petersor@frith.msu.edu (Rick Peterson) (05/24/91)
I have just been catching up on new from the past couple of weeks. It seems that the primary justification from HP for failing to include useful features in HP-UX is ROI (this argument is explicit in postings w.r.t. disk partitions and terminal support in SAM). HP, what is the cost of the lost sales due to the failure to include features your users want/need? A couple of years back we bought a bunch of HP 9000/3xx workstations instead of Sun's (which we already had a lot of), and a little more recently some 400t/425's. The primary reason for this was the low cost negotiated with HP. Since then we have regreted the decision due to the lack of needed features and difficulty to manage (vs. other UNIX OS's). We are once again looking at purchasing some additional machines (as we often are). This time HP's will be a hard sell whatever the price. The graphics performance is great, and the price/performance of the Snakes sure is tempting, but the lack of needed features in HP-UX and the difficulty to manage makes HP a difficult choice. HP, what will it cost you to loose these and other sales? What will it cost you that students here and elsewhere, tomorrow's customers, are learning from first hand experience the failings of HP-UX (in our labs here, Sun's are almost always in use, often with other people waiting, while at the same time most of our HP's sit idle - sort of reminds you of the dog food taste test commercials :-) )? Rather than trying to justify based on ROI the failure to include features your users need, you might be better off to start listening to your customers. I suspect that we would gladly pay somewhat increased prices to get HP's high performance graphics and the performance of the Snakes IF HP-UX met our needs in terms of features and quality. Rick Peterson | petersor@egr.msu.edu College of Engineering | (517) 355-7435 Michigan State University | The opinions expressed are my own and do not represent the opinions of Michigan State University.
dougd@hpfcdc.HP.COM (Doug Drees) (05/29/91)
Rick Peterson writes: > negotiated with HP. Since then we have regreted the decision due to the > lack of needed features and difficulty to manage (vs. other UNIX OS's). > I know of a lot of shortcomings in the management of HP-UX. But I would still like to hear a little more detail on why you say HP-UX is difficult to manage. ___ ___ / ) /) / ) / / ____ ___ // __ / / __ _ _ _/__/ (_) /_/ (_)_(/_(_/_/)__ _/__/ / (_</_</_/)__ (/ Douglas Drees ------------------------------------+------------------------------------- ARPA: dougd@hpfcla.fc.hp.com | UUCP: {ucbvax,hplabs}!hpfcla!dougd USmail: c/o Hewlett-Packard m/s 99 | 3404 E. Harmony Rd. | "These are, of course, my own views Fort Collins, C0 80525 | and don't necessarily represent HP"
petersor@egr.msu.edu (Rick Peterson) (05/30/91)
In article <5570648@hpfcdc.HP.COM> dougd@hpfcdc.HP.COM (Doug Drees) writes: >Rick Peterson writes: > >> negotiated with HP. Since then we have regreted the decision due to the >> lack of needed features and difficulty to manage (vs. other UNIX OS's). >> > >I know of a lot of shortcomings in the management of HP-UX. But I would still >like to hear a little more detail on why you say HP-UX is difficult to >manage. > Since we haven't kept a careful log of all of the problems that we have run into, this is in some ways a difficult question to answer. If we thought about it for awhile we could come up with quite a list of problems that we have run into in HP-UX. But, without giving it a lot of thought, here is a small list of problems. 1) disk partitions are needed 2) quotas are needed (but comming in HP-UX 8.0 as I understand it) 3) it would be nice to have long file names be the default 4) on a cluster, cnodes don't reboot after a server crash. This is a real problem since we have many cnodes running off five servers. It's a pain to have to run around and reboot each cnode individually, particularly since some of the machines are locked in offices to which we don't have access. 5) mounting disks via NFS from HP servers does not work particularly well. At one point we had most of our user disks (SCSI) on HP servers which caused each of the servers to crash on at least a daily basis. We have now moved these to a SPARC and everything is just fine. 6) can't boot into single user mode - must boot to multi user mode and then shutdown to single user. Not only is it a nuisance, it is a real problem since users who immediately log into the cnodes very annoyed when we immediately shut the system down again. 7) shutdown often doesn't complete correctly. It often just hangs at some point, generally forcing us into a hard crash of the system. 