Hi folks,
I have a terminology question.
Rancher 2 ‘clusters’ usually consist of multiple Kubernetes clusters - One cluster hosts Rancher itself, and other clusters host the user applications.
I’ve been calling this the “Rancher 2 cluster”, which is sort of confusing— it’s not really a cluster, it’s a collection of clusters managed by Rancher.
What do you call your Rancher thing? Do you call it a “Rancher Deployment”, a “Rancher Installation”, “Rancher Cluster”?
-= Stefan
Well they are actually separate clusters, each with all of the node types in their own right. The first is the Rancher management cluster (some people refer to this as Rancher HA, but IMO that’s a mis-nomen), and and the other(s) are application workload clusters.
Typically you have one Rancher management cluster for ‘n’ application workload clusters.
Typically you have one Rancher management cluster for ‘n’ application workload clusters.
Right. If you consider all of these clusters together, they are a ‘thing’. What do you call that thing?
Hey Stefan,
Well, whatever you want I suppose, AFAIK there probably isn’t a collective terms for a group of clusters that would necessarily be meaningful to everyone. You might consider this group of resources to represent a ‘platform’. We often refer to ours as the UK Digital Container Platform, not wanting particularly to attach technical terms to what we like to consider as a highly disposable/temporal/ephemeral (pick your favourite adjective) set of compute, network and storage resource provided to support a specific type of workload. Others might refer to them as a platform level service or even an application (certainly in my world Rancher is categorised as such regardless of its implementation detail). On a more general level these clusters represent your application and infrastructure ‘estate’. So, for me, it’s not really whether we have one cluster or a hundred that’s important in terms of what we call it, but of course you may view that entirely differently.
Within your Ops Support community you expect a deeper understanding of how everything hangs together but in the general case of your average application developer or business owner, that level is, in my experience, both unnecessary and sometimes problematic. After all, we use tooling such as Rancher to purposely abstract implementation detail and one reason for that is to avoid the tensions of assumed ‘ownership’ of pieces of kit which we know don’t have much of a shelf-life in this world, and for very good reason. A while back I was talking with a security company about the transient nature or infrastructure and indeed containers, and they very much promoted the idea of the inverse of MTBF as being the preferred way of thinking about it, and that has stuck with me somewhat. It has also to encourage us toward practices such as managed disruption and chaos engineering as a way of increasing resilience and stability. Sorry I digress…
Bottom line, call things using names that are meaningful within the community of people that need to derive some understanding and utility from those names. You may choose a different word or description for a Product Owner, Business Sponsor, App Developer, Ops Support team or Release Manager, and that will depend on what it is that they need to get from some shared understanding. Whilst a common vocabulary is IMO a very valuable commodity (at least within a particular group) it is natural that some people will want to refer to the same thing in different levels of detail (i.e. a support person may refer to a cluster by its internal ID, and a business owner may know only that there is a capability that is enabled which they refer to as ‘Platform X’
Not sure if that helps or not ?
Regards
Fraser.
There is no official answer, but obviously a rancher manages a herd.
1 Like
Yes, this does help. I’m primarily looking for a descriptive name to refer to our implementation. I’m leading our transition from Rancher 1 to Rancher 2/Kubernetes, and from the bird’s eye view, people are tripping over the concept of ‘Cluster’ vs. ‘The collection of clusters, managed by Rancher’ and I’m looking for a good term for use in discussions, slides, and documentation.
“Platform” is a pretty good term.
‘Herd’ will now become an in-joke for those of us implementing the clusters.