| 1 |
Am 12.12.2011 09:43, schrieb Alan McKinnon: |
| 2 |
> On Mon, 12 Dec 2011 07:29:16 +0700 |
| 3 |
> Pandu Poluan <pandu@××××××.info> wrote: |
| 4 |
> |
| 5 |
>>>> It's worked for me ever since I switched all of my machines to |
| 6 |
>>>> OpenRC a year+(?) ago. |
| 7 |
>>> |
| 8 |
>>> You are not a representative sample. |
| 9 |
>>> |
| 10 |
>> |
| 11 |
>> worksforme |
| 12 |
>> |
| 13 |
>> In production servers, even. Virtualized on top of XenServer. All of |
| 14 |
>> them last updated last week. |
| 15 |
>> |
| 16 |
> |
| 17 |
> Same here. All my server VMs work just fine with parallel enabled. |
| 18 |
> There's nothing complex in them, they tend to be single-service |
| 19 |
> machines. |
| 20 |
> |
| 21 |
|
| 22 |
Don't tell me you reboot your servers so often that it is necessary to |
| 23 |
tune the boot process for every last second. And please tell me you make |
| 24 |
the time slots for scheduled reboots large enough for trouble shooting, |
| 25 |
thereby not requiring every last second, either. |
| 26 |
|
| 27 |
> I don't have a current desktop Gentoo system, those necessarily have |
| 28 |
> more complex start-up routines. Perhaps that's where most of the |
| 29 |
> problems are to found? |
| 30 |
> |
| 31 |
|
| 32 |
Guess so. Besides, there is a new init script format in the pipe, for |
| 33 |
example mentioned here: [1] It will also make use of cgroups [2]. IMHO |
| 34 |
loosing a few seconds of boot time is an acceptable price for better CPU |
| 35 |
and IO scheduling. |
| 36 |
|
| 37 |
If these "new style" scripts are written declarative, that means less |
| 38 |
shell scripting and probably better performance even under sequential |
| 39 |
execution. And as I've learned often and hard: You don't parallelize |
| 40 |
until you have properly optimized your sequential execution, not the |
| 41 |
other way around. |
| 42 |
|
| 43 |
WTF do you need fast boot processes, anyway?! If you care about this, |
| 44 |
you hibernate or suspend. Daily shutdown/bootup sounds like something |
| 45 |
you'd do on a diskless client, a pre-ACPI system or some flakey |
| 46 |
hardware. I hardly see a boot screen once per month. My laptop currently |
| 47 |
has an uptime of 15 days, my workstation three months. You probably |
| 48 |
waste more time repopulating your page cache after starting your desktop |
| 49 |
environment than you do with init scripts. |
| 50 |
|
| 51 |
[1] http://blog.flameeyes.eu/2011/10/22/updating-init-scripts |
| 52 |
[2] http://blog.flameeyes.eu/2011/11/28/the-infamous-run-migration |
| 53 |
|
| 54 |
Regards, |
| 55 |
Florian Philipp |
Archives