| 1 |
On Thu, Feb 19, 2009 at 10:04 AM, Beau Henderson <beau@××××××××××××.com>wrote: |
| 2 |
|
| 3 |
> |
| 4 |
> |
| 5 |
> On Thu, Feb 19, 2009 at 9:53 AM, Alan McKinnon <alan.mckinnon@×××××.com>wrote: |
| 6 |
> |
| 7 |
>> On Thursday 19 February 2009 01:38:39 Beau Henderson wrote: |
| 8 |
>> > I've tried manually altering the governor to performance but its the |
| 9 |
>> same |
| 10 |
>> > story. |
| 11 |
>> > |
| 12 |
>> > The system doesn't appear sluggish, I'm really more concerned that |
| 13 |
>> > something is causing the load and this might lead to shorter battery |
| 14 |
>> life |
| 15 |
>> > and and more heat. |
| 16 |
>> |
| 17 |
>> Right in the beginning you said the load was *exactly* 1.00. Now, load is |
| 18 |
>> defined as |
| 19 |
>> |
| 20 |
>> "the _number_ of processes on average waiting for the cpu in the last 1, |
| 21 |
>> 5, 15 |
| 22 |
>> minutes" |
| 23 |
>> |
| 24 |
>> So it does not mean that the cpu is necessarily working hard (but usually |
| 25 |
>> does) if the load is high. Yours is _exactly_ 1.00 (very suspicious) |
| 26 |
>> |
| 27 |
>> This is almost certainly one of two things: |
| 28 |
>> |
| 29 |
>> 1. A stupid kernel config that you should not have done :-) |
| 30 |
>> 2. Some app is blocking hard on IO |
| 31 |
>> |
| 32 |
>> I guess #2 - something waits for IO, it is not available, so immediately |
| 33 |
>> goes |
| 34 |
>> back to sleep waiting for it's next time slice. This happens many times a |
| 35 |
>> second and averaged over a minute looks like the cpu is constantly busy. |
| 36 |
>> Thus, |
| 37 |
>> no real extra cpu load is happening, the machine does not appear at all |
| 38 |
>> sluggish and the only harm is that it is annoying as hell. |
| 39 |
>> |
| 40 |
>> -- |
| 41 |
>> alan dot mckinnon at gmail dot com |
| 42 |
>> |
| 43 |
>> |
| 44 |
>> |
| 45 |
> Woah, now were getting somewhere. |
| 46 |
> |
| 47 |
> After reading that, I had another look at the top output and noticed that a |
| 48 |
> single hald process was in D state. /etc/init.d/hald stop and the load is |
| 49 |
> lowering as I type. I'm going to have to dig into this deeper as time |
| 50 |
> permits. |
| 51 |
> |
| 52 |
> Thanks everyone :) |
| 53 |
> |
| 54 |
> |
| 55 |
> -- |
| 56 |
> Beau Dylan Henderson |
| 57 |
> |
| 58 |
> "No human being should be denied the fundamental right to educate |
| 59 |
> themselves or indulge their curiosities. To deny any person the right to do |
| 60 |
> so, for whatever reason, is nothing more than the safeguarding of ignorance |
| 61 |
> to ensure that enlightenment does not become a threat. For nothing in this |
| 62 |
> world is more dangerous than an open mind." -- Matthew Good |
| 63 |
> |
| 64 |
|
| 65 |
|
| 66 |
The culprit: Hals cdrom polling. Interestingly, the load shot down as soon |
| 67 |
as I stuck a disk. |
| 68 |
|
| 69 |
The fix: hal-disable-polling --device /dev/scd0 'hal' |
| 70 |
-- |
| 71 |
Beau Dylan Henderson |
| 72 |
|
| 73 |
"No human being should be denied the fundamental right to educate themselves |
| 74 |
or indulge their curiosities. To deny any person the right to do so, for |
| 75 |
whatever reason, is nothing more than the safeguarding of ignorance to |
| 76 |
ensure that enlightenment does not become a threat. For nothing in this |
| 77 |
world is more dangerous than an open mind." -- Matthew Good |
Archives