Firts version with ceph-install in provisioning. Nodes synchronization is working. Nodes' iP in /etc/hosts to avoid confusion with dnsmasq.

.signals/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

.ssh/.gitignore

@@ -1 +1,2 @@
-id_rsa
+*
+!.gitignore

.ssh/ceph-admin-config

@@ -1,15 +0,0 @@
-Host node-admin
-	Hostname node-admin
-	User ceph-admin
-	StrictHostKeyChecking no
-
-Host node-osd1
-	Hostname node-osd1
-	User ceph-admin
-	StrictHostKeyChecking no
-
-Host node-osd2
-	Hostname node-osd2
-	User ceph-admin
-	StrictHostKeyChecking no
-

.ssh/ceph-admin-id_rsa

@@ -1,27 +0,0 @@
------BEGIN RSA PRIVATE KEY-----
-MIIEowIBAAKCAQEAq5y73H3NbLPGsS/JRbNNHlR/DiwBilcXygw3F+1o82J5FiWR
-gDB85z4M0DpndvzMvqEtKJo2trCaxHrkfqFcmp8PKzlzkcKJLcY4UDJEJSeU0KKF
-pJhSFTD8ZVZaIB1vLXvwcfmcyly3OEX0HWhep/GQeQiIzxPG+eLQ3wIV7Y08Q9ze
-uREd4pbpEi5G2XDJWmb6+tF234ZoyS+A7j56ItfskzWVHn4sOdMScCvomfcymVaX
-B6UUJuhJsbosEPtoliF6XoIPL1eCFzqeo/nU0qk9OXT+53KVBwiSmPUYCzG4J4Br
-0J6QieGWCvPRZhxfsEIFUbR52PoJFmYHHATeEQIDAQABAoIBAEG5yYls9RxIzjRW
-ZPO49DRFkFPRLtXY/Vd4zjDv0F5GvLFqugsqhuvZq/akJ7Q+pfvspusCBSp7AQt3
-NcP58QmXGawoWMbehsNtQ2wTlF3WhvxRAnZDkExlcoUtSV87ZrU43qIIQ7CNSe7X
-e+M3celIb1irmBoXLQHPiMegeN19tUUEeHjXB2AlPrCZaco4mJ+PQhuh1uzLi4QD
-blGuzNwSOYHk21xvEXm7xoe+Q30utF0xYpOjDDUAlFCJRy/UNqC+2VYzBLtgqSZA
-dU20BuMpzgFokuywfjm7ftUoHtYiiUJZXDhXot9gBEN+WdC83XdT225pZfaW862J
-3azDoAECgYEA4Hg4yzi3A47s82xKXiJUM5DAXcAx1jd9leePhOjhl2NEfTlVPMDZ
-45BFBYhumOcgN/nrM4OqMi/hoCjcC2aDs4jG+12k7Y4MZijyrWmqc7PzJzLzG1yk
-7hlsSsMH454eZ5kjfAyzHumN9/Iz5hWwrF9Fesau5mpbgHR5/ekB1pECgYEAw7fL
-wZvapsIDUQOZeKZxNaDGDS/QG+BuZ5cZk3KGEIisOSv0sKMNmgs1rowPmzgZv2MF
-ubVzMa5wwuYc6q4vrQFijXm9uo2IlaOGXa84XSvCWpzmH4qa8+OZAtdyXZucUQlh
-JztK6plSJw57pxx1go/sJ01UGx9Xe7xXtO1RT4ECgYEAntBwXlgiJEJbI17avv9K
-07D2aV4x5H4ePJE42bBq69EwA+kshW9subS5INEGbugXw4LvBeOB50A5Qjemd+pw
-A/Yh0nbO2cRGwYD3MfowvuvyfSlvF0zU5CDYnYyfaEvW9zVCBeJ2WMilyWdzkLaH
-6Wz+m382/48AbkH6jcrtogECgYAJO96tH8FNJNIrPFQ9UPz+Xrt5W8whxJDw6jwW
-rMyOcyRM/jnpK4g0wFqkBY+m7xqn4PAgNuNccLJ5P4oVuGofa3VO33OIHa1YAgMQ
-4Hd0TeA5jjV9P3jYEODjNW7745+YYyzmQPHj55/sn85NsBYLNucZY5iYPv5NpDEv
-JBM2AQKBgHS5+HaCNSIziYWl7uowwxkJR4crg1KKzWqFbtR21F6cswP96k2+nXj2
-PymRCMWxRugssIe+GOlcCXsopdkvGvb/mFNug9NEmtbgNKQ/2bZRXlRNz2aEOEpt
-ryZn67PmOCV+ttcqna6xfBllsr2wyRrcTMfZQWZZn+gZ9bBYR4WU
------END RSA PRIVATE KEY-----

