[12:02] <neuralis> well, it's probably a month's work for a very talented developer
[12:03] <neuralis> so if someone spends an entire summer on it, he should have a comfortable margin for error
[12:03] <pygi> yup, we've already talked about it
[12:04] <neuralis> yeah, time to take a nap. take care.
[12:06] <pygi> night neuralis
[12:08] <Plug> What's n-a?
[12:08] <lionelp> network authentication
[12:11] <J_P> neuralis: I am not have much experience with Linux cluster, but I am studyng to have :-)
[12:12] <J_P> neuralis: well, so you don't recommend use ubuntu with openmosix beocuse need 2.4 right ? So is better use openmosix + 2.4 in sarge right ?
[12:16] <infinity> J_P: dapper+2.4 will work fine on i386.  For other arches, it won't work at all.
[12:17] <infinity> J_P: If your whole cluster is i386, go ahead and use dapper... Keep in mind that you need modutils (not installed by default) to use modular 2.4 kernels...
[12:19] <J_P> infinity: hummm well, my cluster will be amd k7, p4 and amd64 ...
[12:20] <infinity> Right, the amd64 machines won't run 2.4 on dapper, at all.
[12:21] <infinity> Though, I suspect they may not run it on sarge either.
[12:21] <infinity> Unless you intend to use 32-bit kernels on them.
[01:05] <J_P> infinity: so There aren't solution for linux cluster kernel 2.6  ?
[01:05] <infinity> Probably lots (I'm not a cluster guy, I recommend asking maswan), just not likely to be OpenMosix.
[01:30] <J_P> infinity: thanks man!
[02:27] <Plug> I'm intrigued to see a rc.local in Ubuntu Server
[02:27] <Plug> that was a useful redhatism that Debian lost
[02:29] <ajmitch> afternoon
[02:35] <nictuku> Plug, that will be avaliable in ubuntu 6.06
[02:45] <Plug> yeah, i'm running the preview
[03:51] <J_P> why ubuntu server will have /etc/init.d/rc.local and /etc/rc.local ?
[03:52] <J_P> why not equal as debian ?
[07:28] <fabbione> Plug: it's already there
[07:28] <fabbione> bah these are all FAQ
[07:28] <fabbione> clearly you need to have ini.d/rc.local to run rc.local
[07:58] <Plug> sorry fabbione, i'm intrigued to see that there is one, not "I'd like to see one there"
[07:58] <fabbione> oh ok
[07:58] <fabbione> i misunderstood your wording
[07:59] <ajmitch> must be the accent
[07:59] <fabbione> i blame gkt
[07:59] <fabbione> gtk
[07:59] <fabbione> ajmitch: so... did you put your hands on the T2000 ?
[07:59] <ajmitch> yes, but I still need to take photos for you
[08:00] <Plug> Isn't that a Terminator? :)
[08:00] <ajmitch> 4 of them lined up in the racks
[08:00] <fabbione> ajmitch: eheh neat...
[08:00] <fabbione> ajmitch: did you install Linux???
[08:00] <ajmitch> of course
[08:00] <fabbione> or are you still running slowlaris?
[08:00] <fabbione> eheh
[08:00] <fabbione> cool
[08:00] <ajmitch> I said I got ubuntu on there last wednesday :)
[08:00] <fabbione> oh right
[08:00] <Plug> is the T2000 the one that Sun send you for freE?
[08:01] <ajmitch> and found the installer unbearably slow with the console through ALOM
[08:01] <fabbione> Plug: yeah that kindof box
[08:01] <fabbione> ajmitch: that's an hw design issue
[08:01] <ajmitch> I know
[08:01] <fabbione> ajmitch: already discussed with sun
[08:01] <fabbione> ah ok
[08:02] <fabbione> ajmitch: where are they hosted again?
[08:02] <fabbione> what uni
[08:02] <ajmitch> just with ncurses visibly repainting the whole screen, it was a good exercise in patience
[08:02] <ajmitch> otago uni, NZ
[08:02] <fabbione> ajmitch: ahah yeah
[08:02] <ajmitch> we had linux.conf.au here this year
[08:02] <fabbione> right
[08:03] <ajmitch> just got ubuntu on 1 so far, will get it on another this week
[08:03] <fabbione> did you try raid install?
