Re: [Oscar-devel] New and amusing errors!

[DongInn K. <di...@os...>]

Hi eric,

Thanks for your patience and all your effort.

I am not sure if this is the case but.. since you have said that you should use the "Updates" repo too, I doubt that your F8 machine may not be on a fresh installation status.

If it is, you may have to deal with all the dependencies on your own way that we have not handled.

For the IDE HDD, like I have mentioned on the devel list before, you should use "UYOK" to deploy your image to the client nodes on Fedora 7 / 8. Have you?
http://www.mail-archive.com/osc...@li.../msg08450.html

BTW, for the test machine, I use the real machine whose images are saved on my image server. So, whenever I mess up with my test machine, I can easily go back to the fresh / almost perfectly optimized installation status where I start to test OSCAR again.

If this helps to the OSCAR community, I will post a simple instruction to the OSCAR wiki page.

Regards,

- DongInn


eric gisse wrote:
> 
> 
> On Feb 11, 2008 5:18 AM, DongInn Kim <di...@os... 
> <mailto:di...@os...>> wrote:
> 
>     Hi guys,
> 
>     Well, my fedora-8-i386.url is
>      > [donginn@oscar-f7 ~]$ cat fedora-8-i386.url
>      > file:/tftpboot/distro/fedora-8-i386
>      >
>     http://ftp.ussg.iu.edu/linux/fedora/linux/releases/8/Everything/i386/os
> 
>     As you can see, mine is not the dvd one but everything containing
>     all the rpms (DVD + updates + etc). 
> 
> 
>     Yes, I confirm that the DVD repo and DVD does not have all the rpms
>     that we need and we should use the "Everything" repository.
> 
> 
> This took me a long time to figure out. It was not clear that a) I 
> needed to do this at all and b) that /installing/ the missing packages 
> was not good enough *and* if you did that, the image would be created 
> but would never work. The documentation needs some Fedora 8-specific 
> additions. Which I will try to suggest/add when I get this cluster 
> operational.
> 
> For some reason I still can't discern, OSCAR is incapable of using yum 
> to pull RPM's right off the server like I can do through "yum install" @ 
> the command prompt. Plus OSCAR won't cache the ~250mb of RPM's it pulls 
> off the remote repositories, and refuses to acknowledge them when I 
> directly incorporate them into a local yum repository *and* reference it 
> in the advanced repository. This stuff needs mentioning.
> 
>  
> 
> 
>     I have never tested OSCAR 5.1 b1 or branch-5-1 on VMware and mine is
>     just a normal P4 machine.
> 
> 
> As are my expected server and client nodes. VMware is my way of saying 
> I'm sick and tired of messing with the PCs until I get this sorted out. 
> OSCAR behaves, in my estimation, counter-intuitively sometimes.
>  
> 
> 
>     The VMWare behavior may cause some problems that we have not
>     expected but I have never encountered eric's problems on my physical
>     machines.
> 
> 
> Agreed. I think my current imaging problems I'm seeing on VMware are 
> VMware specific due to the way VMware does harddrives. If I had an extra 
> drive I could directly access with the image, I'm sure I'd be OK. But I 
> don't. I'm going to be either imaging directly off a VMware headnode or 
> a server with OSCAR installed the way I have it installed on VMware 
> later today.
> 
> Yes, I know that FC5 is the real "supported" version but I do not trust 
> it to work reliably on all my machines since I have a mix of IDE and 
> SCSI drives. I don't seem to have that many other options as far as 
> either clustering solutions [ROCKS is too demanding] or other OS' are 
> concerned [Redhat == Fedora, everything else = x86_64 or ppc].
> 
> 
> 
>     So eric, can you please try to play with OSCAR 5.1 b1 on an actual
>     machine when you have a chance?
>     Please post again if you still have the same problems.
> 
> 
> I did, but the responses were less than helpful.
> 
> My current theory is that /Everything isn't good enough, that I need to 
> have the /updates directory as well. Which is what I used as I built the 
> images on VMware. I'm going to use VMware to act as the server node 
> until I'm satisfied I have the kinks worked out. This only because I'm 
> sick of reinstalling Fedora and OSCAR to ensure none of my fuckups are 
> transitioning between attempts.
> 
> It must be the case that I'm missing something simple since so many 
> people seem to be making FC8 work on newer hardware & OSCAR 5.1b1. The 
> stuff I'm using is ~6 years old, but still perfectly functional. P4's 
> and P3's, with 1 possibly 2 Athlon's thrown in.
> 
> I spent most of my time trying to figure out why installing packages 
> directly wasn't good enough. The last week has had me fiddling with 
> online repositories and figuring out PXE & dhcp since I have used 
> neither before.
> 
> My last attempt on the P4 head node was as it is now, except I wasn't 
> using the updates directory to help build the image. The image built 
> correctly, PXE & dhcp played ball, and the image downloaded. Except the 
> image would not install on the harddrive for some reason. Yes, I know a 
> log would be helpful but I can't copy and paste out of the terminal 
> window for some goddamn reason and I can't remember what the error was. 
> I'm sure I'll see it again.
> 
> The only other theory I have is that the dells I have are pieces of shit 
> and that FC8, even though it installs w/o incident on them, somehow 
> chokes anyway. I'll try throwing it on an old non-dell P3 or AMD Athlon 
> system to check that out.
> 
> [...]
> 
> 
> ------------------------------------------------------------------------
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Oscar-devel mailing list
> Osc...@li...
> https://lists.sourceforge.net/lists/listinfo/oscar-devel