8) even when shutdown does complete correctly, on reboot it complains that the system was not shutdown properly and fsck has to fix things up. 9) SAM 10) it is almost always a battle to install software. There are several reasons for this. First, all HP software comes in update format which forces the system to be shut down to single user mode in all cases. Second, since HP software comes in update format it is impossible to do anything at all different from the standard install. Third, most software, even from HP, requires much work to get it running under HP-UX. Clearly this may not indicate a problem with HP-UX. However, regardless of what software or what vendor, typically the same package, from the same vendor, for a different UNIX platform will install without problem on the other platforms. I tend to think this points to a problem with HP-UX. 11) management of cnodes is difficult. When an upgrade to HP-UX is done all/most cdf's are lost. There is no convenient way to completely undo a cdf once it exists. Changes to the cnode definitions using SAM is difficult. 12) our HP's tend to crash more frequently/easily than our other UNIX systems. Also, when they do crash, they need to be rebooted manually (rather than doing an auto-reboot). 13) can't put a printer off of a cnode. This seems like an obvious thing to do since a server tends to be locked away in a room where users don't have access, while the cnodes tend to be sitting in offices or in labs. 14) no support in HP-UX for the parallel interface. 15) can't change the broadcast address. 16) can't change/look at ARP entries. 17) there seems to be arbitrary (and small) limits throughout. For example, in archive files, the names of the included files are truncated at 14 (or is it 15) characters. This causes some interesting things to happen when the individual files are extracted again. In the LaserROM software, the list of printers to which you can send output is limited to the first 15 printers available. 18) cluster protocols are HP proprietary, instead of using protocols which are more standard. 19) this really isn't a problem with HP-UX, but rather with HP procedures... notification of bugs and patches is poor. It seems that often when installing additional hardware, only after much grief and many calls to the HP response center do we find out that there is a patch available for the problem we are having. This has even been the case when installing additional hardware/software from HP. I hope this is of help. Rick Peterson petersor@egr.msu.edu College of Engineering (517) 355-7435 Michigan State University Opinions expressed above are mine alone and do not represent the opinions of Michigan State University.
dougd@hpfcdc.HP.COM (Doug Drees) (05/31/91)
I appreciate Rick's response. I, of course, cannot promise anything will be fixed (that isn't already - see below), but I will try to get these comments to someone who can do something about it. > 2) quotas are needed (but comming in HP-UX 8.0 as I understand it) Yes. > 7) shutdown often doesn't complete correctly. It often just > hangs at some point, generally forcing us into a hard crash > of the system. I have run into this one myself. > 9) SAM I'd be interested in more info here. The new 8.0 version should be much better. > 11) management of cnodes is difficult. When an upgrade to HP-UX is > done all/most cdf's are lost. There is no convenient way to > completely undo a cdf once it exists. Changes to the cnode > definitions using SAM is difficult. I've never seen this and we do updates all the time (for each of our pre-release internal integration cycles). > 13) can't put a printer off of a cnode. This seems like an obvious > thing to do since a server tends to be locked away in a room > where users don't have access, while the cnodes tend to be > sitting in offices or in labs. This is fixed in 8.0. See comments elsewhere in this notes group. > 14) no support in HP-UX for the parallel interface. Also fixed in 8.0 (and 7.5, I think?) > I hope this is of help. We will look seriously at these comments. > > Rick Peterson petersor@egr.msu.edu > College of Engineering (517) 355-7435 > Michigan State University > > Opinions expressed above are mine alone and do not represent the opinions > of Michigan State University. > ---------- ___ ___ / ) /) / ) / / ____ ___ // __ / / __ _ _ _/__/ (_) /_/ (_)_(/_(_/_/)__ _/__/ / (_</_</_/)__ (/ Douglas Drees ------------------------------------+------------------------------------- ARPA: dougd@hpfcla.fc.hp.com | UUCP: {ucbvax,hplabs}!hpfcla!dougd USmail: c/o Hewlett-Packard m/s 99 | 3404 E. Harmony Rd. | "These are, of course, my own views Fort Collins, C0 80525 | and don't necessarily represent HP"
rjn@hpfcso.FC.HP.COM (Bob Niland) (05/31/91)
re: > petersor@egr.msu.edu (Rick Peterson) writes ...> > here is a small list of problems. > 14) no support in HP-UX for the parallel interface. Fully supported in 8.0. Driver present in 7.05 (patch recommended for documentation). Patch includes bits for 7.0 and 7.03. Contact your support person - they can get the patch from me if they don't already have it. Regards, Hewlett-Packard Bob Niland Internet: rjn@FC.HP.COM 3404 East Harmony Road UUCP: [hplabs|hpfcse]!hpfcrjn!rjn Ft Collins CO 80525-9599
okamoto@hpcc01.HP.COM (Jeff Okamoto) (06/01/91)
