From 0414759c745a08c42641bfedf28a9387f8c9c076 Mon Sep 17 00:00:00 2001
From: Nicolas <Nicolas@nonan.net>
Date: Thu, 17 Mar 2022 14:14:49 +0100
Subject: [PATCH] readme added, additional bcache settings added

---
 README.md       | 50 +++++++++++++++++++++++++++++++++++++++++++
 setup_apps.sh   |  6 ++++--
 setup_bcache.sh | 56 +++++++++++++++++++++++++++++++++++++++++++++++++
 setup_cuda.sh   | 20 +++++++++---------
 setup_drives.sh | 41 ------------------------------------
 5 files changed, 120 insertions(+), 53 deletions(-)
 create mode 100644 README.md
 create mode 100644 setup_bcache.sh
 delete mode 100644 setup_drives.sh

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..b97ca3a
--- /dev/null
+++ b/README.md
@@ -0,0 +1,50 @@
+# Setup for setup Tensorflow with GPU
+
+Tested for ubuntu-20.04.4
+
+Steps:
+
+1. Prepare setup:
+
+       git clone https://repos.nonan.net/nicolas/gpu_server_setup.git
+       cd gpu_server_setup
+
+
+2. Setup driver/CUDA:
+
+       sudo bash setup_cuda.sh
+       sudo systemctl reboot
+
+
+3. Setup bcache:
+
+       sudo bash setup_bcache.sh
+
+
+3. Setup apps (Python, JupyterHub (Hub is running as root), Tensorflow etc.):
+
+       sudo bash setup_apps.sh
+
+
+## Notes
+
+### CUDA
+
+Check state of NVIDIA devices (electrical power, temperature, memory etc.):
+
+    nvidia-smi
+
+
+### bcache
+
+Check bcache performance:
+
+    cat /sys/block/bcache0/bcache/state
+    cat /sys/block/bcache*/bcache/stats_five_minute/cache_hit_ratio
+    cat /sys/block/bcache*/bcache/stats_hour/cache_hit_ratio
+
+
+Tune bcache (not permanent):
+
+    echo 64M > /sys/block/bcache0/bcache/sequential_cutoff
+    echo 4096 > /sys/block/bcache0/queue/read_ahead_kb
\ No newline at end of file
diff --git a/setup_apps.sh b/setup_apps.sh
index f1c6185..72cd2d6 100644
--- a/setup_apps.sh
+++ b/setup_apps.sh
@@ -31,9 +31,11 @@ systemctl enable jupyterhub && \
 systemctl start jupyterhub
 
 if [ $? -eq 0 ]; then
+  echo "check nvidia drivers..."
+  nvidia-smi
+
   echo "\n"
   echo "Setup successfully finished"
 fi
 
-#check nvidia drivers
-nvidia-smi
+
diff --git a/setup_bcache.sh b/setup_bcache.sh
new file mode 100644
index 0000000..67d9e77
--- /dev/null
+++ b/setup_bcache.sh
@@ -0,0 +1,56 @@
+#!/bin/bash
+
+#sources:
+#https://wiki.archlinux.org/title/bcache
+#https://stackoverflow.com/questions/20797819/command-to-change-the-default-home-directory-of-a-user
+
+ssddev=/dev/sdb
+hdddev=/dev/sdc
+sequential_cutoff=24000M
+
+if [ "$EUID" -ne 0 ]
+  then echo "Please run as root"
+  exit
+fi
+
+while [ -b /dev/bcache0 ]
+  then echo "Canceld, device /dev/bcache0 is allready present"
+  exit
+fi
+
+echo "Setup bcache... "
+apt-get update && \
+apt-get install -y bcache-tools && \
+dd if=/dev/zero if=$ssddev bs=512 count=8 && \
+dd if=/dev/zero if=$hdddev bs=512 count=8 && \
+wipefs -a $ssddev && \
+wipefs -a $hdddev && \
+make-bcache -C $ssddev -B $hdddev --writeback && \
+echo $sequential_cutoff > /sys/block/bcache0/bcache/sequential_cutoff && \
+echo "ACTION==\"add\", SUBSYSTEM==\"block\", ENV{MAJOR}==\"252\", ATTR{bcache/sequential_cutoff}=\"$sequential_cutoff\"" \
+> /etc/udev/rules.d/99-bcache_sequential_cutoff.rules && \
+udevadm test /sys/block/bcache0
+
+if [ $? -eq 0 ]; then
+  echo "Wait until device is ready... "
+  while [ ! -b /dev/bcache0 ]; do sleep 1; done
+
+  echo "Fromat and mount bcache device... "
+  mkfs.ext4 /dev/bcache0 && \
+  mkdir /mnt/bcache && \
+  mount /dev/bcache0 /mnt/bcache && \
+  echo "/dev/bcache0 /mnt/bcache ext4 rw 0 0" >> /etc/fstab
+fi
+
+if [ $? -eq 0 ]; then
+  echo "Config home directory for new users on bcache partition... "
+  mkdir /mnt/bcache/home && \
+  echo "#" >> /etc/default/useradd && \
+  echo "# Modifications:" >> /etc/default/useradd && \
+  echo "HOME=/mnt/bcache/home" >> /etc/default/useradd
+fi
+
+if [ $? -eq 0 ]; then
+  echo ""
+  echo "Setup successfully finished"
+fi
diff --git a/setup_cuda.sh b/setup_cuda.sh
index 9080bd1..4eb85f8 100644
--- a/setup_cuda.sh
+++ b/setup_cuda.sh
@@ -4,12 +4,15 @@
 #check for NVIDIA device:
 #apt-get -y install pciutils
 #lspci | grep VGA
-#check if driver works with device:
-#nvidia-smi
+#check if driver works with device: nvidia-smi
 
