Net.digest summarizes helpful technical discussions on the HP 3000 Internet newsgroup and mailing list. Advice here is offered on a best-effort, Good Samaritan basis. Test these concepts for yourself before applying them to your HP 3000s.
Edited by John Burke
Changes are afoot at HP. Again. Seems like I just wrote that because I did last month, when the Board of Directors forced the resignation of CEO-as-rock-star Carly Fiorina. The Board lost little time in hiring the anti-Carly, Mark Hurd, formerly the CEO of NCR, to be the new CEO of HP. Hurd is the antithesis of Fiorina, who sometimes seemed more interested in positioning herself for a future in politics than improving HPs future. In contrast to the camera-loving Fiorina, a search party has been sent out to find Hurds personality.
The business press does not know what to make of this new marriage of Hurd with the HP board. The Associated Press reported, An HP director said that the board would look to newly named chief executive Mark Hurd to recommend sweeping changes in strategy at the ailing computer and printer maker. On the same day, the New York Times headline said the hiring of Hurd was A Break With Style, Not With Strategy. I guess we will just have to wait to see which news organization is correct, though the Times article is more consistent with what the Board was saying immediately post-Carly: that there was nothing wrong with the companys overall strategy, just the execution of that strategy.
The 3000-L carried a couple of good April Fools Day stories, with one having Carly taking over IBMs iSeries division and another reporting that Hurds first act as CEO was re-instating the HP 3000 as a key product line. Both are wishful thinking on the part of HP partisans. The HP 3000 and MPE/IMAGE are just pimples on the butt of the lumbering elephant that is the current HP. However, HP has so many problems (like its own $110 million SAP migration project that is two years behind schedule) it is just possible that it will decide to squeeze the pimple, jettisoning MPE/IMAGE to some third party just to get rid of the headache. This is probably our best hope for a future for MPE/IMAGE.
I always like to hear from readers. Even negative comments are welcome. If you spot something on 3000-L and would like someone to elaborate on it, let me know. You can reach me at email@example.com.
When is an update not much of an update?
Answer: When it is an update from HP on the status of MPE. It was mostly a reiteration of items from as long ago as last June. The only new items I spotted were, first, a plan to create a Web page on Jazz listing current beta patches (no way to download them, though) and, second, MPE/iX 7.0 PowerPatch 3, targeted for the end of April. Conspicuously absent was any mention of the progress to put HP in position to make a promised second half 2005 announcement on whether it will license MPE/iX in some form to any third parties.
Telnet: Still very much alive and being fixed/enhanced
Despite inertia on other fronts, James Hofmeister and crew are still busily fixing and enhancing telnet functionality on the HP 3000. Last month James reported, This is a follow-up to past discussions about telnet sucks wind with INETD reporting the error Could not initialize data in path with TCP or Telnet hangs at :. Some folks might have not seen the above conditions and may be blissfully unaware that telnet uses Buffer Manager buffer pools just like you see in NETTOOL -> RESOURCE for the LAN drivers and TCP. This is for those of you who have painfully learned the truth. An enhancement to NETTOOL has been implemented to display these pools as well as several new fixes to buffers leaked by Telnet. Here is what the enhancement to NETTOOL looks like:
THU, FEB 10, 2005, 1:48:04 PM
Item Subsystem Name G/N Description Used High Max
15 TELNET PTOD (G) Write Buf Pool 0 24 2048 :)
16 TELNET PTID (G) Outbnd Buf Pool 0 8 2048 :)
17 TELNET PTID (G) Inbound Buf Pool 0 4 2048 :)
18 TELNET PTOD (G) Negot Buf Pool 0 1 2048 :)
[Note: The Telnet buffer pools are created at the first telnet connection attempt and will not be displayed in NETTOOL until after they are created. This enhancement is available as BETA TEST at the HP Response Center as per SR 8606-388358 in patches NSTHDC7 for C.65.00, NSTHDC8 for C.70.00 and NSTHDC9 for C.75.00.]
System Abort 1007 from Subsystem 101
Usually I do not report on these, since the answer is almost always take a memory dump and send it to HP. However, this time it is from a known problem and, even better, a known problem with a solution. The system receiving the abort was MPE/iX 7.0. Subsystem 101 is Memory Manager. Message 1007 is Unable to create page filler process.