[08:03] <ajmitch> not hardware raid
[08:03] <fabbione> i didn't even check how to make hw raid
[08:03] <fabbione> that's something to test
[08:04] <fabbione> i mean software raid
[08:04] <ajmitch> currently just using 1 disk, ran into a problem on sw raid in partitioning that I have to report
[08:04] <fabbione> what problem?
[08:04] <fabbione> i know what problem you found
[08:04] <ajmitch> that it didn't let me create a raid partition on the 2nd disk for no obvious reason
[08:04] <fabbione> ahhaha
[08:04] <fabbione> wrtong
[08:04] <fabbione> the problem is that md/lvm overwrites the first 512bytes of the disk
[08:04] <fabbione> that is BAD on sparc
[08:05] <fabbione> so the installer blocks you to create a raid/lvm partition as first partition on the disk
[08:05] <fabbione> it is for your own sake
[08:05] <fabbione> so what you do:
[08:05] <ajmitch> interesting
[08:05] <fabbione> leave a 1MB partition empty at the beginning of the disk
[08:05] <fabbione> on both disks
[08:05] <fabbione> say sdX1
[08:05] <fabbione> sdX2 bacomes a 200MB /boot
[08:06] <fabbione> sdX3 lvm ro wahtever
[08:06] <fabbione> note that /boot can be on raid1 now :)
[08:06] <fabbione> i fixed that at easter
[08:06] <fabbione> :)
[08:06] <fabbione> but the first partition is limited
[08:06] <ajmitch> excellent, thanks :)
[08:06] <fabbione> there is really nothing i can do about it :(
[08:07] <ajmitch> doing the partitioning does take awhile in the installer
[08:07] <fabbione> ajmitch: btw did you figure out how to make hw raid?
[08:07] <ajmitch> probably best if I can just use fdisk
[08:07] <ajmitch> no, I haven't found that out yet
[08:07] <fabbione> ajmitch: well you can.. but it's better if you do it with the partitioner
[08:07] <fabbione> ajmitch:  i will not ensure a mixed install
[08:07] <ajmitch> alright
[08:08] <fabbione> also note..
[08:08] <fabbione> there is an interesting issue with OBP/silo interaction when booting from a raid device
[08:08] <fabbione> i *think* i did fix it
[08:09] <fabbione> basically you might see an error on boot that says that "$random_numer is not elf executable"
[08:09] <fabbione> it looks like tha OBP keeps in it's data structure to boot for diskX
[08:09] <fabbione> when with a rait /boot you need to tell the OBP to boot from diskX:$partition
[08:09] <fabbione> the weird thing is that i got this issue in both direction
[08:10] <fabbione> meaning that i had diskX:$partition and i had to revert to disk
[08:10] <fabbione> it's probably OBP version dependent
[08:10] <fabbione> but i don't have enough OBP's around to make that sure
[08:10] <fabbione> (all of these is going in the release notes.. there is no way i can get it properly fixed for dapper=)
[08:11] <ajmitch> at the moment I'm hampered with the machines with lack of decent net access
[08:15] <ajmitch> what's currently working with the hypervisor on the niagara? I saw something mentioned about xen, but haven't looked into it much
[08:15] <ajmitch> neuralis: nice
[08:16] <neuralis> SunOS gawain 5.10 Generic_118822-20 sun4v sparc SUNW,Sun-Fire-T200
[08:16] <neuralis> ubuntu installation pending, after we're done with solaris benchmarks
[08:23] <fabbione> tsk
[08:24] <fabbione> http://www.stdlib.net/~colmmacc/category/niagara/
[08:24] <fabbione> http://www.stdlib.net/~colmmacc/mv-res.html
[08:24] <fabbione> here are your benchmarks
[08:24] <fabbione> now .. go install ubuntu
[08:25] <ajmitch> heh
[08:25] <ajmitch> tune it, and watch it fly :)
[08:25] <fabbione> there is a reason why the guy did not tune it
[08:26] <fabbione> raw benchmarks out of the box is what most users will get because they do not know how to tune it
[08:53] <neuralis> fabbione: we're having benchmarks done by a guy who wrote his thesis on large-system benchmarking, so i'm curious what his tools will come up with
[08:54] <fabbione> neuralis: fair enough :)