Server Overload sem cauxa

Boa noite!

Venho passando por um problema estranho a algum tempo. As vezes meu servidor sobrecarrega. Isso faz o site travar e perco muita visitas. Uso um servidor VPS com 4gb de ram e processador AMD Opterom com 6 núcleos, e o site tem média de 10 mil acessos diários.

Aparentemente esse erro tem algo a ver com o carregamento PHP. Hoje o dia todo estava travando, lento e dificil de acessar. Mudei o sistema PHP Handler no WHM para FCGI. Anteriormente estava em CGI. Melhorou. Agora a pouco recebi um e-mail do servidor, com informações de log “HANG: chkservd”.

Esse problema piora se eu uso algum sistema de cache, como o WP Super Cache.

O relatório contido no e-mail do servidor foi o seguinte:

[2015-09-15 16:44:37 +0000] Service check ....queueprocd [[check command:+][socket connect:N/A]]...named [[check command:+][socket connect:N/A]]...mysql [[check command:+][socket connect:N/A]]...imap [[socket_service_auth:1][check command:+][socket connect:+]]...ftpd [[socket_service_auth:1][check command:+][socket connect:+]]...exim [[check command:+][socket connect:+]]...cpsrvd [[http_service_auth:1][check command:N/A][socket connect:+]]...cpanellogd [[check command:+][socket connect:N/A]]...Done
[2015-09-15 16:44:37 +0000] Disk check .... /tmp (/var/tmp) [7.07%] ... /dev/vda1 (/) [60.14%] ... /usr/tmpDSK (/tmp) [7.07%] ... /dev/vdc (/backup) [64.71%] ... {status:ok} ... Done
Loading services .....cpanellogd....cpsrvd....exim....ftpd....imap....mysql....named....queueprocd..Done
Service Check Started
Service Check Finished
[2015-09-15 16:38:36 +0000] Service check ....queueprocd [[check command:+][socket connect:N/A]]...named [[check command:+][socket connect:N/A]]...mysql [[check command:+][socket connect:N/A]]...imap [[socket_service_auth:1][check command:+][socket connect:+]]...ftpd [[socket_service_auth:1][check command:+][socket connect:+]]...exim [[check command:+][socket connect:+]]...cpsrvd [[http_service_auth:1][check command:N/A][socket connect:+]]...cpanellogd [[check command:+][socket connect:N/A]]...Done
[2015-09-15 16:38:36 +0000] Disk check .... /tmp (/var/tmp) [7.07%] ... /dev/vda1 (/) [60.13%] ... /usr/tmpDSK (/tmp) [7.07%] ... /dev/vdc (/backup) [64.71%] ... {status:ok} ... Done
Loading services .....cpanellogd....cpsrvd....exim....ftpd....imap....mysql....named....queueprocd..Done
Service Check Started
Service Check Finished

[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] Restored /usr/local/apache/logs/modsec_audit.log (size:0) to 0 (requested 0)
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] Restored /var/log/exim_mainlog (size:496796420) to 496731986 (requested 496731986)
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] Restored /var/log/maillog (size:406973) to 405591 (requested 405591)
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] /usr/local/apache/logs/modsec_audit.log opened with inode 195479
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] /var/log/exim_mainlog opened with inode 528324
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [INFO] /var/log/maillog opened with inode 528290
[2354] [2015-09-15 14:31:06 +0000] [Cpanel::TailWatch] [START] 2354 1442327466
[2222] [2015-09-15 14:29:13 +0000] [Cpanel::TailWatch::Eximstats] Loading email sending limits from 1442325600 - 1442329200
[2222] [2015-09-15 14:29:13 +0000] [Cpanel::TailWatch] The tailwatchd driver 'Cpanel::TailWatch::JailManager' is not enabled.
[2222] [2015-09-15 14:29:12 +0000] [Cpanel::TailWatch] [INFO] inotify enabled. watch file is /var/cpanel/.tailwatchd_inotify_alarm_trick

Recentemente fiz o upgrade no servidor, aumentando o CPU de 2 para 6 núcleos e parecia ter resolvido. Ontem voltou a acontecer. O problema é que isso ocorre mesmo com o site tendo um menor número de visitas, como a noite/madrugada ou no final de semana.

Agora esta normal, com Server Load em 0.92, mesmo com 120 usuários online. Se alguém tiver qualquer dica eu agradeço muito.

Obrigado!