.ssh/ceph-admin-id_rsa.pub

@@ -1 +0,0 @@
-ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCrnLvcfc1ss8axL8lFs00eVH8OLAGKVxfKDDcX7WjzYnkWJZGAMHznPgzQOmd2/My+oS0omja2sJrEeuR+oVyanw8rOXORwoktxjhQMkQlJ5TQooWkmFIVMPxlVlogHW8te/Bx+ZzKXLc4RfQdaF6n8ZB5CIjPE8b54tDfAhXtjTxD3N65ER3ilukSLkbZcMlaZvr60XbfhmjJL4DuPnoi1+yTNZUefiw50xJwK+iZ9zKZVpcHpRQm6EmxuiwQ+2iWIXpegg8vV4IXOp6j+dTSqT05dP7ncpUHCJKY9RgLMbgngGvQnpCJ4ZYK89FmHF+wQgVRtHnY+gkWZgccBN4R matou@catsserver-3

Vagrantfile

@@ -5,17 +5,36 @@ Vagrant.configure("2") do |config|
 	# Get the VagrantFile directory
 	vagrant_root = File.dirname(__FILE__)
 	
-	# Trigger the Vagrant preflight script before uping the first VM only
+	# Trigger the Vagrant pre-up script before uping the first VM only
 	config.trigger.before :up, :vm => ["node-admin"], :append_to_path => ["#{vagrant_root}/scripts"] do
-		run "vagrant-preflight.sh"
+		run "pre-up.sh"
 	end
 
+	# Trigger the Vagrant post-destroy script aftre destroying the last VM only
+	# config.trigger.after :destroy, :vm => ["node-admin"], :append_to_path => ["#{vagrant_root}/scripts"] do
+		run "post-destroy.sh"
+	# end
+
 	# Shell provisionner for all VMs
-	config.vm.provision "shell", path: "scripts/ceph-preflight.sh"
+	config.vm.provision "shell", path: "scripts/provision.sh"
 	
 	# All VMs are based on the same box
 	config.vm.box = "bento/ubuntu-16.04"
 
+	# Use nfs for shared folder
+	config.vm.synced_folder ".", "/vagrant",
+  		nfs: true,
+		linux__nfs_options: ['rw','no_subtree_check','all_squash','async'],
+		nfs_version: 4,
+		nfs_udp: false
+
+	# Public host bridge
+	# config.vm.network "public_network", 
+	#	network_name: "public-network",
+	#	dev: "br0", 
+	#	type: "bridge", 
+	#	mode: "bridge"
+
 	# Standard configuration for all VMs
 	config.vm.provider :libvirt do |libvirt|
 		libvirt.memory = 1024
@@ -30,16 +49,17 @@ Vagrant.configure("2") do |config|
 	end
 
 	# osd1 VM with private cluster network
-	# 5 additional disks: 1 for ceph journal and 4 for osd
+	# 3 additional disks: 1 for journals and 2 for osd
 	config.vm.define "node-osd1" do |osd1|
 		osd1.vm.hostname = "node-osd1"
 		osd1.vm.network :private_network,
 			:type => "dhcp",
+			:mac => "52:54:00:79:1e:b0",  
       		:libvirt__dhcp_start => "172.28.128.10",
       		:libvirt__dhcp_stop => "172.28.128.250",
 			:libvirt__network_address => "172.28.128.0",
 			:libvirt__netmask => "255.255.255.0",
-      		:libvirt__network_name => "cluster-net"
+      		:libvirt__network_name => "cluster-network"
 		osd1.vm.provider :libvirt do |libvirt|
 			libvirt.storage :file, 
 				:size => "30G", 
@@ -57,12 +77,13 @@ Vagrant.configure("2") do |config|
 	end
 
