Re: Sparc64 Install Problem
On Thu, 17 Oct 2002 19:01:08 -0700 Andrew Sharp <[EMAIL PROTECTED]> wrote: > > OK, installing anything on an IPX is gonna be painful. But yeah, try > to see if it's as bad as you remember it. ~:^) Just a quick tip, I > think I hear on this list that the best thing for this is just go > straight to the net boot install method, unless you have a cdrom drive > for it. Never say "yes" to anything involving a floppy on the > installation procedure for Debian/Sparc. That's my rule of thumb. Thanks for the tip! It's greatly appreciated!! Yeah, trying to make a boot floppy on the Ultra hung the install enough to require hitting the Big Red Button. Peace! -- Bill Meahan WA8TZG [EMAIL PROTECTED] "Never ascribe to mailce what can be adequately explained by stupidity." -- Robert Heinlein
Re: Sparc64 Install Problem
On Thu, Oct 17, 2002 at 09:10:42PM -0400, Bill Meahan wrote: > On Thu, 17 Oct 2002 16:56:36 -0700 > Andrew Sharp <[EMAIL PROTECTED]> wrote: > > On Thu, Oct 17, 2002 at 05:28:02PM -0400, Bill Meahan wrote: > > [snip] exactly. > > Well, what you are possibly talking about isn't about PC assumptions > > at all, I think. Once the packages are downloaded/copied, and they > > are started to install, some of them have postinstall/configure > > scripts. These are there to help you, and don't really have anything > > Note I put "PC" in quotes. A more general term would be "workstation" > for the point I was aiming at. Example: why force the definition of the > keyboard and mouse when there is no keyboard or mouse? Well, believe it or not, there is almost a good reason for the keyboard thing even if the system doesn't have one, but I'm sure there were other things besides this, as the keyboard question comes and goes quickly and harmlessly. > I don't mean to start any kind of flame-fest! Not at all. And it wasn't taken that way. > What I do want to > accomplish is to sensitize folks to the idea that unconfirmed > assumptions can make life _really_ painful for others and "good > installer" doesn't automatically mean "GUI," it just means it's really > well thought out from a _users_ point of view. That's all. [hp-ux madness gone] > I'm going to be putting woody on a couple of SPARCStations (an IPX and > an LX) next week. I'll try and be really specific if I run into any > problems:-) :-) OK, installing anything on an IPX is gonna be painful. But yeah, try to see if it's as bad as you remember it. ~:^) Just a quick tip, I think I hear on this list that the best thing for this is just go straight to the net boot install method, unless you have a cdrom drive for it. Never say "yes" to anything involving a floppy on the installation procedure for Debian/Sparc. That's my rule of thumb. a
Re: Sparc64 Install Problem
On Thu, 17 Oct 2002 16:56:36 -0700 Andrew Sharp <[EMAIL PROTECTED]> wrote: > On Thu, Oct 17, 2002 at 05:28:02PM -0400, Bill Meahan wrote: > > I take it this is after the part of the install that takes place > before a reboot happens. One should always check silo.conf and such > before rebooting. It usually gets it right, but it doesn't hurt to > check. > Happens quite early in the process IIRC > This is a long and rambling counterpoint to Bill's comparison of > Debian woody install to Solaris 9 install. Delete now if you don't > care. > > [snip] > > Well, what you are possibly talking about isn't about PC assumptions > at all, I think. Once the packages are downloaded/copied, and they > are started to install, some of them have postinstall/configure > scripts. These are there to help you, and don't really have anything Note I put "PC" in quotes. A more general term would be "workstation" for the point I was aiming at. Example: why force the definition of the keyboard and mouse when there is no keyboard or mouse? I don't mean to start any kind of flame-fest! What I do want to accomplish is to sensitize folks to the idea that unconfirmed assumptions can make life _really_ painful for others and "good installer" doesn't automatically mean "GUI," it just means it's really well thought out from a _users_ point of view. That's all. No Debian install (or Solaris for that matter) will ever be as bad as the fiasco I had with an HP-UX machine. That bastard assumed it would be connected to a network (even though the install was from a CD) and by thunder, it would try and do a network query of one sort or another every few steps then wait with a 5-minute timeout before moving on. Even plugging in a network cable didn't help because it was trying to force the connect speed rather than autonegotiate. Oh yeah, because of that, it was impossible to get in and alter anything until well into the install. Literally took a couple of DAYS! Needless to say, HP made some "significant" changes and later editions were much, much better! I'm going to be putting woody on a couple of SPARCStations (an IPX and an LX) next week. I'll try and be really specific if I run into any problems:-) :-) Peace to all! -- Bill Meahan WA8TZG [EMAIL PROTECTED] "Never ascribe to mailce what can be adequately explained by stupidity." -- Robert Heinlein
