Merge pull request #155 from CiscoSystems/aio_install_script

Update README and master.sh script to support collapsed build

README.md

@@ -123,57 +123,71 @@ To create a basic 2 role cluster with a build, compute and control node outside
 
 More information about this tool can be found under the stack-builder directory.
 
-## basic install against already provisioned nodes:
+# Basic install against already provisioned nodes (Ubuntu 12.04.3 LTS):
 
-### install your build server
+### install your All-in-one Build, Control, Network, Compute, and Cinder node:
 
-first, log into your build server, and run the following script to bootstrap it as a puppet master:
+These instructions assume you will be building against a machine that has two interfaces:
 
-If you want to use the havana version of the packages, set the following ENV var:
+'eth0' for management, and API access, and also to be used for GRE/VXlan tunnel via OVS
+'eth1' for 'external' network access (in single provider router mode).  This interface
+is expected to provide an external router, and IP address range, and will leverage the
+l3_agent functionality to provide outbound overloaded NAT to the VMs and 1:1 NAT with 
+Floating IPs.  The current default setup also assumes a very small "generic" Cinder 
+setup, unless you create an LVM volume group called cinder-volume with free space
+for persistent block volumes to be deployed against.
 
-    export openstack_version=havana
+Log in to your all_in_one node, and bootstrap it into production:
 
     bash <(curl -fsS https://raw.github.com/CiscoSystems/openstack-installer/master/install-scripts/master.sh)
 
-### set up your data
+You can over-ride the default parameters, such as ethernet interface names, or hostname, and default ip address if you choose:
 
-on your build server, all of the data you may need to override can be found in:
+scenario           : change this to a scenario defined in data/scenarios, defaults to all_in_one
+build_server       : Hostname for your build-server, defaults to `` `hostname` ``
+domain_name        : Domain name for your system, defaults to `` `hostname -d` ``
+default_interface  : This is the interface name for your management and API interfaces (and tunnel endpoints), defautls to eth0
+external_interface : This is the interface name for your "l3_agent provider router external network", defaults to eth1
+build_server_ip    : This is the IP that any additional devices can reach your build server on, defaults to the default_interface IP address
+ntp_server         : This is needed to keep puppet in sync across multiple nodes, defaults to ntp.esl.cisco.com
+puppet_run_mode    : Defaults to apply, and for AIO there is not a puppetmaster yet.
 
-    /etc/puppet/data/hiera_data/user.common.yaml
+To change these parameters, do something like:
 
-at the very least, you may need to update the controller ip addresses and set the
-interfaces to use.
+scenario=2_role bash <(curl.....master.sh)
 
-Look at the puppet certnames that map to roles in:
+### add additional nodes
 
-    /etc/puppet/data/role_mappings.yaml
+Adding additional nodes is fairly straight forward (for all_in_one scenarion compute nodes can be added, others require a bit of additional effort by expanding the all_in_one scenario)
 
-You may also find a need to change the default scenario in:
+1) on the All-in-one node, add a role mapping for the new node:
 
-    /etc/puppet/data/config.yaml
+echo "compute_node_name: compute" >> /etc/puppet/data/role_mappings.yaml
 
-Choices are in:
+2) Build the phyiscal or virtual compute node
+
+3) Configure the system to point ot the all_in_one node for puppet deployment and set up the right version of puppet on the node:
+
+    export build_server_ip=X.X.X.X ; bash <(curl -fsS https://raw.github.com/CiscoSystems/openstack-installer/master/install-scripts/setup.sh)
+
+After which you may still have to run puppet in "agent" mode to actually deploy the openstack elements:
+
+``
+    puppet agent -td --server build-server.`hostname -d` --certname `hostname -f`
+``
+
+### If other role types are desired
+
+At the scenario leve, choices are in:
 
     /etc/puppet/data/scenarios
 
+And you can extend the all_in_one scenario, or leverage a different variant all together.
+
 Defaults for end user data should be located in one of the following files:
 
     /etc/puppet/data/hiera_data/user.yaml
     /etc/puppet/data/hiera_data/user.common.yaml
     /etc/puppet/data/hiera_data/user.<scenario>.yaml
 
-### install each of your components
-
-first setup each node (unless you're doing all\_in\_one scenario, in which case you'll already have done this from the previous step):
-
-    export build_server_ip=X.X.X.X ; bash <(curl -fsS https://raw.github.com/CiscoSystems/openstack-installer/master/install-scripts/setup.sh)
-
-then log into each server, and run:
-
-``
-    puppet agent -td --server build-server.`hostname -d` --certname `hostname -f`
-``
-
-where build-server is the fully qualified name of the build server (or `` `hostname -f` `` on an all-in-one node), or its IP address that was set in user.common.yaml and ROLE\_CERT\_NAME is the fully qualified name of the local machine (or `` `hostname -f` `` which should return the same thing)
-
-*NOTE: you'll want to run the puppet agent command on any control class nodes (or the all-in-one node) first, before running it on any compute or storage nodes.*
+###Additional information on the data model being leveraged is available in the data directory of this repository.