Bill Cadier provided the answer and fix SA1007 is usually caused by a race condition mapping a file out of memory. Two (or more) processes start the operation, one wins and maps the file out, the other(s) are left holding a lock on an object that no longer exists. The file object in question needs to be > 1GB in order for this race condition to occur because files that large are mapped out incrementally. Files < 1GB are mapped out all at once. Patch MPEMXE5B should correct the problem. The current top of the supersede tree is now MPEMXR6B. If the system has MPEMXE5B installed you should contact the Response Center.
The SYSSTART file (SYSSTART.PUB.SYS) controls what happens when you boot a system. The documentation for SYSSTART (docs.hp.com/en/32650-90907/32650-90907.pdf) shows that you can put different modules in your SYSSTART file for different kinds of reboots. What does this mean? First, if you plan to modify the SYSTART file, it must be created by MANAGER.SYS and needs to reside on LDEV 1 to function properly. Once the file is created, you can copy the file to the appropriate location:
If an error exists in the SYSSTART file, the startup process ignores it and continues execution. The following table indicates what boot commands trigger execution of what modules:
Boot Utility Used SYSSTART Module Executed
START (RECOVERY) warmstart
START (NORECOVERY) coolstart
UPDATE (START) update
INSTALL (START) reload
TurboIMAGE Databases and ACDs, can they co-exist?
As a consequence of SOX, someone asked if it was possible to use ACDs on TurboIMAGE databases. Cathlene McRae of the HP Response Center replied, For TurboIMAGE databases, you can use the ALTSEC command with file system wildcards to add ACDs to database files. For example, :ALTSEC ORDERS@; NEWACD=(R,L:@.STUDENT)
This command allows any user in the STUDENT account to read and lock the files that make up the ORDERS database. Lock access (L) is worth noting because users sometimes forget that DBOPEN mode 5 (a read-only mode) requires lock access. Even if the ACD allows access to the database files, the user must still supply a valid user class password to open the database. With this said, you can have a problem with the database if you restore it to a different account. So, to restore a database with acds, use the ;NOACD option on the :RESTORE command. You should also ensure your database modification utility respects ACDs.
RAID and MPE
Many homesteaders and late migrators are replacing JBOD with Mod 20 or 12H arrays. This is just a reminder that you should NEVER configure a Mod 20 or 12H for RAID 5. Use RAID 1 (mirroring) exclusively. Performance will likely be horrible. This is primarily because the arrays have too small a cache to support RAID 5 with MPE. This same caution does not apply to the VA7xxx line or the XP512.
Did you ever wonder where the formsmsg from the ;FORMS=formsmsg parameter of the file equation for a spoolfile is stored? Believe it or not, it is stored in the actual spoolfile. You can prove this either with FCOPY (char option) or PRINTSPF.
Be careful interpreting CPU utilization in Glance on a multi-processor system. Percent used displayed is for one CPU. Divide by the number of processors to get the system CPU utilization.
Every MPE/iX 7.5 system has six new processes, Fibre Channel IO facilitating processes (called PFPs, port facility process), even those that have no FibreChannel cards. On machines without FC these processes should be idle and not consume CPU after bootup.
Have you noticed a database, called OPSDB, that is opened by the PSMOND,PREDMON.SYS job? It turns out it has been around for some time. It is the Predictive Support Database and contains event history as well as other information.
When transferring files to another OS, variable length records file on the remote system does not mean the same thing as it does on the HP 3000. It almost certainly means bytestream file. If your program cannot create a bytestream file directly, use TOBYTE.HPBIN.SYS, which will convert classic record-oriented MPE files to bytestream files.
Those who have tried the user-defined CI function capability have indicated it works well. However, it is still a beta patch and is still only available for MPE/iX 7.5. This functionality is one of several that OpenMPE, SIGMPE and the MPE Forum have been lobbying HP to backport to MPE/iX 6.5 and MPE/iX 7.0.
The CIs variable table object is large enough to hold approximately 8,000 CI variables. The entries in the variable table are variable length. So, if you use shorter variable names and shorter string values you can create more variables. As an example, a Boolean variable needs one byte for each character in the name, one byte for the Boolean value, and several overhead bytes.
co-Chair of SIGMPE and a Board member of OpenMPE, recently updated
her System Managers Toolbox with an enhanced
sysconf job and Web access to Tim Ericsons script
pages. Check it out at invent3k.external.hp.com/~MGR.GARVERIC/.
Copyright The 3000 NewsWire. All rights reserved.