From 7a06e158f1f6185c015b3c6a87da1c4ea5129839 Mon Sep 17 00:00:00 2001
From: Harshavardhana <harsha@minio.io>
Date: Mon, 6 Jan 2020 00:30:04 -0800
Subject: [PATCH] Add docs for sysctl tuning parameters (#8749)

---
 docs/deployment/kernel-tuning/README.md      |  2 +
 docs/deployment/kernel-tuning/disk-tuning.sh |  0
 docs/deployment/kernel-tuning/sysctl.sh      | 91 ++++++++++++++++++++
 3 files changed, 93 insertions(+)
 mode change 100644 => 100755 docs/deployment/kernel-tuning/disk-tuning.sh
 create mode 100755 docs/deployment/kernel-tuning/sysctl.sh

diff --git a/docs/deployment/kernel-tuning/README.md b/docs/deployment/kernel-tuning/README.md
index 69bb27c54..8a66adf53 100644
--- a/docs/deployment/kernel-tuning/README.md
+++ b/docs/deployment/kernel-tuning/README.md
@@ -58,6 +58,8 @@ echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/enabled
 
 Also, set `transparent_hugepage=madvise` on your kernel command line (e.g. in /etc/default/grub) to persistently set this value.
 
+All these system level tunings are conveniently packaged in [shell script](https://github.com/minio/minio/blob/master/docs/deployment/kernel-tuning/sysctl.sh). Please review the shell script for our recommendations.
+
 ## Tuning Scheduler
 
 Proper scheduler configuration makes sure MinIO process gets adequate CPU time. Here are the recommended scheduler settings
diff --git a/docs/deployment/kernel-tuning/disk-tuning.sh b/docs/deployment/kernel-tuning/disk-tuning.sh
old mode 100644
new mode 100755
diff --git a/docs/deployment/kernel-tuning/sysctl.sh b/docs/deployment/kernel-tuning/sysctl.sh
new file mode 100755
index 000000000..eb2914969
--- /dev/null
+++ b/docs/deployment/kernel-tuning/sysctl.sh
@@ -0,0 +1,91 @@
+#!/bin/bash
+
+cat > sysctl.conf <EOF
+# maximum number of open files/file descriptors
+fs.file-max = 4194303
+
+# use as little swap space as possible
+vm.swappiness = 1
+
+# prioritize application RAM against disk/swap cache
+vm.vfs_cache_pressure = 10
+
+# minimum free memory
+vm.min_free_kbytes = 1000000
+
+# maximum receive socket buffer (bytes)
+net.core.rmem_max = 268435456
+
+# maximum send buffer socket buffer (bytes)
+net.core.wmem_max = 268435456
+
+# default receive buffer socket size (bytes)
+net.core.rmem_default = 67108864
+
+# default send buffer socket size (bytes)
+net.core.wmem_default = 67108864
+
+# maximum number of packets in one poll cycle
+net.core.netdev_budget = 1200
+
+# maximum ancillary buffer size per socket
+net.core.optmem_max = 134217728
+
+# maximum number of incoming connections
+net.core.somaxconn = 65535
+
+# maximum number of packets queued
+net.core.netdev_max_backlog = 250000
+
+# maximum read buffer space
+net.ipv4.tcp_rmem = 67108864 134217728 268435456
+
+# maximum write buffer space
+net.ipv4.tcp_wmem = 67108864 134217728 268435456
+
+# enable low latency mode
+net.ipv4.tcp_low_latency = 1
+
+# socket buffer portion used for TCP window
+net.ipv4.tcp_adv_win_scale = 1
+
+# queue length of completely established sockets waiting for accept
+net.ipv4.tcp_max_syn_backlog = 30000
+
+# maximum number of sockets in TIME_WAIT state
+net.ipv4.tcp_max_tw_buckets = 2000000
+
+# reuse sockets in TIME_WAIT state when safe
+net.ipv4.tcp_tw_reuse = 1
+
+# time to wait (seconds) for FIN packet
+net.ipv4.tcp_fin_timeout = 5
+
+# disable icmp send redirects
+net.ipv4.conf.all.send_redirects = 0
+
+# disable icmp accept redirect
+net.ipv4.conf.all.accept_redirects = 0
+
+# drop packets with LSR or SSR
+net.ipv4.conf.all.accept_source_route = 0
+
+# MTU discovery, only enable when ICMP blackhole detected
+net.ipv4.tcp_mtu_probing = 1
+
+EOF
+
+echo "Enabling system level tuning params"
+sysctl --quiet --load sysctl.conf && rm -f sysctl.conf
+
+# `Transparent Hugepage Support`*: This is a Linux kernel feature intended to improve
+# performance by making more efficient use of processor’s memory-mapping hardware.
+# But this may cause https://blogs.oracle.com/linux/performance-issues-with-transparent-huge-pages-thp
+# for non-optimized applications. As most Linux distributions set it to `enabled=always` by default,
+# we recommend changing this to `enabled=madvise`. This will allow applications optimized
+# for transparent hugepages to obtain the performance benefits, while preventing the
+# associated problems otherwise. Also, set `transparent_hugepage=madvise` on your kernel
+# command line (e.g. in /etc/default/grub) to persistently set this value.
+
+echo "Enabling THP madvise"
+echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/enabled