petersor@egr.msu.edu (Rick Peterson) writes: > But, without giving it a lot of thought, here is a small list of problems. > 1) disk partitions are needed > 13) can't put a printer off of a cnode. > 16) can't change/look at ARP entries. Added/fixed in 8.0. > 3) it would be nice to have long file names be the default Yes, but newfs will use whatever type your root file system is. Change that, and all new file systems added will be default long. Of course, SAM will ask you what kind of file system. > 7) shutdown often doesn't complete correctly. It often just > hangs at some point, generally forcing us into a hard crash > of the system. Sometimes the shutdown script will have problems in the call to cwall to notify all cluster clients of the shutdown. You should be able to interrupt the script (with your interrupt character) and simply call "shutdown 0". > 10) it is almost always a battle to install software. There are > several reasons for this. First, all HP software comes in > update format which forces the system to be shut down to > single user mode in all cases. "In all cases" is a bit of an exaggeration(sp?). Any software that must add drivers to the kernel or alters kernel parameters will force a reboot and should be run in single user mode. But most software does not need to be installed in single user mode. > Second, since HP software comes in update format it is impossible > to do anything at all different from the standard install. Your point is valid, but what did you have in mind, doing something different? > 12) our HP's tend to crash more frequently/easily than our other > UNIX systems. Also, when they do crash, they need to be > rebooted manually (rather than doing an auto-reboot). With what kind of error do your servers panic? > 14) no support in HP-UX for the parallel interface. This is untrue. We have several printers running off parallel interfaces. > 15) can't change the broadcast address. This might be fixed in 8.0. > 17) there seems to be arbitrary (and small) limits throughout. For > example, in archive files, the names of the included files are > truncated at 14 (or is it 15) characters. In the case of ar(1), this is necessary to maintain SVID compliance. Why would HP want me as a spokesman? -- \ oo The New Number Who, \____|\mm Jeff Okamoto //_//\ \_\ HP Corporate Computing & Services /K-9/ \/_/ okamoto@ranma.corp.hp.com /___/_____\ ----------- (415) 857-6236
schales@photon.cs.tamu.edu (Douglas Lee Schales) (06/01/91)
In article <5570652@hpfcdc.HP.COM> dougd@hpfcdc.HP.COM (Doug Drees) writes: I appreciate Rick's response. I, of course, cannot promise anything will be fixed (that isn't already - see below), but I will try to get these comments to someone who can do something about it. Also fixed in 8.0 (and 7.5, I think?) Is a newer version of NFS ever going to be packaged with these machines? We had one of the Snakes here as a demo and according to the Sales Rep, it was running 8.0. The NFS was the same old *EXTREMELY* insecure NFS that has been packaged previously with HP-UX. Any plans for an upgrade to newer stuff? Doug. ----- +---------------------------+ | Douglas Lee Schales | | schales@cs.tamu.edu | | Dept. of Computer Science | | Texas A&M University | +---------------------------+ -- +---------------------------+ | Douglas Lee Schales | | schales@cs.tamu.edu | | Dept. of Computer Science | | Texas A&M University | +---------------------------+
mark@comp.vuw.ac.nz (Mark Davies) (06/01/91)
petersor@egr.msu.edu (Rick Peterson) writes: > Second, since HP software comes in update format it is impossible > to do anything at all different from the standard install. Have you actually looked at what update format is? At least since 7.0 (and from memory earlier on the 800's) it is just a tar file with a standard file structure. I often just go in and tar chunks off without having to get near update. cheers mark
alek@spatial.com ( Alek O. Komarnitsky ) (06/02/91)
In article <1670018@hpcc01.HP.COM> okamoto@hpcc01.HP.COM (Jeff Okamoto) writes: >petersor@egr.msu.edu (Rick Peterson) writes: > >> But, without giving it a lot of thought, here is a small list of problems. >> [much deleted] >> Second, since HP software comes in update format it is impossible >> to do anything at all different from the standard install. > >Your point is valid, but what did you have in mind, doing something different? Another recent posting partly answered this question by stating that the tape is in tar format, so one can strip things off and do with them as one pleases. But does the software itself support non-standard installation locations? The difficulties we run into is supporting multiple versions of the same software - specifically C++. Although one can write shell scripts that redefine the "standard places" for stuff (such as libC.a, cfront, etc.), it's not an all togather elegent solution, and flat-out doesn't seem to work in some situations (end.o [needed for -g] can be found elsewhere with C++2.0, but C++2.1 seems to insist on it being in /usr/CC/lib/end.o). A better (but more difficult for HP) solution would be to built in support for multiple versions, with (perhaps) symlinks to the "default" ones - ala Sun's support of multiple diskless OS's (but it took 'em a while to get there). Remember that us software vendors are between a rock and hard place. We need to support our customers who delay upgrading (and there are a LOT of 'em), but we also want to be able to use the latest and greatest. Alek Komarnitsky 303-449-0649 Software Tools Manager, Spatial Technology, Inc. 2425 55th Street, Bldg A alek@spatial.com Boulder, CO 80301-5704 P.S. And will there be a rdate/date command in HP-UX8.0 that allows one to set the time to the nearest second? versus nearest minute! :-(
