1 |
On Thu, Sep 15, 2011 at 11:06 AM, Neil Bothwick <neil@××××××××××.uk> wrote: |
2 |
> On Thu, 15 Sep 2011 11:40:32 -0500, Paul Hartman wrote: |
3 |
> |
4 |
>> If fetch-only is done unattended, and there's a really slow mirror, |
5 |
>> the interactive emerge later will still be fast, where if using |
6 |
>> parallel-fetch and you hit a slow download, you just have to wait, or |
7 |
>> bring up the fetch log in another terminal and kill the wget session |
8 |
>> etc. |
9 |
> |
10 |
> Yes, which is one of the reasons I do a fetch-only immediately after a |
11 |
> sync - unattended. But when working interactively, running emerge world |
12 |
> three times, once each for download, pretend and install, only slows |
13 |
> things down. Not just the downloads but also the fact that the dependency |
14 |
> tree has to be calculated three times. |
15 |
|
16 |
I think you have a specific view that is likely the very best thing to |
17 |
do for your situation, what ever that is, be it work, office, server |
18 |
farm. I don't know. I'm guessing however, that in your world machines |
19 |
are always turned on, burning power, and running cron jobs in those |
20 |
environment makes lots of sense. |
21 |
|
22 |
In my world, which is just a lowly home user of Linux for nearly 15 |
23 |
years now, many of the machines I take care of spend more time turned |
24 |
off than on. cron jobs don't work when there's no power applied, and |
25 |
while you can let the machine immediately catch up when the machine is |
26 |
powered back up, in my world of futures trading I need to control CPU |
27 |
and network usage to ensure that both downloads and builds don't |
28 |
impact my opportunity to make a trade and hopefully make some money. |
29 |
As I write this email I'm currently in my 23rd S&P futures trade of |
30 |
the day which at this point is just about 5 hours old. Some of these |
31 |
trades take only a few minutes and likely wouldn't execute correctly |
32 |
if portage was building KDE. |
33 |
|
34 |
Like Paul, I run fetch unattended, sometimes 4 or 5 times over |
35 |
multiple days before running emerge -DuN @world. I don't care at all |
36 |
that the tree is calculated multiple times, and I don't care if I |
37 |
download multiple revisions of the same package but only build the |
38 |
last one. When I do sit down to start the build everything is ready |
39 |
and I can be more assured that the build will completely finish and |
40 |
not leave the machine in some funny state. Additionally, it gives me |
41 |
time to watch this list for problems others are having, if any, before |
42 |
I end up having the same problems. |
43 |
|
44 |
It's a different world we live in I think, and different worlds |
45 |
potentially come up with different best practices. The 23rd trade just |
46 |
completed. The computer is now off looking for the 24th. |
47 |
|
48 |
- Mark |