Hadoop recommendation for maximum open file descriptors

1.      Package Size and Inode Count Requirements

·         The recommended maximum number of open file descriptors is 10000 on each host, run below command to check:

ulimit -Sn

ulimit -Hn

·         If the output is not greater than 10000, run the following command to set it to a suitable default:

Edit /etc/sysctl.conf

fs.file-max = 65536

sysctl -p /etc/sysctl.conf

Edit /etc/security/limits.conf

Youruser or *       soft    nofile   10000

Youruser or *       hard    nofile   10000

Hadoop Disable Transparent Huge Pages

Hadoop HDP THP Disable

·         Disable Transparent Huge Pages

§  RHEL

·         Disable THP at boot time

·         Append the following to the kernel command line in /etc/grub.conf:

·         transparent_hugepage=never

·         Add the following commands to your /etc/rc.local file

if test -f /sys/kernel/mm/transparent_hugepage/enabled; then

    echo never > /sys/kernel/mm/redhat_transparent_hugepage/enabled

fi

if test -f /sys/kernel/mm/transparent_hugepage/defrag; then

    echo never > /sys/kernel/mm/redhat_transparent_hugepage/defrag

fi

·       Disable THP at run time

·         Run the following commands to disable THP on-the-fly:

·         echo never > /sys/kernel/mm/redhat_transparent_hugepage/enabled

·         echo never > /sys/kernel/mm/redhat_transparent_hugepage/defrag

·       How to check if HugePages is enabled or disabled

·         if the value of HugePages_Total is "0" it means HugePages is disabled on the system.

·         grep -i HugePages_Total /proc/meminfo

·          if the value in /proc/sys/vm/nr_hugepages file or vm.nr_hugepages sysctl parameter is "0" it means HugePages is disabled on the system

·         cat /proc/sys/vm/nr_hugepages

·         sysctl vm.nr_hugepages

§  Reference Documents

·         https://access.redhat.com/solutions/46111

·         https://docs.mongodb.org/manual/tutorial/transparent-huge-pages/