jsadler@bcstec.UUCP (Jim Sadler) (09/06/89)
I recently ran into a problem. When we wanted to use NFS to make usenet available to some of the local machines, I found out that HP version of NFS will only export whole file systems only and not sub-directories. Not wanting to publish the root partition we decided to re-partition the /users drive. (By the way this is a 825CHX running HP-UX 3.1 with a 7963B that has all 3 305 Megbyte drives in it.) The /users drive is a single drive of 305 Meg. 2 partitions of 150 Meg each sounded good. So after 2 days of reading manuals to make sure things were done correctly, backing up the users files, etc. First we looked in the /etc/ disktab file and found the following: # @(#) $Revision: 62.2.1.1 $ hp7959|hp7959B|hp7963|hp7963B|hp9263B:\ :ty=winchester:ns#15:nt#12:nc#1572:rm#3600:\ :s0#24280:b0#8192:f0#1024:\ :s1#16384:\ :s2#297108:\ :s3#19532:b3#8192:f3#8192:\ :s4#36864:b4#8192:f4#1024:\ :s5#197504:b5#4096:f5#1024:\ :s6#1998:\ :s7#42902:b7#4096:f7#1024:\ :s8#254204:b8#4096:f8#1024:\ :s9#234548:b9#4096:f9#1024:\ :s12#295028:b12#4096:f12#1024:\ :s13#278586:b13#4096:f13#1024:\ :s14#24280:b14#8192:f14#1024:\ :s15#16384: If you will notice the closest you can come to 150 Meg partition is s5 at 197Mb and then you might be able to combine 3 or 4 other sections for the remaining 108 meg. No problem we think we will just change the disktab table to include 2 partitions of the size that we need. We went to it and ended up with a "Write error at block xxxxx" Scratch our heads for a couple of more days to make sure we didn't miss anything, give up and call the HP response center. The first answer I was given I didn't believe, after verifing the the answer I had to scrape my chin off of the floor. HP hardcoded the disktab table into the kernal !!!!!! FLAMES HOW could you believe that the selection of partitions sizes that you chose would fit what yours customers need !? First you have a lousy choices of sizes, either very big or very small. Then the manuals read that you change the disktab entry to have different size partitions, but then that doesn't work because you put the "allowable" sizes in the kernal. I really don't believe it!! I can't figure out if it was stupidity or lots of gall. You know I keep telling my co-workers that HP builds nice machines, after they hear about these types of gaffes they just say "sure jim" and snicker. FLAME OFF Other than hacking the kernal anyone have any ideas ? By the way the response center said that they would submit an "enhancement request". Jim's translation "HP might fix this in the next decade". jim sadler Unix support, BCS Commercial Airplane Support 206-234-9009 uunet!bcstec!jsadler or maybe jsadler@bcstec.boeing.com
mck@hp-ptp.HP.COM (Doug_McKenzie) (09/07/89)
>anything, give up and call the HP response center. The first answer I was >given I didn't believe, after verifing the the answer I had to scrape my chin >off of the floor. HP hardcoded the disktab table into the kernal !!!!!! >FLAMES >HOW could you believe that the selection of partitions sizes that you chose >would fit what yours customers need !? First you have a lousy choices of >sizes, either very big or very small. Then the manuals read that you change >the disktab entry to have different size partitions, but then that doesn't >work because you put the "allowable" sizes in the kernal. I really don't >believe it!! I can't figure out if it was stupidity or lots of gall. >FLAME OFF I notice two parts to your complaint about disk sections -- that they are not configurable, and that a manual says they are. Please note, I'm NOT arguing with you about the value of hard-coded disk sections, but I'd like to know of any reference where an HP manual says to change an /etc/disktab entry. >You know I keep telling my co-workers that HP builds nice machines, after >they hear about these types of gaffes they just say "sure jim" and snicker. Ouch. Doug McKenzie HP/HP-UX Support mck@hpdstma.hp.com
beth@hpfelg.HP.COM (Beth Crespo) (09/09/89)
>I recently ran into a problem. When we wanted to use NFS to make usenet >available to some of the local machines, I found out that HP version of NFS >will only export whole file systems only and not sub-directories. Not wanting >to publish the root partition we decided to re-partition the /users drive. >... ---------- While it is true that you must export an entire file system, it is possible to mount sub-directories. It is also possible to specify read-only as needed. For example, I have X11 NFS-mounted as four parts out of the "/usr" file system which is exported from the host machine. Good Luck, Beth Crespo
