| 1 |
Le 17 Novembre 2005 10:31, Donnie Berkholz a écrit : |
| 2 |
> Eric Thibodeau wrote: |
| 3 |
> > The way I see it, multiple Gentoo roots would be available (different |
| 4 |
> > machine profiles/optimisation) and they would be specified in the DHCP server's |
| 5 |
> > config file as simply as any PXE entry.... So I am not too sure I |
| 6 |
> > understand your "couple of ROOT=/foo" _and_ the busybox one... |
| 7 |
> |
| 8 |
> I just had a single NFS root for all the nodes; didn't have time to get |
| 9 |
> more sophisticated. So the couple of ROOT was one NFS, one busybox. |
| 10 |
|
| 11 |
The way I interpret this is that the nodes actually booted a busybox environment, not Gentoo...am-I getting this wrong? |
| 12 |
|
| 13 |
[...snip...] |
| 14 |
> |
| 15 |
> No reason to chroot, loop mount and all that garbage if you're just |
| 16 |
> doing package management. Just start by unpacking a stage there, and then: |
| 17 |
> |
| 18 |
> ROOT=/path/to/diskless emerge -up world |
| 19 |
|
| 20 |
Ahhh... /me has some polishing up to do about portage (my last read of the documentation dates back 2.5 years ;) |
| 21 |
|
| 22 |
Some other really neat stuff I'd like to implement into the Gentoo-ified PXE/Dynamic cluster (apart from the other obvious stuff like OpenPBS, lam-mpi, etc): |
| 23 |
|
| 24 |
BProc kernel patches http://bproc.sourceforge.net/ |
| 25 |
CheckpointRestart http://ftg.lbl.gov/CheckpointRestart/CheckpointRestart.shtml (for cluster-by-night usage of lab computers at the U ;P ) |
| 26 |
|
| 27 |
Eric Thibodeau |
| 28 |
|
| 29 |
-- |
| 30 |
gentoo-cluster@g.o mailing list |
Archives