| 1 |
On 12/02/2009 05:19 PM, Willie Wong wrote: |
| 2 |
> On Wed, Dec 02, 2009 at 03:50:30PM +0200, Penguin Lover Nikos Chantziaras squawked: |
| 3 |
>> On 12/02/2009 01:22 PM, Dirk Uys wrote: |
| 4 |
>>> This has been bothering me for some time now. I have a Dell PC at work, |
| 5 |
>>> Intel Core2 Duo with 4gb ram etc. Whenever something does a lot of disk |
| 6 |
>>> access, the PC slows down to a halt? I remember some issue between |
| 7 |
>>> Firefox and the kernel causing long pauses, but I've had several |
| 8 |
>>> different kernel versions running. |
| 9 |
>>> [...] |
| 10 |
>> |
| 11 |
>> It's a known problem. I have the same issue. But there is a solution: |
| 12 |
>> start disk I/O heavy tasks with "ionice -c3". For emerge, this can be done |
| 13 |
>> automatically by putting this in your make.conf: |
| 14 |
>> |
| 15 |
>> PORTAGE_IONICE_COMMAND="ionice -c 3 -p \${PID}" |
| 16 |
>> |
| 17 |
>> ionice is in sys-apps/util-linux so it should be installed already. |
| 18 |
> |
| 19 |
> Hum, I had forgotten about this command. It would have come in handy a |
| 20 |
> few days ago. But in the case of FireFox, wouldn't that make it worse? |
| 21 |
|
| 22 |
No. If you run other tasks as "ionice -c3" they will stop blocking Firefox. |
| 23 |
|
| 24 |
|
| 25 |
> Also, what do you mean by known problem? What sort of set-up causes |
| 26 |
> the problem? Is this related at all to the hardware used? Or is this |
| 27 |
> purely in software? |
| 28 |
|
| 29 |
No, it's related to the Linux kernel blocking other applications for too |
| 30 |
long when one of them does heavy I/O. This is done so that each task |
| 31 |
has a chance to get more work done for the amount of time it gets to do |
| 32 |
I/O. As you can imagine, this hurts non-server systems quite badly. |
| 33 |
|
| 34 |
Another solution is to buy more hard disks and set up a RAID. |
Archives