| 1 |
Hi Christian, |
| 2 |
|
| 3 |
Perhaps try booting a standard kernel using the hardened profile and see how long that takes. If it takes closer to 12 hours, then the slowdown is caused by the kernel. If it still takes about 24 hours, then it is caused by the hardened profile. |
| 4 |
|
| 5 |
On another note, have you trying tweaking the options in /etc/mysql/my.cnf? Increasing some of the buffer sizes might improve things, although this would perhaps be more applicable if mysqld was IO bound instead of CPU bound as in your case. Might be still worth googling for mysqld optimisations anyway. |
| 6 |
|
| 7 |
Cheers, |
| 8 |
|
| 9 |
Brad |
| 10 |
|
| 11 |
On Thu, 19 Jul 2007 10:14:06 +0200 |
| 12 |
Christian Parpart <trapni@g.o> wrote: |
| 13 |
|
| 14 |
> Hi all, |
| 15 |
> |
| 16 |
> I just migrated a standard gentoo/linux to a hardened profile |
| 17 |
> including hardened-sources with grsecurity enabled and PaX-disabled. |
| 18 |
> |
| 19 |
> Well, it is basically used as a mySQL server, however, we do a biig |
| 20 |
> sql db import once a week to operate on it. the import took about 12 |
| 21 |
> hours. but now, using the hardened sourced and the hardened profile |
| 22 |
> built system it really taked about 24+ hours. |
| 23 |
> |
| 24 |
> this really confuses me. can it be because of the extra validations |
| 25 |
> the kernel-/userspace has to perform? |
| 26 |
> |
| 27 |
> mysqld still ran at 100%, before the hardened migration and after. |
| 28 |
> |
| 29 |
> because *if* it is really the hardened enabled stuff that lowers the |
| 30 |
> performance that much, than we need to revert to non-hardened - and I |
| 31 |
> don't like that idea yet :) |
| 32 |
> |
| 33 |
> Any hints on this topic are really welcome, |
| 34 |
> Thanks in advance, |
| 35 |
> Christian Parpart. |
Archives