Skip to content

Supported Deployments

Meni Vaitsi edited this page May 14, 2016 · 3 revisions

Introduction

We currently support a subset of all of the possible deployments that AppScale is capable of. In this document we present and describe the layout of these deployments. If you are interested in other deployments, feel free to contact us via IRC and we can help ensure that it works well for you.

There are two types of AppScale deployments: cluster and cloud. Cluster deployments are those in which you start instances of the AppScale VM, verify that you can ssh into them as root, and extract the IP address of each. Cloud deployments are Amazon EC2, Google Compute Engine, and Eucalyptus deployments for which AppScale controls the starting and stopping of VM instances. Cloud deployments assume that there is a registered AppScale image available within the infrastructure of choice. Cloud and cluster deployments can have multiple nodes (VM instances).

Single Node (Test/Dev) Deployments

For single node deployments, the default layout is a single node with all AppScale components (roles) deployed in a single instance. For cluster deployments, the configuration is:

ips_layout:
  controller: 192.168.10.2

For cloud deployments, the configuration is:

min: 1
max: 1

Production Deployments

For a 4-node cluster, the simple, default configuration is:

ips_layout:
  controller: 192.168.10.2
  servers:
  - 192.168.10.3
  - 192.168.10.4
  - 192.168.10.5

For a 8-node cluster, simply add your additional four IP addresses to the end of server list using the same format.

For cloud, the simple configuration is:

min: 4
max: 4

For a 8-node cluster, change these values to 8.

For a 4-node cluster, the advanced configuration is:

ips_layout:
 master : 192.168.10.2
 appengine:
  - 192.168.10.3
 database:
  - 192.168.10.4
 zookeeper:
  - 192.168.10.5

Change the IP addresses to node-X for X=1,2,3,4 for an advanced cloud deployment. In this deployment, please note that since there is only one zookeeper instance, it is a single point of failure for the system. To overcome this, you can place two additional zookeeper roles across the cloud, e.g. on the master and database nodes. Zookeeper requires instances across a majority of nodes.

For a 8-node cluster, the advanced configuration is:

ips_layout:
 master : 192.168.10.2
 appengine:
  - 192.168.10.3
  - 192.168.10.4
 database:
  - 192.168.10.5
  - 192.168.10.6
 zookeeper: 
  - 192.168.10.7
  - 192.168.10.8
  - 192.168.10.9

There are three zookeeper nodes since zookeeper requires a majority of nodes to be zookeeper nodes in multi-node settings. Change the IP addresses to node-X for X=[1,2,3,4,5,6,7,8] for cloud deployments.

Clone this wiki locally