paul@mecazh.UUCP (Paul Breslaw) (09/12/89)
jsadler@bcstec.UUCP (Jim Sadler) writes: >You know I keep telling my co-workers that HP builds nice machines, after >they hear about these types of gaffes they just say "sure jim" and snicker. >Other than hacking the kernal anyone have any ideas ? Pity the poor jerks like us with 9000/3xx machines that do not, and will not, have disk partitions at all. Do you think if we went on and on and on at HP about their stupid disk philosophy and their pathetic 1/4" tape cartridge philosophy, do you think that they would eventually do something about it. I'm the first to look at an idea that may not be 'standard' but is 'better'. An idea that is 'non-standard' and 'worse', you can invent your own fate for. Please HP, WAKE UP TO THESE COMPLAINTS. Paul Breslaw ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Paul Breslaw, Mecasoft SA, | telephone : 41 1 362 2040 Guggachstrasse 10, CH-8057 Zurich, | e-mail : mcvax!cernvax!mecazh!paul Switzerland. |
milburn@me10.lbl.gov (John Milburn) (09/13/89)
In article <358@node17.mecazh.UUCP> paul@mecazh.UUCP (Paul Breslaw) writes: >jsadler@bcstec.UUCP (Jim Sadler) writes: > >>You know I keep telling my co-workers that HP builds nice machines, after >>they hear about these types of gaffes they just say "sure jim" and snicker. A well known feeling, in a predominately Sun/DEC house... >Do you think if we went on and on and on at HP about their >stupid disk philosophy and their pathetic 1/4" tape cartridge >philosophy, do you think that they would eventually do something >about it. [...] This issues are. 1). I shouldn't have to export my root file system whenever I export any directory. I suppose this is my fault for buying large hp discs at outrageous prices :-) 2). I don't like the idea that any bozo user can fill up my entire file system by writing garbage in an otherwise innocuous directory. >Please HP, WAKE UP TO THESE COMPLAINTS. Or at least respond.... -jem John Milburn - Advanced Light Source - Lawrence Berkeley Laboratory INTERNET: JEMilburn@lbl.gov BITNET: JEMilburn@LBL.bitnet UUCP: {...}!ucbvax!lbl.gov!JEMilburn SnailMail: 1 Cyclotron Road 46-161 Berkeley, Ca. 94720 Ph: (415) 486-6969
hesh@lll-crg.llnl.gov (Chris Steinbroner) (09/13/89)
> > Please HP, WAKE UP TO THESE COMPLAINTS. > Or at least respond.... comp.sys.hp is not an HP sponsored news group; far from it: any response you see here from HP personel is an unofficial response. the bottom line is that you can't expect a single thing from this news group. (fortunately, though, HP employees respond with advice just like any other net reader can.) if you have a problem, you should contact your HP representative. i believe they are called "field engineers." that is the official channel. you CAN expect response from your FE; you should even DEMAND response if you have problems. -- hesh
burzio@mmlai.UUCP (Anthony Burzio) (09/14/89)
In article <3786@helios.ee.lbl.gov>, milburn@me10.lbl.gov (John Milburn) writes: > >Please HP, WAKE UP TO THESE COMPLAINTS. > Or at least respond.... They did! They bought Apollo (nice software)! ********************************************************************* Tony Burzio * Watch out for that tree! Martin Marietta Labs * - George song mmlai!burzio@uunet.uu.net * *********************************************************************
rer@hpfcdc.HP.COM (Rob Robason) (09/14/89)
> >Please HP, WAKE UP TO THESE COMPLAINTS. > Or at least respond.... Not wanting you to think that your concerns are falling on deaf ears: HP is aware of your needs and has been for some time. We are currently looking at solutions to the issues you bring up here, but I'm not in a position to promise you anything at this point. Neither is this a hint that anything is in the works. I can tell you that complaining to HP *DOES* help your cause. This discussion group gets pretty wide readership in the lab at HP, though we are reserved about responding because we really can't make commitments to customers. If you want to give input and get commitments about specific features, your best bet is to dog your HP sales rep. She will be able to speak the magic language of "XYZ feature is worth this many $M in sales revenue" to our marketing folks in division headquarters. This in turn results in an "Add XYZ to HP-UX in the N.M release" order to the lab. We in the lab love it when the customer base gives this kind of input, because it usually aligns well with our own needs internally! I hope this insight gives you some idea of how your input can influence HP's products. Rob Robason (Speaking strictly for myself, not HP)
