Contrail DB Split with Mirantis Fuel for releases less than or equal to 3.0.2

Introduction

The recommended best practice for production deployment of Contrail is to separate database for Contrail Config and Contrail Analytics.

The Contrail Analytics along with Analytics DB would be installed on a separate node (physical host or VM) while Contrail Config along with Config DB would be a separate node.

Below are the recommended roles:

Contrail Controller: Contrail Config, Config DB (Config Cassandra and zookeeper), Control, WebUI
Contrail Analytics: Contrail Analytics, Analytics DB (Analytics Cassandra and kafka) 

Mirantis Fuel plugin roles

At present, Fuel has the following roles for Contrail:

contrail-config 
contrail-control
contrail-db 

For splitting up Cassandra for Config and Analytics, we are suggesting to have following roles in Fuel plugin:

contrail-config: This role is contrail config along with config DB (Cassandra and zookeeper) 
contrail-control: Contrail control node 
contrail-analytics: This would be a new role. This is contrail analytics node.
contrail-db: This role would now refer to only Contrail Analytics DB. 

Fab based deployment steps:

Below are the provisioning steps based on current fab scripts (available in 3.0.2-51) to setup the above-described deployment:

1. Populate the testbed.py such that same host(s) have analytics and database role

#Role definition of the hosts.
env.roledefs = {
    'all': [host1, host2, host3, host4, host5, host6, host7, host8, host9],
    'cfgm': [host1, host2, host3],
    'openstack': [host1, host2, host3],
    'control': [host1, host2, host3],
    'compute': [host8, host9],
    'collector': [host4, host5, host6],
    'webui': [host1, host2, host3],
    'database': [host4, host5, host6],
    'build': [host_build],
    #'storage-master': [host1],
    #'storage-compute': [host4, host5, host6, host7, host8, host9, host10],
    #'rally': [host11], # Optional, to enable/setup rally, it can be a seprate node from contrail cluster
    # 'vgw': [host4, host5], # Optional, Only to enable VGW. Only compute can support vgw
    # 'tsn': [host1], # Optional, Only to enable TSN. Only compute can support TSN
    # 'toragent': [host1], Optional, Only to enable Tor Agent. Only compute can
    # support Tor Agent
    #   'backup':[backup_node],  # only if the backup_node is defined
}

2. Install contrail packages in all the nodes

fab install_pkg_all:<path/to/contrail-install-packages>
fab install_contrail 

3. Edit the testbed.py such that host having config role(s) also have database roles

#Role definition of the hosts.
    env.roledefs = {
    'all': [host1, host2, host3, host4, host5, host6, host7, host8, host9],
    'cfgm': [host1, host2, host3],
    'openstack': [host1, host2, host3],
    'control': [host1, host2, host3],
    'compute': [host8, host9],
    'collector': [host4, host5, host6],
    'webui': [host1, host2, host3],
    'database': [host1, host2, host3],
    'build': [host_build],
    #'storage-master': [host1],
    #'storage-compute': [host4, host5, host6, host7, host8, host9, host10],
    #'rally': [host11], # Optional, to enable/setup rally, it can be a seprate node from contrail cluster
    # 'vgw': [host4, host5], # Optional, Only to enable VGW. Only compute can support vgw
    # 'tsn': [host1], # Optional, Only to enable TSN. Only compute can support TSN
    # 'toragent': [host1], Optional, Only to enable Tor Agent. Only compute can
    # support Tor Agent
    #   'backup':[backup_node],  # only if the backup_node is defined
}

4. Install database package in config node and Provision the cluster with config node as DB node.

fab install_database  
fab setup_database  
fab verify_database 
fab setup_common
fab setup_ha
fab setup_rabbitmq_cluster
fab increase_limits
fab setup_orchestrator
fab setup_cfgm
fab verify_cfgm 
fab setup_control
fab verify_control 

5. Change the testbed.py back to as shown in #1

fab setup_database
fab -R database -- "sed -i 's/zookeeper.connect=.*/zookeeper.connect=<host1>:2181,<host2>:2181,<host3>:2181/g'  /usr/share/kafka/config/server.properties"
fab -R database -- "service zookeeper stop"
fab -R database -- 'echo "manual" > /etc/init/zookeeper.override'
fab restart_database
fab verify_database
fab setup_collector
fab -R collector -- "openstack-config --set /etc/contrail/contrail-alarm-gen.conf DEFAULTS zk_list <host1>:2181 <host2>:2181 <host3>:2181"
fab -R collector -- "openstack-config --set /etc/contrail/contrail-collector.conf DEFAULT zookeeper_server_list <host1>:2181,<host2>:2181,<host3>:2181"
fab -R collector -- "openstack-config --set /etc/contrail/contrail-snmp-collector.conf DEFAULTS zookeeper <host1>:2181,<host2>:2181,<host3>:2181"
fab -R collector -- "openstack-config --set /etc/contrail/contrail-topology.conf DEFAULTS zookeeper <host1>:2181,<host2>:2181,<host3>:2181"
fab restart_collector
fab setup_webui
fab verify_webui 
fab setup_vrouter
fab prov_config
fab prov_database
fab prov_analytics
fab prov_control_bgp
fab prov_external_bgp
fab prov_metadata_services
fab prov_encap_type
fab setup_remote_syslog 
fab increase_vrouter_limits
#if there is openstack internal VIP,  do fab setup_cluster_monitors
fab compute_reboot
fab verify_compute
fab setup_nova_aggregate