1 |
----- Original Message ----- |
2 |
From: "Marko McLion" <mclion@×××××.net> |
3 |
To: <gentoo-server@l.g.o> |
4 |
Sent: Tuesday, January 20, 2004 12:53 PM |
5 |
Subject: [gentoo-server] strange bogomips difference |
6 |
|
7 |
|
8 |
> Hi |
9 |
> |
10 |
> We've recently purcased a new 2cpu server and installer gentoo on it. It's |
11 |
a |
12 |
> dual Athlon MP2800. All came smooth, but when the load of the server grew |
13 |
and |
14 |
> mysql got to 180 queries/sec it started to die (mysql daemon). We tried |
15 |
> everything, even the binary distribution from mysql.com and lowering the |
16 |
> maxconnections settings, but with no success... Then we've noticed that |
17 |
gcc |
18 |
> while compiling the last mysql sources died with a strange error, that |
19 |
says |
20 |
> in brief that there is a hardware error maybe... So we've checked the |
21 |
machine |
22 |
> (hardware) and found this: |
23 |
> ----------- |
24 |
> $ cat /proc/cpuinfo |
25 |
> processor : 0 |
26 |
> vendor_id : AuthenticAMD |
27 |
> cpu family : 6 |
28 |
> model : 10 |
29 |
> model name : AMD Athlon(tm) MP 2800+ |
30 |
> stepping : 0 |
31 |
> cpu MHz : 2133.380 |
32 |
> cache size : 512 KB |
33 |
> fdiv_bug : no |
34 |
> hlt_bug : no |
35 |
> f00f_bug : no |
36 |
> coma_bug : no |
37 |
> fpu : yes |
38 |
> fpu_exception : yes |
39 |
> cpuid level : 1 |
40 |
> wp : yes |
41 |
> flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca |
42 |
> cmov pat pse36 mmx fxsr sse syscall mp mmxext 3dnowext 3dnow |
43 |
> bogomips : 4194.30 |
44 |
> |
45 |
> processor : 1 |
46 |
> vendor_id : AuthenticAMD |
47 |
> cpu family : 6 |
48 |
> model : 10 |
49 |
> model name : AMD Athlon(tm) MP |
50 |
> stepping : 0 |
51 |
> cpu MHz : 2133.380 |
52 |
> cache size : 512 KB |
53 |
> fdiv_bug : no |
54 |
> hlt_bug : no |
55 |
> f00f_bug : no |
56 |
> coma_bug : no |
57 |
> fpu : yes |
58 |
> fpu_exception : yes |
59 |
> cpuid level : 1 |
60 |
> wp : yes |
61 |
> flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca |
62 |
> cmov pat pse36 mmx fxsr sse syscall mp mmxext 3dnowext 3dnow |
63 |
> bogomips : 4259.84 |
64 |
> |
65 |
> --------- |
66 |
> |
67 |
> As you can see, the model name is displayed differently by the cpus and |
68 |
the |
69 |
> bogomips is different too.. Could be this the problem? |
70 |
> |
71 |
> We've disasembled the server and to the naked eye the cpu-s look the same. |
72 |
It |
73 |
> is possible that there is so much difference? Or is it a software problem? |
74 |
|
75 |
Once upon a time, a smp kernel could run also on non smp machine. |
76 |
Can you try to take away one CPU, then test the machine, compiling |
77 |
the mysql sources, then if you get the same error, repeat with the other |
78 |
CPU in the other slot, taking away the one you tested before. |
79 |
Which is the strange gcc error that you get, compiling the mysql sources? |
80 |
Isnt' it a sig 11? In such case could be a memory problem. |
81 |
|
82 |
Andrea |