 	# osd2 VM with private cluster network
-	# 5 additional disks: 1 for ceph journal and 4 for osd
+	# 3 additional disks: 1 for journals and 2 for osd
 	config.vm.define "node-osd2" do |osd2|
 		osd2.vm.hostname = "node-osd2"
 		osd2.vm.network :private_network,
 			:type => "dhcp",
-			:libvirt__network_name => "cluster-net"
+			:mac => "52:54:00:dc:51:7c",  
+			:libvirt__network_name => "cluster-network"
 		osd2.vm.provider :libvirt do |libvirt|
 			libvirt.storage :file, 
 				:size => "30G", 

scripts/ceph-install.sh

@@ -27,11 +27,7 @@ VERSION=ceph-install-1.0-120417
 
 # This script is executed in guest context
 source "/vagrant/scripts/cephtest-utils.sh"
-# Directories (depending of the context)
+
-GUEST_USER_DIR="/home/$CEPH_ADMIN_USER"
-GUEST_USER_SSH_DIR="$GUEST_USER_DIR/.ssh"
-GUEST_VAGRANT_SCRIPT_DIR="/vagrant/scripts"
-GUEST_VAGRANT_SSH_DIR="/vagrant/.ssh"
 # Network (dynamically defined by Vagrant)
 IP_ADDRESS=$(ip addr show eth0 | grep "inet\b" | awk '{print $2}' | cut -d/ -f1)
 PUBLIC_NETWORK=$(echo $IP_ADDRESS | awk -F '.' '{print $1"."$2"."$3".0"}')
@@ -42,6 +38,12 @@ if [[ $(whoami) != $CEPH_ADMIN_USER ]]; then
    exit 1
 fi
 
+# Make sure this script is run from the admin node
+if [[ $(hostname -s) != $ADMIN_NODE ]]; then
+   echo "This script must be run from $ADMIN_NODE" 1>&2
+   exit 1
+fi
+
 # Cluster configuration directory
 mkdir -p "$GUEST_USER_DIR/ceph-cluster"
 cd "$GUEST_USER_DIR/ceph-cluster"

scripts/cephtest-utils.sh

@@ -28,7 +28,7 @@ VERSION=cephtest-utils-1.0-120417
 CEPH_ADMIN_USER="ceph-admin"
 CEPH_ADMIN_EXEC="sudo -i -u $CEPH_ADMIN_USER"
 
-# Machines
+# Nodes
 ADMIN_NODE="node-admin"
 OSD_NODES="node-osd1 node-osd2"
 NODES="$ADMIN_NODE $OSD_NODES"
@@ -36,3 +36,18 @@ NODES="$ADMIN_NODE $OSD_NODES"
 # Networks
 CLUSTER_NETWORK="172.28.128.0"
 
+# Guest name
+GUEST_NAME=$(hostname -s)
+
+# Guest directories
+GUEST_USER_DIR="/home/$CEPH_ADMIN_USER"
+GUEST_USER_SSH_DIR="$GUEST_USER_DIR/.ssh"
+GUEST_VAGRANT_DIR="/vagrant"
+GUEST_VAGRANT_SCRIPT_DIR="$GUEST_VAGRANT_DIR/scripts"
+GUEST_VAGRANT_SSH_DIR="$GUEST_VAGRANT_DIR/.ssh"
+GUEST_VAGRANT_SIGNAL_DIR="$GUEST_VAGRANT_DIR/.signals"
+
+# Host directories
+HOST_SCRIPT_DIR="$(dirname "$(readlink -f "$0")")"
+HOST_SSH_DIR="$(readlink -f "$HOST_SCRIPT_DIR/../.ssh")"
+HOST_SIGNAL_DIR="$(readlink -f "$HOST_SCRIPT_DIR/../.signals")"

scripts/post-destroy.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# ========================================================================================
+# Execute post-destroy cleaning
+#
+# Written by    : Denis Lambolez
+# Release       : 1.0
+# Creation date : 16 December 2017
+# Description   : Bash script
+#                 This script has been designed and written on Ubuntu 16.04 plateform.
+#                 It must be executed in vagrant context
+# Usage         : ./post-destroy.sh 
+# ----------------------------------------------------------------------------------------
+# ========================================================================================
+#
+# HISTORY :
+#     Release   |     Date      |    Authors     |       Description
+# --------------+---------------+--------------- +------------------------------------------
+#       1.0     |    12.16.17   | Denis Lambolez | Creation
+#               |               |                | 
+# =========================================================================================
+#set -xev
+
+# Version
+VERSION=post-destroy-1.0-121617
+
+# This script is executed in host context
+source "$(dirname "$(readlink -f "$0")")/cephtest-utils.sh"
+
+# clean-up networks to start with fresh configuration
+for NETWORK in vagrant-libvirt vagrant-private-dhcp; do
+    virsh net-list --all 2> /dev/null | grep $NETWORK | grep active
+    if [[ $? -eq 0 ]]; then
+        virsh net-destroy $NETWORK  2> /dev/null
+    fi
+    virsh net-list --all  2> /dev/null | grep $NETWORK
+    if [[ $? -eq 0 ]]; then
+        virsh net-undefine $NETWORK  2> /dev/null
+    fi
+done

