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
During the holiday period, traffic on 3000-L tends to be down a bit and also tends to the off-topic and wildly off-topic. This year was no exception. However, there was still plenty of technical content to allow me to pick and choose what I considered the most interesting and most useful items for net.digest and Hidden Value.
There are two recent trends on 3000-L that I want to briefly mention, one disturbing and one puzzling. It seems to me that the some of the off-topic threads have become more personal and more mean spirited than in the past. 3000-L has always been noted for its civility when compared to most other listservers out in the ether. It would be more than just a shame if that changed. Please think about this before making a possibly inflammatory posting. The other trend I have noted, and puzzled over, is the number of first-time posters who have clearly never seen an HP 3000 until just recently, much less tried to use one. What does this mean?
Finally, before we get going, December saw a very interesting thread over on the OpenMPE list (subscribe at www.openmpe.org) about MPE licensing for a platform emulator. If you have any interest in MPE continuing, especially on an emulator, you should sign up for this list and contribute your thoughts.
As always, I would like to hear from readers of net.digest and Hidden Value. If something from these columns helped you, let me know. If youve got an idea for something you think I missed, let me know. Are you seeing a pattern here? You can reach me at firstname.lastname@example.org.
How to create a scratch pad area on your HP 3000
The question seemed simple enough, I want to create a directory off the root, say /transfer, and allow any user anywhere on the system full access to it. I want any user to be able to place files in this directory, as well as allowing any user to delete any files that exist in the directory. I thought my answer was ALTSEC /transfer;reppair=(CD,DD,RD,TD:@.@) but that didnt do the trick. What can I do?
The first answer was I can never remember :ALTSEC ACD syntax details. But fortunately POSIX chmod makes your request quite easy::XEQ CHMOD.HPBIN.SYS 777 /transfer
This created a directory like /tmp that anyone could deposit a file into. However, the files created in this manner are not automatically accessible to other users.
The follow on answer was Your ALTSEC does work, however, perhaps you are forgetting to do an ALTSEC on every file placed in the /transfer directory to allow other uses to access those files. For example,
:XEQ CHMOD.HPBIN.SYS 777 /transfer/foo
will permit any user to do anything to foo in the /transfer directory.
But, again, this does not answer the original question. The first suggestion will let anyone CREATE a file in /transfer and the second suggestion will let anyone do anything to the file foo in /transfer, but neither creates a repository that will let anyone create a file in it and let anyone else do anything to the file without an extra step. Here, however, is a way to do what was originally asked by making use of MPEs security features along with the POSIX symbolic link feature.
1. create an account with access=(r,w,l,a,x:any)
2. create a group in this account with access=(r,w,l,a,s,x:any)
3. either use passwords or remove IA, BA access for security
4. create a symbolic link, NEWLINK /transfer,/ACCOUNT/GROUP;symbolic
You now have scratch space that anyone can use without restriction.
A command file tip you may have forgotten
We have probably all created a command file such as PRINTFILE with a first line something like
parm sfid, dev=200, pri=8, copies=1
We then use it like this to print two copies of spoolfile ID O1234 using the default values for dev and pri
Michael Anderson and Greg Chaplin reminded us, in answering someones question, that an alternate format could just as easily be used to keep you out of comma hell:
printfile o1234 copies=2.
Something to consider while using Samba on the Internet
Mark Bixby provided this helpful hint: Ive been doing some Samba 2.2.5 performance testing the past few days, and I have found that adjusting one of the TCP timers in NMMGR can make a substantial performance improvement, at least in my DSL-Internet-Cupertino environment.
As I was doing large cut/paste downloads (about 50Mb) via Samba, I noticed that the status icon for the VPN software needed to access the HP intranet was showing long pauses (5-10 seconds) where no packets were being received. I then cranked up the Samba debugging level in smb.conf and also added some extra debugging code of my own. I was able to determine that it was the send() sockets function that was experiencing the long pauses. In other words, Samba was processing without any undue delays, but was being blocked waiting for the MPE TCP stack to complete the send() operation.
Because my DSL connection is slow compared to the network connection of the Samba server, the MPE TCP stack would eventually time out waiting for an ACK, pause according to the exponential back off algorithm, then try to resend the un-ACKed packet. These retry pauses quickly double to rather large amounts of time that greatly exceed what is needed to retransmit over modern TCP networks, thus the remote client ends up experiencing large delays waiting for retried packets. For greater detail about this, see the 3000-L archives for numerous postings from James Hofmeister.
Adjusting the TCP timers will change the performance of a TCP send(). Go into NMMGR and navigate to the NETXPORT.GPROT.TCP screen. Change Retransmission Interval Lower Bound (Secs) to the minimum of 1. Then validate, exit, followed by a networking restart or a reboot. You should now see better Samba performance.