jason@hpcndjdz.CND.HP.COM (Jason Zions) (06/03/91)
>Have you actually looked at what update format is? At least since 7.0 (and >from memory earlier on the 800's) it is just a tar file with a standard >file structure. > >I often just go in and tar chunks off without having to get near update. Very risky; there's more in the tar files than just bits. You need to execute the various customize scripts as well, in the correct order. In any event, I suspect HP would be reluctant to support a system which was not installed or updated according to the provided mechanism; we'd have no way of knowing how things actually got installed and customized. As for reading and changing the ARP cache, the arp command is included in 8.0. -- This is not an official statement of The Hewlett-Packard Company. No warranty is expressed or implied. The information included herein is not to be construed as a committment on HP's part. The devil made me do it. This won't save me from the lawyers' wrath, but it can't hurt. Jason Zions The Hewlett-Packard Company Colorado Networks Division 3404 E. Harmony Road Mail Stop 102 Ft. Collins, CO 80525 USA jason@cnd.hp.com (303) 229-3800
petersor@egr.msu.edu (Rick Peterson) (06/04/91)
In article <1730097@hpcndjdz.CND.HP.COM> jason@hpcndjdz.CND.HP.COM (Jason Zions) writes: >>Have you actually looked at what update format is? At least since 7.0 (and >>from memory earlier on the 800's) it is just a tar file with a standard >>file structure. >> >>I often just go in and tar chunks off without having to get near update. > >Very risky; there's more in the tar files than just bits. You need to >execute the various customize scripts as well, in the correct order. In any This is the problem that led to the "complaint" about software being distributed in update format (actually the complaint is really that update forces the system to be in single user mode). We tried a couple times quite a while ago to just read the tar files. We only discovered later that some problems we were having were due to the fact that there were "various customize scripts" hidden away that didn't get executed. I have no doubt that using tar to extract the files in some cases works just fine, but it can lead to some real problems. The problem is that you only find out after it is too late about the cases where it doesn't work. Rick Peterson | petersor@egr.msu.edu College of Engineering | (517) 355-7435 Michigan State University | The opinions expressed above are mine alone.
jsadler@misty.boeing.com (Jim Sadler) (06/04/91)
/ misty:comp.sys.hp / okamoto@hpcc01.HP.COM (Jeff Okamoto) / 11:04 am May 31, 1991 / petersor@egr.msu.edu (Rick Peterson) writes: . . . Much deleted. >> Second, since HP software comes in update format it is impossible >> to do anything at all different from the standard install. >Your point is valid, but what did you have in mind, doing something different? How about allowing installs across multiple drives ? >In the case of ar(1), this is necessary to maintain SVID compliance. What about a flag if SVID doesn't allow a command to be a superset. >Why would HP want me as a spokesman? >-- > \ oo The New Number Who, > \____|\mm Jeff Okamoto > //_//\ \_\ HP Corporate Computing & Services > /K-9/ \/_/ okamoto@ranma.corp.hp.com >/___/_____\ >----------- (415) 857-6236 jim sadler 206-234-9009 email uunet!bcstec!jsadler | jsadler@misty.boeing.com This service is brought to you by the computing mafia of Boeing (BCS). Oh ya None of the above is an opinion of The Boeing Co.
mark@comp.vuw.ac.nz (Mark Davies) (06/04/91)
I wrote: |> Have you actually looked at what update format is? At least since 7.0 (and |> from memory earlier on the 800's) it is just a tar file with a standard |> file structure. |> I often just go in and tar chunks off without having to get near update. Jason Zions writes: |> Very risky; there's more in the tar files than just bits. You need to |> execute the various customize scripts as well, in the correct order. In any Rick Peterson writes: |> We tried a couple times quite a while ago to just read the tar files. We |> only discovered later that some problems we were having were due to the |> fact that there were "various customize scripts" hidden away that didn't |> get executed. I would argue that the customize scripts are not hidden away. They are fairly easy to find, and certainly if you were installing something this way you need to look at what they do carefully (but then I would look at what any install did carefully). I mentioned using tar directly mainly for the case when you just want to pull a couple of files off a tape. Certainly I wouldn't recommend doing a full system install that way. cheers mark