scripts/pre-up.sh

@@ -1,35 +1,32 @@
 #!/bin/bash
 # ========================================================================================
-# Execute preflight configuration needed to deploy vagrant cluster
+# Execute pre-up configuration needed to deploy vagrant cluster
 #
 # Written by    : Denis Lambolez
-# Release       : 1.1
+# Release       : 2.0
-# Creation date : 04 December 2017
+# Creation date : 16 December 2017
 # Description   : Bash script
 #                 This script has been designed and written on Ubuntu 16.04 plateform.
 #                 It must be executed in vagrant context
-# Usage         : ./vagrant-preflight.sh 
+# Usage         : ./pre-up.sh 
 # ----------------------------------------------------------------------------------------
 # ========================================================================================
 #
 # HISTORY :
 #     Release   |     Date      |    Authors     |       Description
 # --------------+---------------+--------------- +------------------------------------------
+#       2.0     |    12.16.17   | Denis Lambolez | Renamed pre-up.sh and linked to post-destroy
 #       1.1     |    12.04.17   | Denis Lambolez | Sourcing parameters from cephtest-utils
 #       1.0     |    12.02.17   | Denis Lambolez | Creation
 #               |               |                | 
-#               |               |                | 
 # =========================================================================================
 #set -xev
 
 # Version
-VERSION=vagrant-preflight-1.0-120217
+VERSION=pre-up-2.0-121617
 
 # This script is executed in host context
 source "$(dirname "$(readlink -f "$0")")/cephtest-utils.sh"
-# Directories (depending of the context)
-HOST_SCRIPT_DIR="$(dirname "$(readlink -f "$0")")"
-HOST_SSH_DIR="$(readlink -f "$HOST_SCRIPT_DIR/../.ssh")"
 
 # (re)Create ssh keys
 rm -f "$HOST_SSH_DIR/$CEPH_ADMIN_USER"-id_rsa*
@@ -43,3 +40,9 @@ for NODE in $NODES; do
     echo -e "Host $NODE\n\tHostname $NODE\n\tUser $CEPH_ADMIN_USER\n\tStrictHostKeyChecking no\n" >> "$HOST_SSH_DIR/$CEPH_ADMIN_USER-config"
 done
 chmod 644 "$HOST_SSH_DIR/$CEPH_ADMIN_USER-config"
+
+# Clean up IP and PROVISION signals
+for NODE in $NODES; do
+    rm -f "$HOST_SIGNAL_DIR/$NODE-IP"
+    rm -f "$HOST_SIGNAL_DIR/$NODE-PROVISION"
+done

scripts/provision.sh

@@ -3,35 +3,32 @@
 # Execute preflight configuration needed to deploy ceph distributed storage
 #
 # Written by    : Denis Lambolez
-# Release       : 1.1
+# Release       : 2.0
-# Creation date : 04 December 2017
+# Creation date : 17 December 2017
 # Description   : Bash script
 #                 This script has been designed and written on Ubuntu 16.04 plateform.
 #                 It must be executed in privileged mode
-# Usage         : ./ceph-preflight.sh
+# Usage         : ./provision.sh
 # ----------------------------------------------------------------------------------------
 # ========================================================================================
 #
 # HISTORY :
 #     Release   |     Date      |    Authors     |       Description
 # --------------+---------------+--------------- +------------------------------------------
+#       2.0     |    12.17.17   | Denis Lambolez | Adding /etc/hosts modification and 
+#               |               |                | synchronization with other nodes. Renamed 
+#               |               |                | to provision.sh 
 #       1.1     |    12.04.17   | Denis Lambolez | Sourcing parameters from cephtest-utils
 #       1.0     |    12.02.17   | Denis Lambolez | Creation
 #               |               |                | 
-#               |               |                | 
 # =========================================================================================
 #set -xev
 
 # Version
-VERSION=ceph-preflight-1.0-120217
+VERSION=provision-2.0-121717
 
 # This script is executed in guest context
 source "/vagrant/scripts/cephtest-utils.sh"
-# Directories (depending of the context)
-GUEST_USER_DIR="/home/$CEPH_ADMIN_USER"
-GUEST_USER_SSH_DIR="$GUEST_USER_DIR/.ssh"
-GUEST_VAGRANT_SCRIPT_DIR="/vagrant/scripts"
-GUEST_VAGRANT_SSH_DIR="/vagrant/.ssh"
 
