| 1 |
I have been running a small group of Gentoo servers for the past year |
| 2 |
(up to 11 now). Initially I used a build server that "emerge sync"d |
| 3 |
daily and built packages (emerge -ub <world|package>). The build |
| 4 |
server's /usr/portage and /usr/portage.local were shared via nfs to the |
| 5 |
other servers. Then the client servers would use "emerge -uk <package>" |
| 6 |
to install. But I ran into some issues with this process since the the |
| 7 |
client servers never ran an "emerge sync", large tbz2s over NFS seemed |
| 8 |
to be non-stable, and some other issues. |
| 9 |
|
| 10 |
So just recently I converted to a multi-stage process. |
| 11 |
|
| 12 |
1) I have a "tree-server" which is just a local rsync mirror of the |
| 13 |
portage tree (and my portage.local tree). |
| 14 |
2) The build server -- which builds the packages (emerge -ub |
| 15 |
<world|package>). |
| 16 |
3) The Package server (currently same server as "tree-server") receives |
| 17 |
the build packages from the build server (rsync), and shares them via |
| 18 |
httpd over a back door to my client servers. |
| 19 |
4) The client servers have their own portage tree (synced to the |
| 20 |
tree-server) and grab the binary packages from the package server via |
| 21 |
"emerge -ug <package>". |
| 22 |
|
| 23 |
The emerge -g is slower than emerge -k (needs to re-check the metadata |
| 24 |
cache every time) but it has worked everytime. Things seem more stable, |
| 25 |
and I can get rid of a bunch of NFS links that were only intermittantly |
| 26 |
needed. |
| 27 |
|
| 28 |
jbw |
| 29 |
|
| 30 |
Jon Kinred wrote: |
| 31 |
|
| 32 |
>> i rsync with the fixed portage tree then use buildpkg |
| 33 |
> |
| 34 |
> That should be usepkg... |
| 35 |
> |
| 36 |
> Jon |
| 37 |
> |
| 38 |
> |
| 39 |
> Jon Kinred wrote: |
| 40 |
> |
| 41 |
>> Hi Andy, |
| 42 |
>> |
| 43 |
>> I too am working on migrating our Red Hat 7.3 servers to Gentoo. I |
| 44 |
>> will outline what i have done to address some of the issues when running |
Archives