burzio@mmlai.UUCP (Anthony Burzio) (09/15/89)
In article <32970@lll-winken.LLNL.GOV>, hesh@lll-crg.llnl.gov (Chris Steinbroner) writes: > > > Please HP, WAKE UP TO THESE COMPLAINTS. > > Or at least respond.... > > comp.sys.hp is not an HP sponsored news group; > far from it: any response you see here from > HP personel is an unofficial response. the > bottom line is that you can't expect a single > thing from this news group. (fortunately, > though, HP employees respond with advice just > like any other net reader can.) For those on software support, you can send an SPR over the HP bulletin board. I tried it out and it looks real nice. ********************************************************************* Tony Burzio * Oh no, not again... Martin Marietta Labs * - The flower pot mmlai!burzio@uunet.uu.net * *********************************************************************
rjn@hpfcso.HP.COM (Bob Niland) (09/15/89)
re: "Do you think if we went on and on and on at HP about their stupid disk
philosophy and their pathetic 1/4" tape cartridge philosophy, do you
think that they would eventually do something about it."
Quite possibly.
HP's "pathetic 1/4" tape cartridge philosophy" consisted of getting into the
cartridge tape game about 5 years before QIC, back when there were no
obvious standards other than the one HP chose, the 3M "HCD" format. HCD has
some unique advantages (random access, update-in-place, ECC) that are too
widely used to abandon.
HCD and QIC are different enough that no one has ever built a drive that
handles both, nor do I think it possible to build one economically. HCD
and QIC drives can also mis-rewind (or even unspool) each other's tapes.
QIC drives can destroy HCD factory pre-formatting. Until the Apollo
acquistion, HP tried to keep as much distance as possible between HP and
QIC, lest HP customer risk media accidents.
However, "Standard is better than better" and "the customers have all the
votes". It is likely that for at least write/read, HP will support both a
built-in and external SCSI QIC drive in a future release (they can be made
to work today). There is also some chance of supporting boot from QIC.
Speaking for myself and not.......................... Hewlett-Packard
Bob Niland 3404 East Harmony Road
ARPA: rjn%hpfcrjn@hplabs.HP.COM Fort Collins
UUCP: [hplabs|hpu*!hpfcse]!hpfcla!rjn CO 80525-9599taylor@limbo.Intuitive.Com (Dave Taylor) (09/16/89)
Paul Breslaw responds to a note from Jim Sadler (Hi Jim!) with: > Do you think if we went on and on and on at HP about their > stupid disk philosophy and their pathetic 1/4" tape cartridge > philosophy, do you think that they would eventually do something > about it. I think it's worth pointing out that, for example, Hewlett-Packard was very involved with the initial development of the R-DAT data storage mechanism, especially as related to computer usage, as well as with WORM and other optical media too. It would not be surprising to find that HP is poised and ready to ship many interesting and exciting new media for their computers, including CD-ROM, for the 9000 series (300 and 800), DAT backup units (that, I believe, have been floating around Bristol Peripherals Division in England for at least a year now), WORM systems, and so on. Further, with the current migration from slow HP-IB to the much faster SCSI interface, I also expect that HP will allow you much more freedom to configure your disk/etc in any way you'd like. Further, you'll be able to buy ANY SCSI disk and plug it in (I hope!!) [besides, you can get most of these, including WORM units and 8mm backup systems, from third parties TODAY] As far as 800 disk partitioning, I too would have to place my vote with Jim that hard coded partitioning tables is a pretty poor solution to the problem -- I mean, even MS-DOS has a better solution than that! The whole idea of partitioning is to allow the end user site MORE ability to slice their disk up into 'virtual disks' with hard byte limits after all... 300 disk partitioning is something that I can see would also be of possible value, but I am not aware that there are any plans afoot within the company to work on that particular problem. (err, enhancement request) Now, for my own two cents worth, what