- 400 * (1 Core , 2 GB) VM instances
- 10Gb Redundant Network Between VMs
- Controller HA
CPU overcommit ratio : 1:6
RAM overcommit ratio : 1:1.5
400 * 1 Core/6 = 66 physical cores
400 * 2 GB/1.5 = 533 GB physical Memory
4 * Compute nodes each with 16 cores CPU and 16 * 8 GB of RAM
Each node with 2 * 10Gb NIC and 1 * 1Gb management NIC
3 * Controller nodes each with 16 cores CPU and 8 * 8 GB of RAM plus LSI SAS 9207-8E(HBA) for JBOD ST7110-30A connection.
>> SV7220-2S with mezzanine card(Intel x520 DA2) is selected.
3 * ST7110-30A each with 15 * 2TB SATA HDDs to provide space for glance images and cinder volume.
1 * Edge-core AS4600-54T for pXe and ipmi/dcmi network
2 * Edge-core AS5600-52X for OpenStack related network
- Rack, servers, switches and power preparing
- Switch configuration
Controller Node Cabling
Compute Node Cabling
1G Switch (1~8 port on default vlan 1 Untagged)
10G Switches (1~7 port for default vlan 1 Untagged, vlan 101 tagged, vlan 102 tagged and 2 * 10G switches are stacking)
Fuel Master Node Preparation
- Fuel is a graphical web tool that deploys OpenStack cloud environment.
- Install Fuel on any x86 bare metal machine or VM. Refer to Fuel Documentation.
- Plug Fuel interface to 1G switch port.
Defining OpenStack Environment
Login and create environment
- Make sure all the target nodes’ BIOS setting are configured to be PXE boot and they are within Fuel Administrative or PXE network. Power on all the target nodes which will be used as controller and compute nodes. These nodes will PXE boot and load bootstrap image from Fuel server. The bootstrap OS will then collect the node’s hardware information like processor, memory, hard drive and network interfaces for OpenStack environment defining.
- Click “New OpenStack environment”
- The following options will be selected
- Base OS : Ubuntu 12.04
- Multi-node with HA
- Hypervisor type in compute node : kvm
- Neutron with Gre segmentation
- Storage configuration : Cinder Default , Glance Default
- No additional services chosen
- Click “Create” to create an empty OpenStack environment.
Defining OpenStack Networking
Configure the kernel parameter
- SV7220 server is non-VGA. ”console=ttyS0,57600n8” will redirect console from debug card to laptop for troubleshooting.
- Click “Setting” tab and edit as the following
Assign Role to target nodes
- Click “Add Nodes” button to assign the Controller and Compute role to available node list(Cinder storage is also included in controller node).
Assign networks to node’s interfaces(NICs)
- Drag and drop the boxes with network names to specific interface. The example bonded 10Gbe network interfaces as Balance SLB mode to provide storage network, management network and neutron network and 1Gbe network interface to provide PXE network. Make sure the two 10G switches are configured to be stacking when using Balance SLB mode. Or use Active Backup bonding mode which does not need further configuration of switch.
Define Disk Partition for nodes
- Define disk layout for Cinder and Glance services on Controller nodes.
- Define disk layout for Compute nodes.
Verify the Network and Start Deploy
- Click “Networks” tab and “Verify Networks” botton to verify the predefined networks.
- If the network verification completed without error, then click “Deploy Change” to start deploy OpenStack.
- Fuel takes two stages to deploy OpenStack : Stage1: Base OS deployment > Reboot > Stage2: OpenStack services deployment. It takes about 1~2 hours to complete the whole deployment process.
- Once completed, you will see success message and Horizon Dashboard IP address wil be printed on Fuel’s Web UI.
- Click “Health Check” and select the function to be validated.
- Login Horizon Dashboard using Chrome or Firefox browser. (Default user and password is admin/admin)
- Basically the Fuel UI and CLI is enough to provision and manage the new OpenStack environment.If you want to deep dive and troubleshoot the environment, please see Dockers section in Mirantis OpenStack Operation Guide. For example “#docker ps” lists all containers and “#dockerctl shell cobbler” logins cobbler container. Cobbler container serves dhcp/tftp/dns/http functions.We usually login the cobbler container and edit the “/var/lib/tftpboot/pxelinux.cfg/default” file to manipulate bootstrap’s initial ram disk behavior.…LABEL bootstrap kernel /images/bootstrap/linux MENU LABEL bootstrap append initrd=/images/bootstrap/initramfs.img ksdevice=bootif lang= locale=en_US text mco_user=mcollective priority=critical url=http://10.10.10.1:8000/api biosdevname=0 mco_pass=pjTDlDJO kssendmac ks=http://10.10.10.1/cblr/svc/op/ks/profile/bootstrap intel_pstate=disable console=ttyS0,57600n8…intel_pstate=disable : Disable the intel_pstate driver at boot
console=ttyS0,57600n8 : Wiwynn SV7220 servers are non-VGA, so we need to add this parameter to redirect console from debug card to laptop for troubleshooting.