On Tue, Sep 04, 2007 at 04:15:23AM +0200, Adriaan wrote:
Depends if all your boxes are the same architecture.
In any case, this is more or less sketched out in ports(7) (BULK PACKAGE
BUILDING, in particular).
Roughly sketched:
- there are distinct areas in /usr/ports. There's the main stuff, there are
the distfiles, there are the packages, there are the cookies, and there are
the working directories. You will want to use separate rules for each.
- You probably want to fetch the distfiles on the NFS server itself
(using the mechanism described in mirroring-ports(7)
- keep the work directory local. That's what WRKOBJDIR is for.
- the ports tree has some mechanism to ensure that two thingies do not build
the same package at the same time. With separate workdirs, this is less of
an issue, as the chances of collision are less... but be careful. it's
probably a good idea to lock stuff on machines that share some not read-only
stuff (e.g., packages for the same architecture).
- you might want to put the built packages on a separate partition with
distinct rules: they're all that's needed if you just want to install stuff
on a machine.
- think about what you do with the various update and bulk cookies.
There. The only thing we do not handle at all is shared installation through
NFS. pkg_add really does not cope with machines which share only /usr/local
and not /etc, for instance...
| Heiko Carstens | Re: -mm merge plans for 2.6.23 -- sys_fallocate |
| Greg KH | [GIT PATCH] driver core patches against 2.6.24 |
| Vladislav Bolkhovitin | Re: Integration of SCST in the mainstream Linux kernel |
| Ingo Molnar | Re: suspend2 merge (was Re: [Suspend2-devel] Re: CFS and suspend2: hang in atomic ... |
git: | |
| Sander | 'struct task_struct' has no member named 'mems_allowed' (was: Re: 2.6.20-rc4-mm1) |
| David Miller | [GIT]: Networking |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Natalie Protasevich | [BUG] New Kernel Bugs |
| Gerrit Renker | [PATCH 15/37] dccp: Set per-connection CCIDs via socket options |
| Hans Reiser | Re: silent semantic changes with reiser4 |
| Nick Piggin | [ANNOUNCE] new new aops patchset |
| Christoph Hellwig | Re: [patch 01/15] security: pass path to inode_create |
| johnrobertbanks | Re: Reiser4. BEST FILESYSTEM EVER. |
