Chef Standalone vs Tiered vs High Availability


I am trying to find out differences between three different types of Chef deployments:

  1. Standalone Deployment
  2. Tiered Deployment
  3. High Availability Deployment

Correcct me if I am wrong, I noticed that Chef High availability server needs a different installable but Standalone and Tiered uses same installable. Then what is the difference between these two?


Sorry but I can’t do best than directing your to read this.

Standalone is all on one machine, tiered is multiple machines to separate computing resources (cpu/memory) needs.

For example with chef-server components: API server (nginx/erchef) on machine 1, postgres DB on machine 2, solr on machine 3, bookshelf on machine 4.

A HA deployment is usually two standalone instances with a replication system to keep both in sync. IIRC the chef-server HA architecture is a failover mode, if the primary (active) server fail, the second one (passive) become active and take over the various floating IPs.

Thanks @Tensibai
What about licensing of Tiered and HA?
Are they free for 25 nodes or they follow different licensing?

As far as I know there’s no licensing requirement for the HA with DRBD, if you’re looking at the only architecture speaking about premium feature is the HA with AWS.

All premium features are free under 25 nodes as far as I know (analytics and delivery are separated products, I’m quite unsure there).

You can still contact Chef Inc. on for licensing questions, or wait until someone from Chef pass here :slight_smile:

Yeah, I also read in that “You can use Chef server for free — and premium features are free for up to 25 nodes.”

Thanks Tensibai. I will try contacting them if nobody from Chef passes this discussion. :slight_smile:

Woah, is HA with DRBD a non-premium feature for reals? If so, this might
number the days of our multiple chef server topology.

You might not want to use DRBD – see this post from 2014 by (I think) a Chef support engineer, where DRBD is specifically not recommended. (Look for the headline that says “DRBD: Don’t do it.” :smile: )

HA with DRBD is a non-premium feature for reals but as @saintaardvark points out, it is not a very reliable/operable setup and our customers don’t like it.

We intend to ship a better HA solution in Q1 2016 and it will be a premium feature.

Hello, author of the aforementioned article here. We have a lot of customers that run the DRBD-HA mode and are happy with it, but in those cases both of the following are true:

  1. They have operational experience with DRBD and use it in other parts of the infrastructure
  2. They run physical servers (not VMs) and do things to guarantee high network bandwidth at low latency between their backends (ex: 10GbE NICs cross-connected between the two backends)

As I stated in the article, DRBD+keepalived are great at protecting you against hardware failure. But nowadays most of our customers choose to run their Chef servers in virtualized environments where hardware fault tolerance is handled at a lower layer, which makes the complexity (and performance hit) of DRBD undesirable.

1 Like