Caution, though, your mileage may vary. Also, I can imagine certain configurations where you wouldnt want to lower the retransmission lower bound, i.e., if your 3000 predominantly talks to remote TCP clients with really slow network connections.
Donna Garverick graciously provided the URL for one of James postings that I have found particularly useful in the past. I used it to solve a problem I had using Lars Appels port of htDig with Apache. Marks tip and James posting (see raven.utc.edu/cgi-bin/WA.EXE?A2=ind0112A&L=hp3000-l&P=R49) should be added to your tips folder.
In case you missed it
Courtesy of Gary Jackson and Bill Cadier, if you use Mirror/iX and plan to install MPE/iX 7.5, be sure to install GR patch MPEMXC6B or its successor. Failure to do so will lead to unpleasant consequences.
This migration stuff is not easy to plan for
Many people have been asking questions like this that appeared recently on 3000-L: Im trying to help someone size an HP-UX system to replace their HP 3000/969/100. Has anybody done something similar? What HP-UX box did you go to? Anybody have any benchmarks that can be used to compare performance amongst various HP 3000s and 9000s?
If you are going to replace your applications with purchased applications, then your vendor of choice should be able to help you size the hardware you will need. For example, if you plan to join the big boys and move to something like SAP R/3, both SAP and your hardware vendor of choice, HP for example, have detailed sizing programs available to help you. From personal experience, I recommend you and your management be sitting down when you read the results.
Birket Foster, of Platinum partner MB Foster, had this take on the question: In the matter of performance, it depends. Is the application moving or being replaced? If it is being replaced by off the shelf software, then there may be performance characteristics available from the vendor. If you are moving it, there are greatly differing performance footprints depending on the number of connections, the workload including OLTP and batch jobs and the database chosen. You must clearly understand the workload and the pieces that make it up before anyone can give an estimate or model it. HP has a benchmark center as part of the migration center which may help with estimates.
Alan Yeo added, Benchmarking is really only useful if the benchmark actually benchmarks something that you do. It is utterly useless to anybody else other than as a vague finger in the air gauge. In the above example what you have is a measure of how that particular vendor chose to migrate/re-write their application. There are just too many variables to make it useful to anybody else.
While there are some measures that may give you some sort of guidance as to the size of HP 9000 you might need, in reality its only when you have done the work that you will know how efficient what you have migrated is, and therefore what configuration you will need. And as I suspect for those companies migrating more than a single standalone application what you actually end up running could be spread across several servers, and you will never have a box that was running everything exactly as your HP 3000 was.
So by all means gather together as much anecdotal benchmark information as you can. Accumulated together it will give you a guide to the ballpark you will be operating in. But would I go out and order a processor based on other peoples results? Never. Also most migrations are going to take a reasonable amount of time so why worry about sizing the production system now? Instead, worry about the size and cost of the development/testing box and tools that you need now.
From the seriously cool department
Mark Wonsil has ported text2pdf, a freeware text to Adobe PDF converter program written by Phil Smith of the Electronic Publishing Research Group. Actually, he did it some time ago and Im only just now checking it out. Check it out at invent3k.external.hp.com/ ~MGR.WONSIL/text2pdf.html
Remote reboots are something I have not thought much about lately because I have had Telamons Console Engine available giving me remote console access. So, when someone asked about ways to do remote reboots, it was interesting to see all the possibilities that now exist.
Christian Lheureux offered the classic solution: Go to the console issue CTRL-B and then enable the remote console with ER. Of course, you need to have a modem ready in auto-reply mode and a phone line.
Jonathan Backus added that another great way is the WebConsole.
Cory Black suggested using a PC equipped with PCAnywhere as the system console. Ive successfully used this technique also, although it requires that you can access your corporate LAN remotely.
John Korb suggested using a power switch that you can access via Telnet or web browser and to power cycle the CPU. If you have AUTORESTART (an extra cost software product) or AUTOBOOT configured, your system will come back up. Kind of crude, but workable.
In the same vein, Gavin Scott noted, If you have one of the recent A/N-Class systems then you can also turn the system power on and off from the built-in Service Processor I believe. The externally accessible power switch turns out to be little more than a hint to the system.
Finally, Chris Bartram reminded everyone about the 3000-L FAQ (available at www.3kassociates.com) and item 3.1.4, How can I make the 3000 reboot automatically if it crashes? It shows how to setup your 3000 to reboot automatically when it crashes (no additional software needed just FOS).
John Burke is the editor of the NewsWires HiddenValue and net.digest columns and has more than 20 years experience managing HP 3000s.
Copyright The 3000 NewsWire. All rights reserved.