Announcement

Collapse
No announcement yet.

Linux settings for high load systems

Collapse
This topic is closed.
X
This is a sticky topic.
X
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux settings for high load systems

    There are a few basic settings you have to adjust for high load systems to make sure the server have enough resources to handle big number of network connections.

    The main parameter is a maximum number of opened files allowed for the process to keep at the same time. Each network connection uses a file handler therefore if the limit is too low you can quickly run out of handlers and the server can not accept any more connections.

    This limit is set on 2 levels - on the kernel level (fs.file-max) and on the system level (nofile).

    Another kernel property which can be important in certain configurations (like transports installations or when you use proxy for Bosh connections) is: net.ipv4.ip_local_port_range. This parameter can be set the same way as the fs.file-max property.
    fs.file-max

    The fs.file-max kernel property is set via sysctl command. You can see current settings executing command:
    1 # sysctl fs.file-max
    2 fs.file-max = 358920

    If you plan to run high load service with big number of server connections then this parameter should be at least as twice big as the number of network connections you expect to support. You can change this setting executing command:
    1 # sysctl -w fs.file-max=360000
    2 fs.file-max = 360000
    net.ipv4.ip_local_port_range

    You can see current settings executing command:
    1 # sysctl net.ipv4.ip_local_port_range
    2 net.ipv4.ip_local_port_range = 32768 61000

    You can change this setting executing command:
    1 # sysctl -w net.ipv4.ip_local_port_range="1024 65000"
    2 net.ipv4.ip_local_port_range = 1024 65000
    TCP_keepalive

    According to http://blog.kolargol.eu/2006/06/tcpkeepalive.html?m=0 some keepalive settings should be changed to improve reliability.
    1 # sysctl -w net.ipv4.tcp_keepalive_time="60"
    2 net.ipv4.tcp_keepalive_time = 60
    3 # sysctl -w net.ipv4.tcp_keepalive_probes="3"
    4 net.ipv4.tcp_keepalive_probes = 3
    5 # sysctl -w net.ipv4.tcp_keepalive_intvl="90"
    6 net.ipv4.tcp_keepalive_intvl = 90
    /etc/sysctl.conf

    Above commands let the system remember new settings until the next system restart. If you want to make the change permanent you have to edit file: /etc/sysctl.conf and add the property at the end of the file:
    1 fs.file-max=360000
    2 net.ipv4.ip_local_port_range=1024 65000

    It will be automatically loaded next time you start the server. Command:
    1 # sysctl -p

    Causes the /etc/systcl.conf to be reloaded which is useful when you added more parameters to the file and don't want to restart the server.
    nofile

    This is the property used by the system limits. For example running the command ulimit -a shows you all limits set for the current user:
    01 # ulimit -a
    02 core file size (blocks, -c) 0
    03 data seg size (kbytes, -d) unlimited
    04 file size (blocks, -f) unlimited
    05 pending signals (-i) 38912
    06 max locked memory (kbytes, -l) 32
    07 max memory size (kbytes, -m) unlimited
    08 open files (-n) 40960
    09 pipe size (512 bytes, -p) 8
    10 POSIX message queues (bytes, -q) 819200
    11 stack size (kbytes, -s) 8192
    12 cpu time (seconds, -t) unlimited
    13 max user processes (-u) 38912
    14 virtual memory (kbytes, -v) unlimited
    15 file locks (-x) unlimited

    To make it even more interesting and more complex there are 2 types of system limits: soft limit which can be temporarily exceeded by the user and hard limit which can not be exceeded. To see your hard limit execute command:
    01 # ulimit -a -H
    02 core file size (blocks, -c) unlimited
    03 data seg size (kbytes, -d) unlimited
    04 file size (blocks, -f) unlimited
    05 pending signals (-i) 38912
    06 max locked memory (kbytes, -l) 32
    07 max memory size (kbytes, -m) unlimited
    08 open files (-n) 40960
    09 pipe size (512 bytes, -p) 8
    10 POSIX message queues (bytes, -q) 819200
    11 stack size (kbytes, -s) unlimited
    12 cpu time (seconds, -t) unlimited
    13 max user processes (-u) 38912
    14 virtual memory (kbytes, -v) unlimited
    15 file locks (-x) unlimited

    The hard limits are usually bigger then the soft limits or sometimes the same.

    For us the most important parameter is: open files. You can change the property in file: /etc/security/limits.conf. You have to append 2 following lines to the end of the file:
    1 jabber soft nofile 350000
    2 jabber hard nofile 350000

    Where the jabber is the user name of the account running you IM service. You can also set the limits for all users on the machine in a following way:
    1 * soft nofile 350000
    2 * hard nofile 350000

    For those changes to make an effect you have to logout from the modified account and login again. New limits should be applied.


    su and init script

    If one intends to use init scripts for startup purposes (or simply wants to be able to start the server utilizing su command it's necessary to adjust PAM configuration by modifying /etc/pam.d/su file and uncomment following line:
    1 session required pam_limits.so

    Afterwards that init scripts will respect configured limits.


    Source: http://www.tigase.org/content/linux-...h-load-systems
    www.AYKsolutions.com
    From Shared to Dedicated
    Professional. Painless. Polite.
Working...
X