 # Make sure only root can run the script
 if [[ $EUID -ne 0 ]]; then
@@ -39,8 +36,11 @@ if [[ $EUID -ne 0 ]]; then
    exit 1
 fi
 
-# Create user ceph-admin
+# Create user ceph-admin if not existing
+cat /etc/passwd | grep $CEPH_ADMIN_USER
+if [[ $? -ne 0 ]]; then
     useradd -m -s /bin/bash $CEPH_ADMIN_USER
+fi
 
 # Make ceph-admin passwordless sudoer
 echo "$CEPH_ADMIN_USER ALL = (root) NOPASSWD:ALL" | tee "/etc/sudoers.d/$CEPH_ADMIN_USER"
@@ -72,15 +72,15 @@ apt-get update
 # vnstat for network stats, htop for system monitor and ceph-deploy
 apt-get -y install chrony gdisk vnstat htop ceph-deploy
 
-# Modify /etc/hosts to allow ceph-deploy to resolve the guest
+# Modify /etc/hosts to allow ceph-deploy to resolve the guests
-# Need to replace the loopback address by the real address
-GUEST_NAME=$(hostname -s)
 IP_ADDRESS=$(ip addr show eth0 | grep "inet\b" | awk '{print $2}' | cut -d/ -f1)
+# Need to replace the loopback address by the real address
 sed -i "s/127.0.0.1\t$GUEST_NAME\t$GUEST_NAME/$IP_ADDRESS\t$GUEST_NAME\t$GUEST_NAME/g" /etc/hosts
 
 # Create partitions on journal disk for osd nodes only
 for NODE in $OSD_NODES; do
-    if [[ NODE == $GUEST_NAME ]]; then
+    if [[ $NODE == $GUEST_NAME ]]; then
+        sgdisk --zap-all
         sgdisk --new=0:0:10G /dev/vda > /dev/null 2>&1
         sgdisk --new=0:0:20G /dev/vda > /dev/null 2>&1
         sgdisk --largest-new=0 /dev/vda > /dev/null 2>&1
@@ -91,3 +91,50 @@ done
 # Full update
 #apt-get -y dist-upgrade
 #apt-get -y autoremove
+
+# Signal that IP is ready
+echo -e "$IP_ADDRESS\t$GUEST_NAME" > "$GUEST_VAGRANT_SIGNAL_DIR/$GUEST_NAME-IP"
+
+# Wait for all nodes IP and update /etc/hosts
+TIMER_MAX=300
+echo >> /etc/hosts
+for NODE in $NODES; do
+    if [[ $NODE != $GUEST_NAME ]]; then
+        TIMER=0
+        until [[ -r "$GUEST_VAGRANT_SIGNAL_DIR/$NODE-IP" ]]; do
+            sleep 1
+            TIMER=$(($TIMER + 1))
+            if [[ $TIMER -gt $TIMER_MAX ]]; then
+                echo "Can't get IP from $NODE" >&2
+                exit 1
+            fi
+        done
+        # Remove record if existing
+        sed -i "/$NODE/d" /etc/hosts
+        # Add new record
+        cat "$GUEST_VAGRANT_SIGNAL_DIR/$NODE-IP" >> /etc/hosts
+    fi
+done
+
+# Signal that provision is done
+echo "$(date --rfc-3339=ns) - Done!"  | tee "$GUEST_VAGRANT_SIGNAL_DIR/$GUEST_NAME-PROVISION"
+
+# Continue provisioning on admin node only
+[[ $GUEST_NAME != $ADMIN_NODE ]] && exit 0
+
+# Wait for all nodes to be ready
+TIMER_MAX=300
+for NODE in $NODES; do
+    TIMER=0
+    until [[ -r "$GUEST_VAGRANT_SIGNAL_DIR/$NODE-PROVISION" ]]; do
+        sleep 1
+        TIMER=$(($TIMER + 1))
+        if [[ $TIMER -gt $TIMER_MAX ]]; then
+            echo "Waited too long for $NODE!" >&2
+            exit 1
+        fi
+    done
+done
+ 
+# Install ceph in ceph-admin context
+$CEPH_ADMIN_EXEC $GUEST_VAGRANT_SCRIPT_DIR/ceph-install.sh