Re: Sparc64 Install Problem
On Thu, Oct 17, 2002 at 05:28:02PM -0400, Bill Meahan wrote: > On Thu, 17 Oct 2002 12:11:21 -0400 > philo vivero <[EMAIL PROTECTED]> wrote: > > > > > > > Instead, I booted with the "rescue" image which, for whatever reason, > > did a normal install. Here's what I did: > > > > ok> boot cdrom > > . . . blah blah blah . . . > > SILO: rescue > > . . . normal installation occurs here . . . > > > > > 5. The computer begins to boot, I see the normal Linux screen as it > > > boots. > > > 6. During boot a kernel panic occurs... > > > > > > Kernel panic: I have no root and I want to scream > > > Press L1-A to return to the boot prom I take it this is after the part of the install that takes place before a reboot happens. One should always check silo.conf and such before rebooting. It usually gets it right, but it doesn't hurt to check. > I had exactly the same experience, also on an Ultra-1. > > Once installed via the "rescue" path, everything was fine. Of course, so > much of the install just _assumes_ you're on a "PC"-like device with a > video card and attached keyboard and mouse. My Ultra-1, however, was > headless and I was using a serial console. I was able to find > workarounds to the assumptions but that made it a lot more painful than > it needed to be. Took me about 6 hours to get everything installed. This is a long and rambling counterpoint to Bill's comparison of Debian woody install to Solaris 9 install. Delete now if you don't care. I haven't seen an u1 in a long while, and I haven't done an install of Debian 3.0 for a month or so, but when I did, it was on a couple of CP1500s (a USII based cPCI system popular with telco's) which has video/keyboard/mouse hardware at all. In fact, it has two serial ports, and HME port and a scsi port, that's all. Prior to that, I did an install on my ultraEnterprise2 at home, and I have no Sun console stuff (no keybds or monitors or whathaveyou), just an x86 computer, a serial cable, and microcom (pathetically simple terminal emulator). Installs on all three machines went quite smoothly, and I didn't notice anything bad that was assuming that I was on a PC with a console. I didn't have the dreaded "can't find vmlinuz64" problem, but maybe that's a more recent change. Anyway, I agree with your [possibly deleted] notes on how hard writing an install program is. I've done it. It's like hell. Every bastard with a pencil neck wants progress meters, ways to do special weird things that 99.99% of the customers won't ever want to do, you name it. It's quite thankless and a butt load of work. Just to get it to work I had to find and fix a bug in ksh (this was in the 80's -- don't ask). > For comparison, I loaded Solaris 9 on the same machine and everything > was smooth as silk. Very few "assumptions", good support (via > "suninstall") for a serial console. Took me about an hour since once > set-up for the install, the only required interaction was changing the > CD's and I could walk away and do something else. Wow, I installed Sol9 on one of the aforementioned CP1500s and it was THE biggest pain. Not only did it suck, but it sucked worse than installing sol8 with a serial console! That's impressive. Gone was any attempt to do cursor addressing, and instead everything was done with the moral equivalent of more(1) instead! Yikes! It was ... words fail me. Yes, it's true, you no longer have to hit the F2 key a million times, but holly scrolling 9600 baud terminal, batman, having to go through I-forget-how-many hundreds of packages by scrolling and re-scrolling and aagh. It was horrible. Took me hours. Imagine dselect, but w/o cursor addressing: every time it wants to redraw the screen, it would just scroll a screen full of lines up from the bottom. At 9600 baud. You'd go completely mad after about an hour. But at least sol9 comes with bash. ~:^) Bottom line, AFAIC, the Debian 3.0 install was head-and-shoulders better and easier than either the sol8 or sol9 installs. > I understand _where_ the assumptions come from (given the history of > Linux) and I'm not really complaining or faulting anyone but I truly > believe more effort needs to be put into the installation process. No, > I'm most assuredly NOT talking about a whiz-bang, themable, skinable > all-singing, all-dancing GUI, I'm talking about a well-thought-through > procedure that gathers all the necessary information up front so a busy > sysadmin doesn't have to spend all day interacting with the installer. > I'm also talking about CONFIRMING assumptions before they are acted > upon. (Example: does the presence of a video card _really_ mean that the > installer can go ahead and assume a video display or does it just mean > the box came with one and it wasn't worth the trouble to remove it?) Well, what you are possibly talking about isn't about PC assumptions at all, I think. Once the packages are downloaded/copied, and they are started to install, some of them have postinstall/confi
Re: Sparc64 Install Problem
On Thu, 17 Oct 2002 12:11:21 -0400 philo vivero <[EMAIL PROTECTED]> wrote: > > > Instead, I booted with the "rescue" image which, for whatever reason, > did a normal install. Here's what I did: > > ok> boot cdrom > . . . blah blah blah . . . > SILO: rescue > . . . normal installation occurs here . . . > > > 5. The computer begins to boot, I see the normal Linux screen as it > > boots. > > 6. During boot a kernel panic occurs... > > > > Kernel panic: I have no root and I want to scream > > Press L1-A to return to the boot prom > I had exactly the same experience, also on an Ultra-1. Once installed via the "rescue" path, everything was fine. Of course, so much of the install just _assumes_ you're on a "PC"-like device with a video card and attached keyboard and mouse. My Ultra-1, however, was headless and I was using a serial console. I was able to find workarounds to the assumptions but that made it a lot more painful than it needed to be. Took me about 6 hours to get everything installed. For comparison, I loaded Solaris 9 on the same machine and everything was smooth as silk. Very few "assumptions", good support (via "suninstall") for a serial console. Took me about an hour since once set-up for the install, the only required interaction was changing the CD's and I could walk away and do something else. I understand _where_ the assumptions come from (given the history of Linux) and I'm not really complaining or faulting anyone but I truly believe more effort needs to be put into the installation process. No, I'm most assuredly NOT talking about a whiz-bang, themable, skinable all-singing, all-dancing GUI, I'm talking about a well-thought-through procedure that gathers all the necessary information up front so a busy sysadmin doesn't have to spend all day interacting with the installer. I'm also talking about CONFIRMING assumptions before they are acted upon. (Example: does the presence of a video card _really_ mean that the installer can go ahead and assume a video display or does it just mean the box came with one and it wasn't worth the trouble to remove it?) Writing a good installer, one that helps novices while not getting in the way of experts, is a really big job and is as much an art as a science. But it ain't near as much fun as eye-candy or the 432nd unique text editor! BTW, the Ultra is still running Solaris while my 4 Intel boxes (3 desktops and a laptop) run Woody or Sid. I don't do Windows!! -- Bill Meahan WA8TZG [EMAIL PROTECTED] "Never ascribe to mailce what can be adequately explained by stupidity." -- Robert Heinlein
Re: Sparc64 Install Problem
> I'm trying to install 3.0 on a Sun Ultra2 using CDs created with jigdo. > This is my first experience with Debian. Here is what I've done... I just did this on an Ultra1. > 1. Booted with CD1, got the boot: prompt and hit Enter. > 2. Got an error "/boot/sparc64.gz not found" or something similar. > 3. Looked at the CD contents and found /boot/sparc64 without the gz > suffix. > 4. At the boot: prompt I entered "/boot/sparc64" and hit Enter. Instead, I booted with the "rescue" image which, for whatever reason, did a normal install. Here's what I did: ok> boot cdrom . . . blah blah blah . . . SILO: rescue . . . normal installation occurs here . . . > 5. The computer begins to boot, I see the normal Linux screen as it > boots. > 6. During boot a kernel panic occurs... > > Kernel panic: I have no root and I want to scream > Press L1-A to return to the boot prom I love Linux error messages. A little humour can go a long way. You would have to specify where your root is. With a CD boot, I don't know what it'd be... you specify it like this: SILO: linux root=/dev/XYZ But since you're trying to install, what device you mount as root I can't remember. It'd be the RAMdisk image that is loaded up. But again, like I said, when I chose "rescue" target to boot, that actually got me into the normal installation. -- tim ellis senior database architect