net.digest Febraury 2005

February 2005

net.digest tracks each month’s message traffic on the 3000-L mailing list and comp.sys.hp.mpe Internet newsgroup. Advice offered from the messages here comes without warranty, so be sure to test before you implement.

Edited by John Burke

January saw a number of stories about Sun Microsystems’ open sourcing of Solaris, which occasioned a query on 3000-L. “Why doesn’t HP just open source MPE/iX?” Of course no one outside the HP executive suites knows the true answer, but we can speculate that considering the earnings pressures on HP, heading the list of reasons is it sees no value accruing to itself by releasing MPE into the wild, only costs. If anyone out there has a spare $25 million or so lying around, we should talk.

Perhaps second on the list of reasons is the prevailing attitude within HP that it knows best what customers need. After many years of benign neglect towards MPE/IMAGE, HP has decided it knows what we should be doing; i.e., buying and migrating to other HP systems (preferably HP-UX). By the way, in case you are leaning in the HP-UX direction, while Sun was announcing the open sourcing of Solaris, HP announced it is shelving plans to bring Tru64’s much-ballyhooed TruCluster and Advanced File System to HP-UX. Instead, it will resell technology from storage software vendor Veritas. Some on the list wondered if this was one more nail in the HP-UX coffin.

The off-topic discourse on 3000-L was somewhat less rancorous in January than in the months leading up to the US elections, but there was still plenty of name-calling and inflexibility. However, there was also lots of good technical commentary, some of which is described below.

I always like to hear from readers of net.digest and Hidden Value. Even negative comments are welcome. If you spot something on HP3000-L and would like someone to elaborate on it, let me know. You can reach me at john@burke-consulting.com.

Transient space: What is it, and how much do you need?

The disk and volume management utility VOLUTIL has three commands: NEWVOL, NEWSET, and ALTERVOL, with parameters that make it possible to specify the maximum allowable amount of both transient and permanent space for each disk drive on the system. Even though the process is simple and well-documented in the Volume Management manual (see in particular page 27 of the PDF file at docs.hp.com/en/32650-90491/32650-90491.pdf), there remains considerable mystery surrounding transient space, particularly what it is, how much you need and where you need it.

Transient space is the disk space MPE/iX uses for virtual memory. Transient space is only relevant on the system volume set, and even then, most relevant on LDEV 1. For user volumes — you do make use of user volumes don’t you? — permanent space should be set to 100 percent. Transient space is usually set to either 0 percent (my preference) or 100 percent - it does not matter since no transient space is ever needed on a user volume set. Now what about the system volume set, and LDEV 1 in particular? The rule of thumb for LDEV 1 has always been 75 percent permanent and 75 percent transient.

For other system volume set members, typically they are configured as 100 percent permanent and 100 percent transient. Now, about the 75 percent/75 percent allocations for LDEV 1: This is to ensure that, first there is always sufficient permanent disk space on LDEV 1 to boot the system; and, second, that there is always sufficient virtual memory space available on LDEV 1 for the system to operate. The 75 percent values were chosen back in the days when DEV 1 was typically less than 1 GB.

Today, most systems have an LDEV 1 that has at least 4GB of usable space, so you can safely configure values greater than 75 percent. However, with disk costs relatively low these days, don’t push the envelope on DEV 1 too far – the cost could be an unusable system. [Note that in VOLUTIL, you can change the allocation percentages dynamically; i.e., a reboot is not necessary.]

To wait for “SHUT 6” or NOT, that is the question

It turns out there is a bug in Transaction Manager that only reveals itself if you have written to KSAM64 files (which you may be using without realizing it, if you have third-party indexing tools) and then you shutdown the system without reaching “SHUT 6”. [Note: there is a patch, MPEMXTOA, which appears to fix the problem.] Of course, a System Abort is a shutdown without reaching “SHUT 6”, but when shutting down the system on purpose, how important is it generally to wait for “SHUT 6” and are there ways to shorten the time it takes to get to a “SHUT 6”? The posting of this bug to 3000-L started a discussion about “SHUT 6”.

