Document operator system requirements

[maltesander]

We want to document user-guidelines for the operator in terms of e.g. min/max CPU, memory and disk space. These are non-binding (as an accurate assessment is dependent on a specific use-case) but may be referred to from GTCs and other contract documents.

This should be documented as described in stackabletech/documentation#289

Acceptance

• run the following
  • zookeeper operator managing single ZK instance (serving HDFS, HBase and Kafka clusters)
  • zookeeper operator managing multiple ZK instances in different namespaces (each serving HDFS, HBase and Kafka clusters)
  • vary cluster sizes
• recommendations for CPU, memory and disk space per zookeeper instance managed by the operator, for
  • Kind cluster
  • Managed k8s cluster (at least one)
• The results are added to a table in the documentation as described in Document operator system requirements documentation#289

[lfrancke]

Thank you, this looks good.
Is there any reason to believe we need to test every operator or can we assume that we can copy the requirements from one to the other?

Especially as this will potentially change with every new PR?

I would suggest we do one, double or quadruple the memory and disk space required and set that as the minimum.

What do you think?

[adwk67]

Thank you, this looks good. Is there any reason to believe we need to test every operator or can we assume that we can copy the requirements from one to the other?

Especially as this will potentially change with every new PR?

I would suggest we do one, double or quadruple the memory and disk space required and set that as the minimum.

What do you think?

I think that makes sense: we could look at this one (zookeeper) in some detail (as it will typically serve multiple product clusters) and then use that as rule-of-thumb benchmark.