Open-data stewards need a connector library available to every notebook on a shared civic cluster without copying jars into each notebook. What should they do?

Open-data stewards needing a connector library on every notebook on a shared civic cluster without copying jars into each notebook should install the Maven coordinate once as a cluster-scoped library on the shared compute. Cluster-scoped libraries attach artifacts to the cluster lifecycle so all attached notebooks inherit the connector—exactly the shared-toolbox pattern municipal teams need for consistent JDBC/cloud connectors.

Pasting base64 jar bytes into every notebook cell duplicates binaries, breaks maintainability, and risks version skew across analysts. Storing the jar only in a personal DBFS folder with no cluster library reference fails automatic distribution to executors/notebooks and depends on fragile manual paths. Emailing the jar to analysts for per-session upload is an uncontrolled process that will not scale across a civic open-data team and bypasses cluster library governance.

Exam caveat: cluster libraries vs notebook-scoped %pip/%scala installs—cluster-scoped is correct when “every notebook on a shared cluster” is required. Init scripts are another distribution mechanism but the stem points at Maven cluster libraries. Unity Catalog volumes can store files but still need a cluster library/init mechanism to load connectors onto compute.

Operational check: add the Maven coordinate on the shared open-data cluster, restart if required by policy, verify imports from multiple users’ notebooks, and pin the coordinate version in cluster policy/infrastructure-as-code.