| 1 |
Grant wrote: |
| 2 |
|
| 3 |
>>what kernel are you using ? in 2.6 yes definetively there is a way, also |
| 4 |
>>some of the 2.4 series kernel was swapping too much, google for "swappines". |
| 5 |
>>normally it's set to 60, lower that value and your swap will empty. |
| 6 |
>>if you have a pre 2.4.19 kernel upgrade to a newer one also of the same |
| 7 |
>>series. |
| 8 |
>> |
| 9 |
>> |
| 10 |
> |
| 11 |
>I'm using 2.4.27-hardened, and I think swappiness is only available |
| 12 |
>for 2.6. Maybe it's time to move to 2.6, but this is a commercial |
| 13 |
>server so I'm not so sure. |
| 14 |
> |
| 15 |
> |
| 16 |
> |
| 17 |
If it's a commercial server and it's load under the 40% |
| 18 |
it's easy, buy a disk and build a chrooted gentoo with hardened-dev-sources |
| 19 |
then schedule a downtime of some hour to make the change. |
| 20 |
Before to do this anyway it's better to figure out what really is the |
| 21 |
problem, try to isolate it, |
| 22 |
reboot, cat somewhere that output of "top -b -n1" and diff one hour |
| 23 |
after and one day after, |
| 24 |
try also "ps ax -eo user,vsize,size,command" and "man ps" I bet you will |
| 25 |
find what is strange |
| 26 |
|
| 27 |
btw 2.4.27 it's *not* an old kernel. |
| 28 |
|
| 29 |
>>Also it's not true with older kernels that running out of memory it's |
| 30 |
>>safe, *really* it's not if you are also writing a lot to a disk. |
| 31 |
>> |
| 32 |
>> |
| 33 |
> |
| 34 |
>Even if there is no crashing and the program just fails to load. |
| 35 |
>That's no good either. I don't think disabling swap is the solution |
| 36 |
>at all (we agree). |
| 37 |
> |
| 38 |
>- Grant |
| 39 |
> |
| 40 |
>-- |
| 41 |
>gentoo-user@g.o mailing list |
| 42 |
> |
| 43 |
> |
| 44 |
> |
| 45 |
> |
| 46 |
|
| 47 |
|
| 48 |
-- |
| 49 |
gentoo-user@g.o mailing list |
Archives