1 |
Its been solved in the past ... designed for just this purpose: |
2 |
|
3 |
moriah ~ # esearch http-replicator |
4 |
[ Results for search key : http-replicator ] |
5 |
[ Applications found : 1 ] |
6 |
|
7 |
* net-proxy/http-replicator |
8 |
Latest version available: 3.0-r2 |
9 |
Latest version installed: 3.0-r2 |
10 |
Size of downloaded files: 38 kB |
11 |
Homepage: http://sourceforge.net/projects/http-replicator |
12 |
Description: Proxy cache for Gentoo packages |
13 |
License: GPL-2 |
14 |
|
15 |
|
16 |
moriah ~ # |
17 |
|
18 |
I chain them together (two levels, avoiding expensive download costs) so |
19 |
the remote site doesnt have it in its cache, upstream is the master |
20 |
cache, which if it doesnt have it will fetch from the repo. You can |
21 |
specify what port it runs on, and then use the http_proxy entry in |
22 |
make.conf to point the clients to it thus avoiding port 80 and any |
23 |
existing webserver. Handles concurrent fetches transparently. Overall, I |
24 |
have found it preferable to NFS which has been a bit flaky at times in |
25 |
the past. |
26 |
|
27 |
Recommended! |
28 |
|
29 |
|
30 |
BillK |
31 |
|
32 |
|
33 |
|
34 |
|
35 |
On Sat, 2011-11-12 at 22:01 +0000, Neil Bothwick wrote: |
36 |
> On Sat, 12 Nov 2011 19:40:08 +0700, Pandu Poluan wrote: |
37 |
> |
38 |
> > During my drive home, something hit my brain: why not have the 'master' |
39 |
> > server share the distfiles dir via NFS? |
40 |
> |
41 |
> No reason at all, I've been doing it for years without a single |
42 |
> problem. |
43 |
> |
44 |
> > So, the question now becomes: what's the drawback/benefit of |
45 |
> > NFS-sharing vs HTTP-sharing? The scenario is back-end LAN at the |
46 |
> > office, thus, a trusted network by definition. |
47 |
> |
48 |
> The benefit is that everything is centralised. With an HTTP proxy, you |
49 |
> still have to download from the server to each client. The only drawback |
50 |
> that I experience is that if several packages use the same, large source |
51 |
> file, as so many of the KDE packages do, you are repeatedly pulling the |
52 |
> same file over the network, which is a little slower. |
53 |
> |
54 |
> |