cim vic jader CPU, tim pomalejsi

Miroslav Lachman 000.fbsd at quip.cz
Tue Mar 30 01:59:03 CEST 2021


On 30/03/2021 00:45, Miroslav Lachman wrote:

> Ani jsem netusil, ze VirtualBox neco takoveho umi, takze jsem zase o 
> neco chytrejsi. No u me to vypada, ze kdyz mu pridelim 6 jader, tak i 
> guest vidi 6 jader. Zkusim to ted rebootovat a zvednout to na 10 a 
> pozorne si prohlidnout, jestli jich je skutecne 10.

Opravdu system vidi 10 jader
FreeBSD/SMP: 1 package(s) x 10 core(s)

Problem je, ze je s temi 10 jadry opravdu neuveritelne pomaly. Takhle to 
vypadalo ctvrt hodiny po rebootu (v noci, kdy je na serveru minimalni 
provoz a zatez by mela byt okolo 10% / load 1)

last pid:  3561;  load averages: 20.78, 14.75,  8.14     up 0+00:16:32 
01:15:55
170 processes: 14 running, 152 sleeping, 1 zombie, 3 lock
CPU: 26.3% user,  0.0% nice, 65.6% system,  8.1% interrupt,  0.0% idle
Mem: 1913M Active, 321M Inact, 1913M Wired, 1563M Buf, 12G Free
Swap: 2047M Total, 2047M Free

   PID USERNAME THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
  1348 root       1 -16    0  6852K  2940K CPU5    5   6:57 111.36% pfctl
  3093 www        1  78    0   185M 59940K CPU4    4   0:03  92.79% httpd
   811 mysql     75  20    0  4806M  1345M select  2   3:17  57.92% mysqld
  3086 www        1  52    0   187M 45504K select  3   0:18  48.35% httpd
  3089 www        1  33    0   185M 57304K CPU0    0   0:12  46.55% httpd
  3012 www        1  52    0   198M 64900K lockf   6   0:18  43.90% httpd
  3098 www        1  80    0   187M 71556K RUN     2   0:06  39.39% httpd
  3015 www        1  82    0   193M 68112K CPU3    3   0:20  36.74% httpd
  3021 www        1  52    0   185M 75052K select  6   0:14  29.85% httpd
  3024 www        1  52    0   187M 74820K RUN     1   0:12  21.30% httpd
  3087 www        1  78    0   187M 85288K RUN     5   0:15  18.51% httpd
  3022 www        1  25    0   185M 56596K select  2   0:12  17.54% httpd
  3095 www        1  52    0   187M 70400K RUN     3   0:16  16.00% httpd
  1357 postfix    1  23    0 48104K 12660K select  0   0:06  14.71% smtpd
  3091 www        1  77    0   187M 55568K RUN     0   0:13  14.12% httpd
  3099 www        1  52    0   189M 75456K lockf   9   0:09  12.67% httpd
  3071 www        1  36    0   187M 59348K lockf   3   0:12  11.83% httpd

Zkusil jsem to jeste s 8 jadry a to vypadalo skoro stejne spatne jako s 
10, se 7 jadry uz to bylo o neco lepsi. Ted jsem to zkusil se 4 jadry a 
system je skutecne temer bez zateze

last pid:  3354;  load averages:  0.76,  0.68,  0.40       up 0+00:09:06 
  01:55:56
109 processes: 1 running, 107 sleeping, 1 zombie
CPU:  4.1% user,  0.0% nice,  1.1% system,  0.1% interrupt, 94.7% idle
Mem: 1907M Active, 456M Inact, 1886M Wired, 1563M Buf, 11G Free
Swap: 2047M Total, 2047M Free

   PID USERNAME THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
  1389 www        1  23    0   194M   134M select  0   0:24  14.53% httpd
   812 mysql     50  20    0  4765M  1339M select  3   0:27   2.95% mysqld
   595 root       1  20    0  7152K  3308K select  3   0:04   1.14% p0f
   970 nobody    10  20    0 73832K 34284K kqread  3   0:03   0.62% 
memcached
  1433 root       1  20    0  7812K  3992K select  2   0:00   0.08% tmux
  1454 root       1  20    0  7912K  4316K CPU0    0   0:00   0.05% top
  1236 root       1  20    0   185M 35144K select  3   0:00   0.02% httpd
  1459 www        1  22    0   194M   128M select  3   0:19   0.01% httpd
   842 mailnull   1  20    0  6304K  2232K select  0   0:00   0.01% postsrsd
   567 root       1  20    0 12636K  8084K select  3   0:00   0.01% proftpd
   459 root       1  20    0  6420K  2440K select  1   0:00   0.01% syslogd
   300 root       1  20    0  9168K  5040K select  3   0:00   0.01% devd
  1358 www        1  20    0   202M   137M lockf   0   0:15   0.01% httpd


Porovnaval jsem nastaveni virtualu s tim, co poslal Dan a krome 
vypnuteho PAE tam byly ty podstatne veci stejne. Tak jsem zkusil zapnout 
PAE, ale zadny rozdil v tom chovani nepozoruji.

Mirek


More information about the Users-l mailing list