Is there an existing issue for this?
Current Behavior
Running scripts/setup-lakebase.sh with a current Databricks CLI (v0.250+) exits 1 with no output at all. With bash -x you can see HOST= resolves empty and the script dies inside the _cfg helper before its own error guard can print.
Even if host/token resolution is patched locally, two further steps are broken against the current CLI/API:
- Existing databases are never detected (
postgres list-databases now prints a bare JSON array, the parser expects {"databases": […]}), so re-runs attempt duplicate creation — the script is no longer idempotent.
- The database-creation fallback posts
{"name": …}, but the databases API now requires {"spec": {"postgres_database": …, "role": "projects/…/branches/…/roles/…"}} and rejects the request.
Expected Behavior
The script provisions the Lakebase instance + branch + database as documented in its header, is idempotent on re-run, and fails loudly with a diagnostic when auth cannot be resolved.
Steps To Reproduce
- Install a current Databricks CLI (>= 0.250) and log in with OAuth:
databricks auth login --profile DEFAULT
- Run
./scripts/setup-lakebase.sh --name my-instance --database my_db --profile DEFAULT
- Observe exit code 1 with no output
Root causes:
databricks auth env now prints JSON ({"env": {…}}), not export KEY="…" lines, so _cfg's grep matches nothing; with set -o pipefail the failed grep aborts the script before the empty-value guard runs.
- OAuth profiles (
auth_type=databricks-cli) expose no DATABRICKS_TOKEN in auth env at all — a token must be minted via databricks auth token.
Cloud
AWS
Browser
Other
OntoBricks Version
0.6.1
Relevant log output
$ ./scripts/setup-lakebase.sh --name my-instance --database my_db --profile DEFAULT
$ echo $?
1
# (no output; bash -x shows HOST= resolving empty inside _cfg)
Additional Context
CLI-independent verification: the same provisioning steps executed directly against the REST API succeed, confirming the API contract moved rather than the workspace being misconfigured. A fix is prepared.
Is there an existing issue for this?
Current Behavior
Running
scripts/setup-lakebase.shwith a current Databricks CLI (v0.250+) exits 1 with no output at all. Withbash -xyou can seeHOST=resolves empty and the script dies inside the_cfghelper before its own error guard can print.Even if host/token resolution is patched locally, two further steps are broken against the current CLI/API:
postgres list-databasesnow prints a bare JSON array, the parser expects{"databases": […]}), so re-runs attempt duplicate creation — the script is no longer idempotent.{"name": …}, but the databases API now requires{"spec": {"postgres_database": …, "role": "projects/…/branches/…/roles/…"}}and rejects the request.Expected Behavior
The script provisions the Lakebase instance + branch + database as documented in its header, is idempotent on re-run, and fails loudly with a diagnostic when auth cannot be resolved.
Steps To Reproduce
databricks auth login --profile DEFAULT./scripts/setup-lakebase.sh --name my-instance --database my_db --profile DEFAULTRoot causes:
databricks auth envnow prints JSON ({"env": {…}}), notexport KEY="…"lines, so_cfg's grep matches nothing; withset -o pipefailthe failed grep aborts the script before the empty-value guard runs.auth_type=databricks-cli) expose noDATABRICKS_TOKENinauth envat all — a token must be minted viadatabricks auth token.Cloud
AWS
Browser
Other
OntoBricks Version
0.6.1
Relevant log output
Additional Context
CLI-independent verification: the same provisioning steps executed directly against the REST API succeed, confirming the API contract moved rather than the workspace being misconfigured. A fix is prepared.