This document is meant to be used by RackConnect "power users". Power users typically use the API to build new servers and need to be able to spin up large numbers of servers quickly. For power users, the time it takes to spin up many servers is very important, thus build times and build success rates are key success measurements.
"Burst" the server builds - A RackConnect configuration represents a deployment of the RackConnect solution. The configuration is comprised of your Edge and Connected physical network devices, along with one or more Cloud Accounts that you associate with the configuration. A single RackConnect can support up to 200 cloud servers per configuration based on network device.
We recommend that you create no more than 20 Windows servers or 30 Linux servers at a time. It is best to create all 20 or 30 simultaneously, rather than issuing the API call to create one, waiting for a while, and then issuing the command for the next. If you need to build more than 20 or 30 servers, wait until the first batch is complete before kicking off the next batch. RackConnect has an API you can use to determine when the last server is fully deployed. More information on how to programatically determine the RackConnect automation status for cloud servers can be found here: How to programmatically determine the RackConnect Automation Status of your Cloud Servers
Vet the image - Once the basic cloud server build has completed, RackConnect logs onto the newly created server and performs some configuration changes. For RackConnect 2.x customers, the automation creates a RackConnect "user", disables the public interface, reconfigures the private interface so that it can connect to your dedicated environment, and configures the software firewall according to your network policies. If you are not using a standard Rackspace image that has been verified to work with RackConnect, there are a number of configuration changes that could be made that may prevent RackConnect automation from running successfully. For example, images that have SSH disabled, have SSH configured for a non-standard port, or prevent root login via SSH will all fail to build correctly. Please refer to RackConnect Best Practices for more information about image configuration: RackConnect Best Practices
Allow Rackspace automation to complete - If you are running your own automation scripts after the cloud server build is complete, be sure to allow all Rackspace automation to complete before running your scripts. If your automation and Rackspace automation runs at the same time it could cause one or both to fail. For customers with a Managed Cloud Service Level, there are two automation passes that Rackspace runs: one for RackConnect and one for the managed Cloud Servers. If you are not a Managed Cloud Service Level customer, you only need to wait for the RackConnect automation to complete before running your own. To programmatically determine when Rackspace automation is complete refer to this guide: How to programmatically determine the RackConnect Automation Status of your Cloud Servers
Build cloud servers in the correct region - Make sure that you create your cloud servers in the same region as your RackConnect configuration. In other words, if your dedicated environment is in the Rackspace DFW region be sure to build any cloud servers you want to be RackConnected in the same region. Servers build in a different region will not be accessible from your dedicated environment.
Use manual disk mode - The Rackspace next generation Cloud Servers API has an extension called "diskConfig" that allows you to control how the cloud server disk is partitioned when servers are created, rebuilt, or resized. There are two modes: manual and auto.
Manual disk mode creates a 20GB partition for the OS, so unless you need more than 20GB of space you should use manual mode as this will decrease the server build time significantly. You can also extend the partition manually post-build.
© 2011-2013 Rackspace US, Inc.
Except where otherwise noted, content on this site is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License