1 |
Le 21 Novembre 2005 11:10, Robin H. Johnson a écrit : |
2 |
> On Sun, Nov 20, 2005 at 08:51:13PM -0500, St?phane Lacasse wrote: |
3 |
> [snip discussion about installing] |
4 |
> |
5 |
> I've done the cluster system (128 node+ 1 master) in a similar fashion |
6 |
> to what you are after. |
7 |
> 1. PXE-boot install environment for performing installs of both the |
8 |
> master and all of the nodes. |
9 |
PXE-boot even for the Master?...so where do the images reside...how do you |
10 |
manage the slightly varying config items such as hostname and all? This |
11 |
approach still seems a little bit time consuming since all nodes are still |
12 |
individual entities (not NFS roots to a single maintained image). Though |
13 |
granted that the nodes being all identical, emerge -K should in theory be a |
14 |
breeze....but it's not the case for maintaining all the config files |
15 |
consistent. |
16 |
|
17 |
> 2. The install environment uses the Gentoo Installer, with the CLI |
18 |
> frontend I wrote for the GLI project, and performs complete installs of |
19 |
> nodes in under 20 minutes (depending on network traffic). |
20 |
So switching a machine's purpose/profile requires a complete re-install on the |
21 |
node? You state 20 minutes for re-installing, is it a _real_ install or the |
22 |
dump of a "reference" root? (Pardon my ignorance of the CLI installer you are |
23 |
referring to... I'll read the http link you'll send me ;) ) |
24 |
|
25 |
> By using GLI, it's a simple matter of altering the install profiles to |
26 |
> reconfigure the cluster, and wipe the nodes for changing their purpose |
27 |
> (presently we have an MPI mode and a MOSIX mode), some of the cluster |
28 |
> users need assurances that none of their data remains on the cluster |
29 |
> after they are done, hence being able to reinstall easily. |
30 |
[...] |
31 |
> Also, make use of your cluster tools to administer the cluster. OpenPBS |
32 |
> allows running a job on all nodes, so use it to emerge -K [package]. |
33 |
> (not -k as binpkgs don't currently have any locking in $PKGDIR, and can |
34 |
> get corrupted if two emerge processes try to create a binpkg at the |
35 |
> same time.) |
36 |
|
37 |
Actually, I would have thought you use _one_ node to compile the packages |
38 |
(using distcc at your description) and _then_ propagate the package onto the |
39 |
other nodes with -K....still, I would think maintaining an NFS mounted ROOT |
40 |
would be less cumbersome.... |
41 |
|
42 |
-- |
43 |
Eric Thibodeau |
44 |
|
45 |
-- |
46 |
gentoo-cluster@g.o mailing list |