|Click here for Hardware House sponsor message|
|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.
We admit that we like to mix it up on strategic issues, having started this thread ourselves. At issue: is NT a suitable substitute for the 3000s in mission-critical, cant-reboot uses? In an August Computerworld, three stories appeared that cast doubt on NTs ability to scale and become a mission-critical server. Our pointers to the Computerworld articles were suspect in some eyes because they came out of Computerworld, but others noted NT shortfalls of their own. James Trudeau said, I find NT quite well behaved and do not understand the underpinnings of your attack on a really fine OS which has... guess I sort of missed the part where you said mission-critical. Okay, you win.
Defender of the NT crown Denys Beauchemin weighed in with defense of NTs shortcomings such as the reboot frenzies the OS is encountering. The story refers to a user being unable to stop a misbehaving application and having to bring the entire system down, he said. There are several utilities in the NT resource kit that will do this for you, thus precluding a reboot. Beauchemin went on to add that Yes, NT 5.0 is supposed to address many of these reboot issues, because 5.0 is billed as a Plug-n-Play OS. Will it live up to its claim? Probably not completely until Service Pack 2 or 3 for NT 5.0. I have tested 5.0 a while back, on my laptop. The PnP concept was definitely there, but more work was required.
Users with differing NT experience had other views. I have to use a machine with NT 4.0 at a client site, said Cecile Chi, and when it kept dying someone there installed Service Pack 3. Now I have to reboot it (because it hangs) only a couple of times a week. I read the messages people post here about having NT stay up for a month or two at a time, but Ive never been any place where that really happens. It must be nice.
Lost in the discussion was clarity on whether the NT in question ran on a desktop workstation or was used as a server for dozens of users. The majority of the posts seemed to point to desktop use, not server work. One clear posting came from NT expert Joe Geiser: I too, will not defend MS here but I will defend the OS. I have one NT server that has been running since I brought it up, back in March. It has not burped yet, and still hums along quite nicely thank you. A user of Maestro on the NT platform said that Tivoli reported Maestro was much more stable on Unix: This is not intended to be a slam on Tivoli, said Patrick Santucci. Rather, its to make the point once more that NT is not yet enterprise-ready.
It was up to James Clark to flag another tar-pit of managing NT, its use of DLLs: The problem I have with NT is when you change something or add some software you can break the OS, he said. This changing system DLLs for software additions is for the birds. Patches are one thing, and they may work or maybe not. But putting a good software on a good working system which then crashes is poor design and stupid. This has happened to me and continues to happen, making me afraid of touching a working NT system. Traffic on the thread was mixed enough to consider the debate a draw, meaning its not a slam-dunk pro or con about NTs mission-critical abilities unlike those of the HP 3000.
Tracking malfunctions in the 3000 involves using its provided tools provided you know which ones. Operator replies to console messages through system logging can provide answers, and capturing the replies sparked some ideas from the newsgroup. Pete Crosby of the Atlanta HP Response Center said Just enable event type 115 logging in SYSGEN. Then LOGTOOL can be used to extract the records.
A little lower-tech solution came from Cindee Bridges-Fowler, who said At my former job we attached a slow dot matrix printer to our console and log bottom to capture the console messages. Having it has helped us trace numerous problems and reconstruct events that would have otherwise been unexplainable.
Gavin Scott noted that console logging not only gets console messages logged, but some or all of the ordinary terminal input and output that happens on the console you at least used to get :REPLY commands logged as I (input) type console log records. Does anyone know what the criteria is for normal console I/O to go into the logfile? Doug Werth of Beechglen Development replied: On the output side it is all console messages that appear on the logical console. On the input side it is all CONSOLE commands, regardless of whether issued from the logical console or not. This would all REPLYs, ABORTJOB, ABORTIO, everything that is considered a console command.
Werth then added that the log record on the input record captures the session/job info and the LDEV number of the user who issued the command. What it does not capture is every keystroke entered on the console. I would give a dollar for each time someone wanted to trace back a PURGE command that was issued from the console but couldnt because it is not an operator/console command. Michael Hensley resolved that problem, noting that A PURGE is just an FCLOSE with DISP=DELETE. If you log FCLOSEs (105 and 160), you can see who find out who purged the file (userid, ldev), and when.
Chris Bartram asked that when he noted that DLT drives from HP not only are more expensive than DDS-3 counterparts, but have lower Mean Time Between Failure rates. Martin Slover replied We have been using DLTs here for both our HP 3000 and HP 9000s without any real glitch. Now that we have the full implementation of 5.5 and the latest TurboStore updates, the things are pretty rock solid from my experience on the 3000s.
Mark Klein of ORBiT, which makes backup solutions for the 3000 that span the range of media, offered this experience: We have a longer history with DLT than with DDS-3. To my knowledge, weve not seen any failure with either device to date. Well see tapes go bad from excessive use before well lose a drive. I think in all the time that weve had these automated suites running, weve only had a single drive failure, and that was on a DDS-1 drive. Weve never seen a failure with DDS-2 or DDS-3 drives in house. And never with DLT.
The discussion didnt linger on hardware, but went straight to the differences in media between the tape backup devices. John Clogg said, Although HPs numbers seem to indicate a longer MTBF for the DDS-3 drives, I believe the real issue for most of us is the reliability of the media and their contents. Because of DLTs linear format (no helical scan mechanism to get out of alignment), and robust error-correction schemes, the error rate on DLT media is breathtakingly small. If the [DLT] drives need service a bit more frequently, Im willing to pay that price for reliable backups.
A question about how to get an HP 3000 to dial out to a pager with alphanumeric messages about system errors got the same reply in two different sources see Telamon. Randy Medd of the company pointed out that a freeware program for the 3000 is available at the companys FTP site: ftp://ftp.telamon.com/dist/pager.nm a>. The firm also sells its TelAlert solution, of which Guy Smith noted We do paging, but we also send messages to e-mail, voice mail, phones, loudspeakers, and electronic signboards (we can even attach a pager to an electronic signboard and have it receive the text of the message to display). Due to oddities of MPE, we have not ported the TelAlert server to MPE, but we do have the TelAlert client piece available if you want to run the server on NT or Unix. But then, that depends on if you consider paging mission-critical, and how you feel about NT versus Unix, doesnt it?
Copyright 1998, The 3000 NewsWire. All rights reserved.