net.digest October 2003

October 2003

Net.digest summarizes helpful technical discussions on the comp.sys.hp.mpe Internet newsgroup and 3000-L 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

As you read this, we are less then two weeks from the end of sales for the HP 3000 – and not much closer to answers for many of the questions asked almost two years ago, when HP first made its HP 3000 discontinuance announcement. Will there ever be an emulator of any kind? Will HP allow the transfer of source code to OpenMPE, Interex or any other third party under any circumstances? Will HP allow the conversion of used HP 9000s that correspond to HP 3000 models? The list goes on.

Let me be perfectly blunt. In my opinion, HP has done little more than throw its HP 3000 customers enough crumbs to prevent mass defections. Furthermore, it has no real intention to allow MPE-IMAGE to survive in any form. It has dragged its collective feet at every opportunity; searching for reasons why something cannot be done, rather than searching for ways to make something positive happen for its once-loyal HP 3000 customers. The truth is that when HP made its announcement, it had no reasonable solution for its small and medium-sized business HP 3000 customers running mostly homegrown software. In my opinion, it still has no reasonable solution two years later.

September marked the 58th anniversary of the first real computer “bug” when a “moth [was] found trapped between points at Relay # 70, Panel F, of the Mark II Aiken Relay Calculator while it was being tested at Harvard University.” In the same vein, postings on the list included a pun-fest when someone quoted HP’s new ad that claims the company might be able to make a cell phone someday “so small an ant could use it.” We had the “antsaphone,” text messaging as an “antscillary” function, “ANTicipation” of the first shipment, a question whether it would have an “ANTenna”, and of course the observation that HP marketing has “HIGH hopes. They’ve got HIGH hopes. They’ve got high, apple pie in the sky hopes...”

A less amusing note on 3000-L was the report that this summer HP bought two $30 million Gulfstream jets. Many people wondered how many layoffs would have to occur to pay for HP execs to fly around the world in style. Then there was the ironic comment on HP’s “Invent” slogan, when CNET reported that “HP buys into Web services management — the company plans to acquire start-up Talking Blocks to accelerate its push into the Web services management market.”

There were many more off-topic threads of interest as well as the usual political and religious rants; however, there was still a lot of good technical advice, some of which follows.

I always like to hear from readers of net.digest and Hidden Value. Even negative comments are welcome. You can reach me at john@burke-consulting.com.

Homesteader Alert – Cut This Out and Save

Paul Edwards, former HP SE, long-time independent MPE and IMAGE trainer, and MPE and IMAGE consultant recently faced the problem of replacing the disk drives in his Series 9x8 system. He decided to pretend to be a homesteader or self-maintainer, go out on the Internet and acquire the new drives, and install them in his system. Edwards’ experience, and the great checklist he developed, along with commentary follows:

LDEV 2 on my 9x8 with MPE/iX 7.5 was not responding. A power down and restart resolved the problem for a few hours, but then it would stop responding again. The drive was an internal HP branded Seagate ST34573N 4gb device. I decided to replace both internal drives with larger drives to test 7.5’s solution to the 4GB limit on LDEV 1. After I consulted IODFAULT.PUB.SYS for SE disk drives, I purchased two Seagate ST19171N 9GB drives on the Internet. When I first tried to do an INSTALL to the two new drives, I received the following error:

Start reading and building labels
Status -4 ($FFFC) From subsystem 113 ($71)
ERROR on first read disk page 1 from rendezvousio
System Abort 341 from SUBSYSTEM 148
Secondary Status: Info= -4, Subsystem 113
SYSTEM HALT 7, $0155

I could see the drives with ODE MAPPER, so they were configured OK, and there were no problems with jumpers or terminators. It turns out the drives needed low-level formatting before MPE could use them.

After several false starts, I came up with the following guide for installing disk drives. Of course if you do not need to replace the system volume set or any of its components, you do not need to do an INSTALL. [Caveat: This may not be the only solution but it worked for me. Other system configurations may not work this way.]

1. Make sure you have a good current full backup and CSLT. A CSLT should be done at least twice monthly, in my opinion, during the full backup cycle. Make sure you use DIRECTORY on your STORE command and, to be safe, run BULDACCT.PUB.SYS prior to each backup. Run VSTORE.PUB.SYS against all your backups to verify the quality of the media and run CHECKSLT.MPEXL.TELESUP to verify your CSLT each time. If you can run this on different DDS drives, that will provide an additional verification.

2. Have a PC available with an inexpensive SCSI card in it. I have an Adaptec 2930 PCI card. After setting the proper SCSI address and connecting the drive, power up the PC and use the SCSI utilities to format the drive. Adaptec uses Cntl-A during bootup to access their utilities. The format took about 3 hours for each drive.