I would like to see is for HP to offer a 'binary compatability mode' feature similar to the one that allows MPE-V programs to be run under MPE-XL. In the 9000 universe it would mean that I could take a binary that was generated on a 300 machine, say, and run it in "300 compatability mode" on the 800 machine. It would probably have a 5-10% performance hit, but that's cool. Think of the massively improved migration path (especially when the 800 is so much more expensive!). Of course, 800 compatability on the 300 series would be required too. Anyone else think that'd be a useful feature? (we've not even started talking about a PORT/BSD type application either, which would aid in the porting of sofware from a 4.3 BSD environment into the HP-UX environment, partially due to the muddying of the waters that OSF/1 is going to have...!) Keeping a close eye on things, -- Dave Taylor Intuitive Systems Unix Editor Mountain View, California "The HP Chronicle" taylor@limbo.intuitive.com or taylor%limbo.uucp@decwrl.dec.com {uunet!} { decwrl, apple } !limbo!taylor
burzio@mmlai.UUCP (Anthony Burzio) (09/17/89)
In article <5570291@hpfcdc.HP.COM>, rer@hpfcdc.HP.COM (Rob Robason) writes: > > >Please HP, WAKE UP TO THESE COMPLAINTS. > > Or at least respond.... > > If you want to give input and get commitments about specific features, > your best bet is to dog your HP sales rep. She will be able to speak > the magic language of "XYZ feature is worth this many $M in sales > revenue" to our marketing folks in division headquarters. This in turn > results in an "Add XYZ to HP-UX in the N.M release" order to the lab. > We in the lab love it when the customer base gives this kind of input, > because it usually aligns well with our own needs internally! Here's one I yap at my sales rep all the time: I have found from talking to our local AI group and people at the University of Maryland that the inability to put local user files on a "discless" node leads to the reaction "Oh. What a shame, SUN can do it" right before they walk away. Any leads on when HP "discless" nodes may become a cluster with user disks and printers on every node? While on the subject of discless nodes, who was the genious who restricted the 834 to 2 users? I don't know about you all, but if the 834 comes up as a discless node it inherits the licenses from the parent. Doesn't this bypass the 2 user restriction? Just curious, I'd NEVER do this you see :-) ********************************************************************* Tony Burzio * Hi ho, Hi ho, it's off to hack we go!! Martin Marietta Labs * mmlai!burzio@uunet.uu.net * *********************************************************************
mike@fionn.lbl.gov (Michael Helm) (09/18/89)
In article <119@limbo.Intuitive.Com> taylor@limbo.Intuitive.Com (Dave Taylor) writes: >.... Further, with the current migration from slow HP-IB to the >much faster SCSI interface, I also expect that HP will allow you >much more freedom to configure your disk/etc in any way you'd like. >Further, you'll be able to buy ANY SCSI disk and plug it in (I hope!!) We've been using 600 MB WREN-V's on HPUX series 300 machines w/ the SCSI adapter for about 4 months now; a very nice combination. We did have some trouble initially with the unusual connector[1] HP uses on its SCSI board, but once that was overcome everything's spun along just fine. I haven't tried any other 3rd party stuff yet. ----- [1] Just unusual for our environment. HP uses a centronics-style cable, & in our Sun-oriented environment we were used to Sun's DB-50 style SCSI connectors. [I hope you can figure out what I mean, that's not the correct terminology for these connectors.] But I think both connector styles are kosher for the SCSI bus. Compatible centronics connectors were a bother to find, however. ----- >I am not aware that there are any plans afoot >within the company to work on [300 disk partitioning] too bad >Now, for my own two cents worth, what I would like to see is for >HP to offer a 'binary compatability mode' feature ... [in the] >9000 universe it would mean that I could take a binary that was >generated on a 300 machine, say, and run it in "300 compatability >mode" on the 800 machine. It would probably have a 5-10% performance >hit, but that's cool. Doesn't / didn't Apollo have some feature like that, you could manufacture a mixed-architecture executable? I don't think this is based on emulation, but rather had different instruction set- based executables rolled up together. I wonder if you could get as little as 5-10% performance hit from an emulator. (5-10% less than what? Probably not a 370!) Is the "Precision Architecture" really fast enuf to do this? Michael Helm (my opinions only)
