1 |
On Fri, 15 Oct 2004, Istvan Soos wrote: |
2 |
|
3 |
>> Stage 3, so your system is not fully compiled and thus not fully |
4 |
>> optimised, however it should not impact your disk I/O ;p |
5 |
> |
6 |
> sorry, I've emerge -pv system (recompile gcc and others...) |
7 |
|
8 |
I hope you made a typo there; -p is for pretend. |
9 |
|
10 |
I know this is rather unfashionable, but I'd look to see what top said |
11 |
was going on while the system was being sluggish. I'd probably have top |
12 |
on a 15 second or more update, so you have a chance to really look at |
13 |
its output. If you have more io wait (wa%) when certain processes are |
14 |
nearly the top CPU users, they may have some specific issues. Note that |
15 |
some processes can do vast amounts of disk I/O with almost no CPU usage, |
16 |
so it can be difficult to find the culprit in this manner. |
17 |
|
18 |
I/O wait issues can also be caused by failing hardware. Normally, there |
19 |
would be a fair amount of kernel syslog complaints (these could be in |
20 |
/var/adm/messages, /var/log/syslog, or a variety of other places. I |
21 |
can't remember where Gentoo's standard location is; all I remember for |
22 |
certain was that its default syslog put too many facilities in one |
23 |
location for my preference; so I fixed it.) |
24 |
|
25 |
Finally - just to verify, you are actually seeing visible slowdowns, and |
26 |
not merely noticing the io wait stat? I have seen some people get very |
27 |
concerned about io wait climbing when they did io intensive commands; |
28 |
this is something that will always happen on a system which has more CPU |
29 |
bandwidth than disk bandwidth (i.e. almost every system around today.) |
30 |
This can be even more pronounced on systems where people've decided to |
31 |
do RAID on IDE or ATA devices (including, apparently, SATA devices, |
32 |
although I've not had a chance to play with these myself.) |
33 |
|
34 |
Ed |
35 |
|
36 |
-- |
37 |
gentoo-performance@g.o mailing list |