data/hiera_data/common.yaml

@@ -48,3 +48,7 @@ openstack::swift::storage-node::storage_devices:
   - 3
 
 apache::mpm_module: prefork
+
+quantum::agents::ovs::local_ip: "%{ipaddress}"
+neutron::agents::ovs::local_ip: "%{ipaddress}"
+

install-scripts/master.sh

@@ -4,14 +4,65 @@
 # it runs on into a puppetmaster/build-server
 #
 
-export build_server_ip="${build_server_ip:-127.0.0.1}"
+# All in one defaults to the local host name for pupet master
+export build_server="${build_server:-`hostname`}"
+# It'd be good to konw our domain name as well
+export domain_name=`hostname -d`
+# We need to know the IP address as well, so either tell me
+# or I will assume it's the address associated with eth0
+export default_interface="${default_interface:-eth0}"
+# So let's grab that address
+export build_server_ip="${build_server_ip:-`ip addr show ${default_interface} | grep 'inet ' | tr '/' ' ' | awk -F' ' '{print $2}'`}"
+# Our default mode also assumes at least one other interface for OpenStack network
+export external_interface="${external_interface:-eth1}"
+
+# For good puppet hygene, we'll want NTP setup.  Let's borrow one from Cisco
+export ntp_server="${ntp_server:-ntp.esl.cisco.com}"
+
+# Since this is the master script, we'll run in apply mode
 export puppet_run_mode="apply"
 
+# scenarios will map to /etc/puppet/data/scenarios/*.yaml
+export scenario="${scenario:-all_in_one}"
+
+sed -e "s/2_role:*/$scenario/" -i /root/openstack-installer/data/config.yaml
+
+if [ "${scenario}" == "all_in_one" ] ; then
+  echo `hostname`: all_in_one >> /root/openstack-installer/data/role_mappings.yaml
+  export FACTER_build_server_ip=${build_server_ip}
+  export FACTER_build_server=${build_server}
+  cat > /root/openstack-installer/data/hiera_data/user.yaml<<EOF
+domain_name: "${domain_name}"
+ntp_servers:
+  - ${ntp_server}
+
+# node addresses
+build_node_name: ${build_server}
+controller_internal_address: "${build_server_ip}"
+controller_public_address: "${build_server_ip}"
+controller_admin_address: "${build_server_ip}"
+swift_internal_address: "${build_server_ip}"
+swift_public_address: "${build_server_ip}"
+swift_admin_address: "${build_server_ip}"
+
+# physical interface definitions
+external_interface: ${external_interface}
+public_interface: ${default_interface}
+private_interface: ${default_interface}
+
+internal_ip: "%{ipaddress}"
+nova::compute::vncserver_proxyclient_address: "%{ipaddress}"
+swift_local_net_ip: "%{ipaddress}"
+nova::compute::vncserver_proxyclient_address: "0.0.0.0"
+EOF
+fi
+
 bash <(curl -fsS https://raw.github.com/CiscoSystems/openstack-installer/master/install-scripts/setup.sh)
 
-cp -Rv /root/openstack-installer/modules /etc/puppet/
-cp -Rv /root/openstack-installer/data /etc/puppet/
-cp -Rv /root/openstack-installer/manifests /etc/puppet/
+cp -R /root/openstack-installer/modules /etc/puppet/
+cp -R /root/openstack-installer/data /etc/puppet/
+cp -R /root/openstack-installer/manifests /etc/puppet/
 
-puppet apply /etc/puppet/manifests/site.pp --certname build-server --debug
+
+puppet apply /etc/puppet/manifests/site.pp --certname ${build_server} --debug
 puppet plugin download --server `hostname -f`; service apache2 restart

install-scripts/setup.sh

@@ -21,7 +21,7 @@ else
 fi
 # puppet's fqdn fact explodes if the domain is not setup
 if grep 127.0.1.1 /etc/hosts ; then
-  sed -i -e "s/127.0.1.1.*/127.0.1.1 $(hostname).$domain $(hostname)/" /etc/hosts\n
+  sed -i -e "s/127.0.1.1.*/127.0.1.1 $(hostname).$domain $(hostname)/" /etc/hosts
 else
   echo "127.0.1.1 $(hostname).$domain $(hostname)" >> /etc/hosts
 fi;

manifests/setup.pp

@@ -53,6 +53,10 @@ if $::puppet_run_mode != 'agent' {
       before  => Package['puppet'],
       require => Package['puppet-common']
     }
+    package { 'puppetmaster-passenger':
+      ensure  => $puppet_version,
+      require => Package['puppet'],
+    }
   }
 
   # set up our hiera-store!