mjs@hpfcso.HP.COM (Marc Sabatella) (09/18/89)
>Now, for my own two cents worth, what I would like to see is for >HP to offer a 'binary compatability mode' feature similar to the >one that allows MPE-V programs to be run under MPE-XL. In the >9000 universe it would mean that I could take a binary that was >generated on a 300 machine, say, and run it in "300 compatability >mode" on the 800 machine. It would probably have a 5-10% performance >hit, but that's cool. Think of the massively improved migration >path (especially when the 800 is so much more expensive!). Of >course, 800 compatability on the 300 series would be required too. Incredibly unlikely. This would require an entire virtual machine to be written to emulate the 68K on the HP-PA, which was not designed with this purpose in mind (it was designed with the 3000 in mind, hence the easier migration path on the MPE side). Such an effort would be almost guaranteed to have significantly greater than a 5-10% performance hit, and in any case, would be a gigantic design and development effort (unlike disk partitioning, would seems to me to be just a relatively quick hack).
ckw@hpcupt1.HP.COM (Chick Webb) (09/19/89)
>Here's one I yap at my sales rep all the time: I have found from >talking to our local AI group and people at the University of Maryland >that the inability to put local user files on a "discless" node leads >to the reaction "Oh. What a shame, SUN can do it" right before they >walk away. Any leads on when HP "discless" nodes may become a cluster >with user disks and printers on every node? Not at the HP-UX 7.0 Release, which was just announced. Sorry. Local disks and printers are supported in raw (vs. character) mode, though. >While on the subject of discless nodes, who was the genious who restricted >the 834 to 2 users? I don't know about you all, but if the 834 comes up >as a discless node it inherits the licenses from the parent. Doesn't this >bypass the 2 user restriction? Just curious, I'd NEVER do this you see :-) I don't believe that this is the case. The "uname" structure is initialized at boot time, and even if the 834 is running as a diskless node, it will have its own kernel, which will correctly identify the hardware as an 834, thus filling in the uname structure correctly. This will preserve the 2-user restriction. Also, please note that the 2-user restriction applies to logins via login(1), rlogin(1), and telnet(1) only. Access via remsh(1) is unrestricted. >********************************************************************* >Tony Burzio * Hi ho, Hi ho, it's off to hack we go!! >Martin Marietta Labs * >mmlai!burzio@uunet.uu.net * >********************************************************************* Hope this helps. Chick Webb "Common sense is not so common." Hewlett-Packard Company -- Voltaire Cupertino, CA UUCP: {ucbvax, etc.}!hpda!ckw ARPA: ckw@hpda.HP.COM
taylor@limbo.Intuitive.Com (Dave Taylor) (09/19/89)
Marc Sabatella (marc@hpfcso.HP.COM) responds to my note about the possibility of a 300 binary emulation mode for the 800 and vice versa with the following comment: > Incredibly unlikely. This would require an entire virtual machine to be > written to emulate the 68K on the HP-PA... Such an effort would be almost > guaranteed to have significantly greater than a 5-10% performance hit... While I don't for a minute believe it would be easy to accomplish what I'm talking about, the fact is that HP has a fragmented hardware line with two very different architectures, and the reality that many customers in the field are extremely anxious about this and its effect on migration paths, etc. As Don Souza, a highly placed DEC marketing manager, has indicated in an interview a few months ago, DEC recognizes this as a weakness of HP; he showed me slides that demonstrate that their particular product line (DEC) has a much better hardware growth path, for example. The significance isn't that DEC is right or not, but rather that they think it's an important enough issue to put quite a bit of time and money into pushing with industry analysts/journalists. Why? Because, bluntly, customers don't want to have obsoleted hardware and don't want to buy new computers every year or two. Further, they want to be able to gracefully upgrade -- the 950 -> 955 -> 960 upgrade is EXACTLY THE RIGHT SOLUTION to the problem, from a customers point of view; everything is the same, you get a trade in value on the old CPU board, and lo! things are just lots faster. When we look at the fragmentation within the HP/Apollo market, however, we can see a rather shocking jumble, with machines all the way from the now HP-endorsed "SCO Xenix" on Vectras to the 300/800 schism to the dual Apollo lines too. FIVE DIFFERENT UNIX PLATFORMS! That's a lot of confusion for the customer, isn't it? Or look at the problems of dealing with RAM cards when migrating up in the 9000/300 line itself. There are at least three different types of cards required, each incompatible. Not a good way to protect MY investment in hardware... The point I'm trying to make is that I don't see it really as a question of whether it's a hard implementation or not, but rather a point of possible differentation in the marketplace; lots of vendors currently are confusing their customers with RISC and CISC hardware lines, and HP could be first under the wire with complete binary migration capabilities across the line... Remember; if Unix is to continue to be successful, it's going to have to move from being an engineer's operating system to being a commercial viable product environment, and that means that difficulties in the implementation of desired features (like a fast and compact COBOL for the 9000 series that accepts MPE-V and MPE-XL programs untouched) will become more and more irrelevant. If HP doesn't do it, someone else will, and that'll just be another mark against the 9000 series and another hit against them in competitive sales situations... Besides, if it's true that the HP-Precision Architecture line wasn't designed for 68K compatability, well, add that functionality in the new PA-PLUS that is being designed as we speak at IAG and with the help of the new Japanese partners... Solutions, not excuses. Or isn't there "a better way" ? One other quickie: > ... unlike disk partitioning, would seems to me to be just a > relatively quick hack. Yup, and maybe that's why customers in this group and elsewhere are complaining that it isn't a very good partitioning scheme and that it, again, isn't available across the entire HP 9000 series? Hacks are the last thing that a commercial customer is going to be interested in. And with HP, at least in the "good old days", hacks were the last thing that you'd expect... Just my two cents worth, -- Dave Taylor Intuitive Systems Mountain View, California taylor@limbo.intuitive.com or {uunet!}{decwrl,apple}!limbo!taylor
is813cs@pyr.gatech.EDU (Cris Simpson) (09/20/89)
In article <5570291@hpfcdc.HP.COM> rer@hpfcdc.HP.COM (Rob Robason) writes: >to customers. > >If you want to give input and get commitments about specific features, >your best bet is to dog your HP sales rep. [...] Oh, thanks! I can't even get my sales rep to call me back with prices on shipping products I know I want to buy! OF course, the other problem is that she doesn't know anything about the products. Sheesh, cris -- || Gee, do you think it'd help if I plugged in both ends of this cable? || Cris Simpson Computer Engineer VA Rehab R&D Center GATech Atlanta,GA is813cs@pyr.gatech.edu ...!{Almost Anywhere}!gatech!gitpyr!is813cs
milburn@me10.lbl.gov (John Milburn) (09/20/89)
In article <760004@hpfelg.HP.COM> beth@hpfelg.HP.COM (Beth Crespo) writes: >>I recently ran into a problem. When we wanted to use NFS to make usenet >>available to some of the local machines, I found out that HP version of NFS >>will only export whole file systems only and not sub-directories. Not wanting >>to publish the root partition we decided to re-partition the /users drive. >>... >---------- > >While it is true that you must export an entire file system, it >is possible to mount sub-directories. It is also possible to >specify read-only as needed. You are missing the point. I don't want to rely on the person doing the mounting (the nfs client machine) to decide which directory from my machine (the nfs host) he wishes to mount. I want the control to determine what the client has access to. While it is indeed possible to mount directories read-only, it is NOT possible to export read only. This is almost pointless. John Milburn - Advanced Light Source - Lawrence Berkeley Laboratory INTERNET: JEMilburn@lbl.gov BITNET: JEMilburn@LBL.bitnet UUCP: {...}!ucbvax!lbl.gov!JEMilburn SnailMail: 1 Cyclotron Road 46-161 Berkeley, Ca. 94720 Ph: (415) 486-6969