search

stackabletech / druid-operator

An Operator for Apache Druid for Stackable Data Platform
Other
10 stars 0 forks source link

fix(docs): Update getting-started script pre-24.11 #643

Closed adwk67 closed 1 week ago

adwk67 commented 1 week ago

Check and Update Getting Started Script

Part of https://github.com/stackabletech/issues/issues/657. Uses --server-side for deploying druid resource so that null fields are honoured.

[!NOTE] During a Stackable release we need to check (and optionally update) the getting-started scripts to ensure they still work after product and operator updates.

# Some of the scripts are in a code/ subdirectory
# pushd docs/modules/superset/examples/getting_started
# pushd docs/modules/superset/examples/getting_started/code
pushd $(fd -td getting_started | grep examples); cd code 2>/dev/null || true

# Make a fresh cluster (~12 seconds)
kind delete cluster && kind create cluster
./getting_started.sh stackablectl

# Make a fresh cluster (~12 seconds)
kind delete cluster && kind create cluster
./getting_started.sh helm

popd
adwk67 commented 1 week ago

I don't fully understand how this fixes it, but given it is for demo purposes I think if it works it is ok.

Does this change much for production users?

It seems that some client tools in some versions (in this case, kubectl) try to resolve the nullable-yet-having-a-default character of CRD fields before sending it to the server: with --server-side, the CRD definition is honoured as-is, but has the downside of being a bit "brute-force" (i.e. possibly causing problems if resources are patched concurrently). So this may affect how customers deploy resources.

adwk67 commented 1 week ago

Opened an issue here: https://github.com/stackabletech/issues/issues/663