CAUTION: Never use FORMATVOL in VOLUTIL because it will never work or finish in your lifetime. Be aware of the terminator jumpers on the drives and set them correctly. Mine were missing, so I used the ones from the old drives. It really helped to download the installation instructions from the Seagate Web site for the drives.

3. Since I replaced LDEVs 1 and 2, I removed both internal drives and installed the new ones making sure I set the proper SCSI address on both new drives. I then had to reinstall MPE from my CSLT and restore my full backup.

4. Use the Checklist J in the MPE software Maintenance Manual for the Install. Follow the directions very carefully.

5. After the initial system startup, logon as MANAGER.SYS at the console. Using DSTAT ALL, LDEV 2 showed being in an UNKNOWN state. Use debug as follows with n being the drive LDEV (2 in my case):

Debug
>msec #n.0,8
>msec #n.100,40
>exit

There will be a lot of prompts for each line and enter zero to each one. This resets the drive header information. [Editor’s note: Someone asked about using the SCRATCHVOL command in Volutil. This will not work in this case; you need to reset more. Stan Sieler contributed the following: SCRATCHVOL only flips a bit somewhere in one of the first two sectors, while the two “Modify Secondary storage” commands above zap a lot more data (68 bytes worth, to be exact, assuming current input base is still hex).]

6. Go into VOLUTIL. Use the NEWVOL command to add the drive to the volume set. CAUTION: Use a different Member name so that you won’t get a duplicate member name message and have to start all over (including a three-hour re-formatting like I did). I used Membr2 instead of Member2. Exit VOLUTIL. DSTAT ALL should show the new drive properly. [Editor’s note: We are unsure why using a different name was necessary, since we’ve all had the experience of rebuilding a volume set after replacing one or more drives and using all the old member names. If anyone has an explanation I will gladly publish it.]

7. You should be able to finish the rest of the checklist J steps to restore the backup and finish the Install and reboot. There are steps that give you the option of using DIRECTORY for the restore or BULDJOB1 and 2 to get your accounting structure completed. I restored BULDJ@.@.@ and streamed the two jobs. Then I restored the rest of my files with KEEP and OLDDATE. Everything works okay now.

Thanks to Paul for this great guide for homesteaders and self-maintainers.

Going Headless

A user posted to the list, “The console port on our N-Class locked up. HP says the control board the console is hooked into needs to be reset. The memory on the board filled up because console messages were routed to memory because there was no display for the messages to go to. So it’s not possible to go headless? This seems strange since the console logs are going to a file. Is there any way to prevent this?”

One suggestion involved configuring the port to ignore flow control. The other suggestion, and the reason for including this here, was from Mike Hornsby, with some additional information supplied by Lars Appel: “Take a look at EASYTIME.PUB.SYS. When enabled, it buffers the console messages to a file and nothing goes to LDEV 20. And if you change your mind and want console messages back on the console, you can use ETSETUP.PUB.SYS to change the settings.”

I always thought Easytime had a lot of potential, but was pretty much ignored by HP marketing for some reason. It seemed to me the trapping mechanism could have been used as the basis for a client server management console.

Primaries, Secondaries, and Synonym Chains

It is always a good idea to refresh one’s memory of how IMAGE works. And it is especially good when the Father of IMAGE, Fred White, can do the refreshing.

Here was the question as posted: Say you have a record in a master set that lives at address 100, a second record tries to write to the same address, but is sent to record 133 with a pointer written at record 100. This process continues and we have a synonym chain. If you have a program that deletes records by record address (probably a bad choice), do the records move up in the chain? If the record at location 133 is deleted, what happens to the records further down in the chain? Do they move up one slot? Say you delete record 100; does record 133 move to slot 100?

David Powell replied, “I think the only time records actually move is when you delete the chain head (address 100 in your example). When you nuke a secondary, all that happens to other records is that pointers get updated.” This prompted Fred White to note: “David is absolutely right (almost). The chain count in the primary also gets decremented. Master entries may be moved but only if they are Secondaries (of a Synonym chain) and only under one of the following two circumstances:

“A DBDELETE of a Primary entry which has one or more Synonyms. In this case, the first Synonym is MOVED to the Primary address to become the new Primary of that Synonym chain.

“A DBPUT which discovers that the correct Primary address for this new entry is occupied by a Secondary of some other Synonym chain. In this case, DBPUT finds an empty location and MOVES that Secondary to this new location and changes the pointers of the predecessor and successor (if any) of that MOVED Secondary. Finally, DBPUT stores the new Primary in the location formerly occupied by the MOVED Secondary.”

John Burke is the editor of the NewsWire’s HiddenValue and net.digest columns and has more than 20 years’ experience managing HP 3000s.