Product Announcements

Does a Reverse Web Proxy Protect Against An Auto Deploy Server Outage?

Kyle Gleed, Sr. Technical Marketing Architect, VMware

A while back I did a write-up on using reverse web proxy servers with auto deploy.  At the time I thought that using a reverse web proxy would not only help with scalability, but would also help protect against an outage affecting the auto deploy server.  However, I’ve learned this is not the case – using a reverse web proxy does not protect against an auto deploy server outage.

There are two parts to the auto deploy server, a rules engine and a web server.  In setting up a reverse web proxy we are creating a redundant web server, but we still only have one rules engine, and that’s on the auto deploy server. 

The way the communication flows when using a reverse web proxy with auto deploy is as follows:

  1. ESXi host PXE boots and contacts the auto deploy server.
  2. Host is processed against the rules engine to identify the image profile (VIBs to be installed), host profile, and vCenter location.
  3. Host is passed off to the reverse web proxy where the image profile is copied and the host profile applied (note: existing hosts get the host profile from the web proxy, new hosts will get the host profile from the auto deploy server).
  4. The host informs the auto deploy server that it has been installed and is ready to be added/reconnected to vCenter.

So even with a reverse web proxy, if the auto deploy server becomes unavailable you will not be able to deploy new hosts or reboot existing hosts.  The role of the reverse web proxy is limited to delivery of the image profile and host profile.  The auto deploy server is still needed at the beginning of the boot operation to parse the rules, and at the end where it connects the host to vCenter.

So in summary, using reverse web proxy servers with auto deploy helps with scalability by offloading the task of streaming the image profiles to the ESXi hosts, but it does not protect against an outage of the auto deploy server.  

To address HA of the auto deploy server my recommendation is to run the auto deploy server in a separate “management” cluster (along with vCenter and other core infrastructure VMs) and to leverage VMware HA/DRS to ensure availability.  VMware Tech Pubs has a couple of short videos that give a great overview on using reverse web proxy with auto deploy as well as how to implement a highly available auto deploy infrastructure.  Definitely worth checking out!

Get notified on future  posts by following me on twitter @VMwareESXi