 #based on https://docs.nvidia.com/cuda/cuda-installation-guide-linux/
 #and https://www.tensorflow.org/install/gpu
 #Versions required: https://www.tensorflow.org/install/source#gpu
+
+#remove cuda:
+#sudo apt-get remove cuda-11-2
+#sudo apt-get autoremove
 #------------------------------------
 
 if [ "$EUID" -ne 0 ]
@@ -47,17 +50,18 @@ if [ $? -eq 0 ]; then
   #Use the --no-install-recommends option for a lean driver install without any dependencies on X packages
   apt-get update && \
   apt-get install -y --no-install-recommends cuda-11-2
-  #apt-get install --no-install-recommends libcudnn8=8.1.0.44-1+cuda11.2 libcudnn8-dev=8.1.0.44-1+cuda11.2
 fi
 
 if [ $? -eq 0 ]; then
   echo "\n"
   echo "install libcudnn8..."
-  #Latest available version from ubuntu1804 repo:
+  #Latest available version from nvidia ubuntu1804 repo:
   wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/libcudnn8_8.1.1.33-1+cuda11.2_amd64.deb && \
-  wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb && \
   apt-get install -y ./libcudnn8_8.1.1.33-1+cuda11.2_amd64.deb && \
-  apt-get install -y ./libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb
+  rm ./libcudnn8_8.1.1.33-1+cuda11.2_amd64.deb && \
+  wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb && \
+  apt-get install -y ./libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb && \
+  rm ./libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb
 fi
 
 if [ $? -eq 0 ]; then
@@ -67,7 +71,3 @@ if [ $? -eq 0 ]; then
   echo "Setup successfully finished"
   echo "Type \"systemctl reboot\" to reboot system"
 fi
-
-
-#sudo apt-get remove cuda-11-2
-#sudo apt-get autoremove
\ No newline at end of file
diff --git a/setup_drives.sh b/setup_drives.sh
deleted file mode 100644
index 5def89e..0000000
--- a/setup_drives.sh
+++ /dev/null
@@ -1,41 +0,0 @@
-#!/bin/bash
-
-#https://stackoverflow.com/questions/20797819/command-to-change-the-default-home-directory-of-a-user
-
-#move user with files:
-#usermod -m -d /newhome/username username
-
-#Simply open this file using a text editor, type:
-#vi /etc/default/useradd
-#The default home directory defined by HOME variable, find line that read as follows:
-#HOME=/home
-#Replace with:
-#HOME=/iscsi/user
-#Save and close the file. Now you can add user using regular useradd command:
-# useradd vivek
-# passwd vivek
-#Verify user information:
-# finger vivek
-
-#echo "/dev/bcache0 /mnt/bcache ext4 rw 0 0" | tee -a /etc/fstab > /dev/null
-
-
-apt-get update && \
-apt-get install -y bcache-tools && \
-dd if=/dev/zero if=/dev/sdb bs=512 count=8 && \
-dd if=/dev/zero if=/dev/sdc bs=512 count=8 && \
-wipefs -a /dev/sdb && \
-wipefs -a /dev/sdc && \
-make-bcache -C /dev/sdb -B /dev/sdc --discard --writeback && \
-mkfs.ext4 /dev/bcache0 && \
-mkdir /mnt/bcache && \
-mount /dev/bcache0 /mnt/bcache && \
-mkdir /mnt/bcache/home && \
-echo "/dev/bcache0 /mnt/bcache ext4 rw 0 0" >> /etc/fstab
-
-
-if [ $? -eq 0 ]; then
-    echo "#" >> /etc/default/useradd
-    echo "# Modifications:" >> /etc/default/useradd
-    echo "HOME=/mnt/bcache/home" >> /etc/default/useradd
-fi
\ No newline at end of file