Do you have performance issues on the VPS? [message #38238] |
Tue, 01 December 2009 11:45 |
fernandomm
Messages: 11 Registered: November 2009
|
Junior Member |
|
|
Hello,
Last weekend i tried to move a high traffic website to a VPS on the same hardware that i was running it before.
I have set the ubc and other parameters like this:
# UBC parameters (in form of barrier:limit)
KMEMSIZE="999999999:999999999"
LOCKEDPAGES="256:256"
PRIVVMPAGES="524288:643072"
SHMPAGES="21504:21604"
NUMPROC="2000:2000"
PHYSPAGES="0:9223372036854775807"
VMGUARPAGES="524288:524288"
OOMGUARPAGES="524288:524288"
NUMTCPSOCK="999999999:999999999"
NUMFLOCK="999999999:999999999"
NUMPTY="16:16"
NUMSIGINFO="1024:1024"
TCPSNDBUF="999999999:999999999"
TCPRCVBUF="999999999:999999999"
OTHERSOCKBUF="999999999:999999999"
DGRAMRCVBUF="262144:262144"
NUMOTHERSOCK="999999999:999999999"
DCACHESIZE="3409920:3624960"
NUMFILE="999999999:999999999"
AVNUMPROC="180:180"
NUMIPTENT="800:800"
# Disk quota parameters (in form of softlimit:hardlimit)
DISKSPACE="211048576:211153024"
DISKINODES="400000:440000"
QUOTATIME="0"
# CPU fair sheduler parameter
CPUUNITS="153424"
After a while running it ( it's the same OS, same apache version and conf, same mysql conf etc. ) the apache processes stops working and use a lot of CPU ( load goes up to 400-500 ).
If i strace one of the processes, they do not show anything. Also, the apache logs are not updated the this starts to happen.
It is important to note that user_beancounters do not show any failcount, but the system is unusable.
We have about 400 requests per second, but our current hardware was running fine until me moved to a VPS.
Do you see any configuration issues at my ubc parameters? Could the diskinodes limitation be causing this problem ( limiting apache access to disk and making it goes crazy? )?
Anyway, any opinions and experiences about this issue are welcome
Thanks!
[Updated on: Tue, 01 December 2009 11:45] Report message to a moderator
|
|
|