpbin.sys” is in HPPATH, I can issue “man tar” and see the documents.

No matter what I put in for this printer with an add-on tray, it still prints from the add-on and not from the original tray, marked “2.” How do I make it print from the original?

Sam Knight replies:
You need to specify a “setup_file” in npconfig for the printer. In the “setup_file” add a line for the tray selection. It should be something like “<esc>&l8H”. See your printer manual for the PCL code.

Jeff Kell adds:
The “data_tray” directive is wiped out by any subsequent setup or environment file (bear in mind, if no setup or environment file is used, MPE supplies you a landscape-mode 132 col x 66 line default), as they generally begin with a hard reset, putting the printer back to default. The “banner_tray” does work (no setup).

From time to time, I have seen (in Glance) P11 MANAGER.SYS take up all our disk IO. The event will last for a minute or two, and in that period of time all other processes seem to hang. After P11 finishes doing whatever it was doing, everything continues to run fine. What is P11 is doing? The machine in question is running MPE/iX 5. PowerPatch 7.

Christian Lheureux, Bill Cadier, Andreas Schmidt and Richard Bayly reply:
Pin 11 is probably the XM (Transaction Manager) checkpoint server. It checks the Transaction Manager logs on each volume set and ensures that corresponding transactions have been completed and posted OK to disk. If you have a lot of activity (changes) in files covered by XM such as Image data bases or KSAM/XL files the XM checkpoint server will need to run often. This is a typical reason for XM overhead but certainly not the only reason.

Carl McNamee adds:
One possible way to ease the pain, if you are on 6.0 or higher, is to go into VOLUTIL and enable ALTERCHKPTSTAT. This is done per volume set and a reboot is required to activate it. This option changes the way MPE handles a checkpoint so that doesn’t impact the system quite as bad. See the Communicator for MPE/iX 6.0 for more detailed information.

We had the same problem when loading up databases. About every 10 minutes XM would start posting and everything on the system would grind to a halt for 30-90 seconds. I’ve been using the ALTERCHKPSTAT option for over a year now on several systems without any problems.

I am trying to select trays on an LaserJet 5N using MPE’s network printing and have entered this in npconfig:

520 (network_address = 192.78.126.79
pjl_supported=false
banner_trailer=false
data_tray = 2
banner_header=false) #Scheduling LJ5N

Where do I look to find the maximum number of processes that are available on an HP e3000? Is it configurable? (I couldn’t find it in SYSGEN.)

Mark Bixby notes that there is a paper on the CSY Jazz server that contains a wealth of information on various system limits, at jazz.external.hp.com/papers/limits/intro.html

How do I turn on auto reply and auto eject for LDEV 7?

Tony Newton and Barry Lake reply:
For auto reply do

:sysgen
sysgen> io
io> mdev ldev=7;mode=autoreply
io> hold
io> exit
sysgen> keep
sysgen> exit

then a Start Norecovery to activate. To enable auto eject you can do

:xeq devctrl.mpexl.telesup dev=7;eject=enable

This does not survive reboots, so it is best to have it in a job launched by SYSSTART if you want it to be always enabled.

We got FTP running under inetd recently on our Series 937RX with MPE/iX 6.0. It works fine but the following error is output on the console many times during the day:
Stat on “/etc/bootptab”: No such file or directory.
This occurs whether we’re using FTP or not. What is this file, why don’t I have it, and what should I put in it?

Doug Werth and Kevin Miller reply:
This error is generated because you have the BOOTP service enabled but no configuration file. Rather than creating a configuration file you can disable the BOOTP service.
• Log on as MANAGER.SYS,NET.
• Edit the file INETDCNF.NET.SYS
• Comment out the line that begins with “bootps” by placing a “#” in the first column.
• Enter the command “:inetd.net -c” to make the inetd job re-read the configuration file.

Is it possible to change the database creator without rebuilding the database?

Barry Lake replies:
You can STORE the database, then RESTORE it with the “CREATOR=” option.

Jeff Kell notes:
The fastest way is the Posix shell ‘chown’ command.

Paul Christidis adds:
For those that are not comfortable on the Posix shell, the following is pretty fast too:
:ALTFILE filename ;OWNER=newname.account

I have a logfile which indicates the following when I try to purge it:
BAD VARIABLE BLOCK STRUCTURE (FSERR 105)
Unable to purge file “LOG7815.PUB.SYS”. (CIERR 384)
What can I do?

Leonard Berkowitz and Tom Madigan reply:
You can use the “PURGEFILE” command in FSCHECK.MPEXL.TELESUP. This gets rid of the file by brute force. The utility gives you come dire warnings, and there is a slight risk of corrupting your directory, but it usually works.

Lars Appel adds:
PURGELINK or RM.HPBIN.SYS should work. They do not attempt to open the file for purging and thus should not complain about the internal file structure being damaged.

My question concerns the percentages, both minimum and maximum, specified in workload manager. Is this value the percentage of the total CPU available? That is, if I specify a 5 percent maximum for a work group on a four-processor system, does that mean 5 percent of one processor, or 5 percent of four processors for a result of 20 percent of one processor?

John Hornberger replies:
The CPU minimum and maximum deals with the global percentage of CPU available. For instance, a CPU limit of 25 percent on a four-processor system equates to 100 percent of one processor or 25 percent of each of four processors. There is no way to nail a workgroup to a given processor, so at times more than one CPU can be servicing a given workgroup’s processes.

What does LTU stand for in the HP support contract?

Phil Anthony jumps on this first:
“License To Use.”

What’s the file code for a store-to-disc file?

Patrick Santucci replies:
Code: 2501, Mnemonic: STORE

And Stan Sieler points out an MPE bug:
The STORE mnemonic does not work with the BUILD command.