Before you get to “SHUT 6”, you have to pass “SHUT 4”. Gary Jackson reminded posters that running STMSHUT.DIAG.SYS before taking your network/system down will help you speed through the “SHUT 4” on your way to “SHUT 6”. If you are using Mirrored Disk/iX it is critical that you reach “SHUT 6”, or have done a VSCLOSE on all mirrored volume sets before shutdown, or else your disks will go out of sync and will have to re-sync on boot, a possibly lengthy process and performance hit.

Closing the volume set (VSCLOSE) eliminates the need for XM recovery and guarantees that the mirrors are marked as being in sync. If your user data is on user volumes and you VSCLOSE those volumes prior to shutdown, you can probably afford to be less patient about waiting for “SHUT 6”.

The recent addition of the MPE/iX SHUTDOWN command allows for the scripting of a planned shutdown process. (Of course, there is a shutdown UDC already written, available at jazz.external.hp.com, that you can adapt to your site.)

What the blank is a MIB? And should I care?

No, it does not stand for Men In Black, at least not in this case. MIB stands for Management Information Base and is used by SNMP/XL (see the HP SNMP/XL Users Guide at docs.hp.com/en/36922-90036/36922-90036.pdf — heck, it is only 11 years old, so we know it has all the latest information). The subject came up this month in relation to a question about the network query tool “What’s Up Gold.” It appears many people are unaware that MPE/iX can respond to many SNMP (Simple Network Management Protocol) requests, despite the fact HP’s own OpenView product has been doing it for years.

MPE/iX’s SNMP definition file (MIB) is MPEIXMIB.NET.SYS. The contact and location information are in SNMPCONF.NET.SYS (it is your responsibility to supply this information). For example, if I use a network query tool I have available on my network, the information returned about my R&D machine includes:

IP Address: 192.168.1.20

MAC Address: 08-00-09-22-96-F0

SysDescr: HP3000 SERIES 927LX, MPE XL version C.65.03 NS Transport version B.06.05

SysUpTime: 20 days 11 h. 17 m. 48 s.

Active TCP ports: 21, 23, 80

Http: Burke Consulting and Technology Solutions

Interface: Hewlett Packard NIO Lan Interface, partnum: 5181-6138; date code: 3135

Quick Cuts

• As long as there are human beings using computers, there will be I D 10 T errors. One IT Manager reported that a remote site was experiencing strange connectivity problems. Some devices could communicate with the central site, but some could not. It took almost a week to resolve the issue. It turns out the remote site had received a new router and when told to unplug the original one, they simply unplugged it from the external communications line, leaving it powered up and on the network creating havoc. The moral? If you want precise computer communications, you had better have precise human communications.

• Suppose you want to determine the firmware version of disk or tape drives. Running ODE (Offline Diagnostic Environment) will get you the information. Boot to the ISL prompt, enter ODE and then run mapper (or mapper2, depending upon your HP 3000 model). Unfortunately, this does not get you the firmware version for any FWSCSI cards. Having the right firmware version is critical if you plan to attach certain peripherals (e.g., a Model 20 array). As a result of community advocacy, a tool (called FWSCSI) has been written by HP and made freely available (see jazz.external.hp.com/src) to all users that will display the firmware level of all FWSCSI cards on a system – all without having to shut down the system.

• If you’ve been looking for the script files that Tim Ericson maintained, it seems that they were moved to a server that is case-sensitive from one that was not and many links are broken. For the time being, thanks to Donna Garverick, the scripts, without all the extra niceties, are available on invent3k at invent3k.external.hp.com/~MGR.GARVERIC/xeqs/.

• If you are used to using DSCOPY to copy files between HP 3000s and now want to copy bytestream files, you need to either use FTP directly between systems to move the files, or, if you do not want to use ftp directly on the bytestream files, STORE the files to disk, DSCOPY (or FTP) the resulting file, then RESTORE the files on the target system. The STORE to disk, copy, and RESTORE trick also works well for NM spool files. Note: depending upon your version of FTP, it may be broken for transferring spool files.

• DBCLOSE mode 3 is a no-cost rewind of a dataset, allowing you to re-read the dataset serially, without having to really close and then re-open the database.

• The www.HP3000links.com Web site is back (thanks to John Dunlop and an anonymous HP 3000 supporter). This is an excellent source of information on the HP 3000 and should definitely be bookmarked in your browser.