Deploy & run

Step 5 — Deploy the jobs

This deploys five serverless jobs: stream_ticks (continuous), partition_manager, rollup_refresh, tiering, and retention.

dev — runs as you

Jobs are prefixed with your name and run as you; schedules are paused, so you start runs explicitly.

cd demo/live/bundle

databricks bundle deploy -t dev \
  --var="lakebase_project=$PROJECT" -p $PROFILE

prod — service principal

Jobs run as a service principal that owns a Lakebase Postgres role with the required privileges.

cd demo/live/bundle

databricks bundle deploy -t prod \
  --var="lakebase_project=$PROJECT" \
  --var="service_principal_name=<service-principal-application-id>" -p $PROFILE

---

---

Step 6 — Start the stream and watch it move

Start the continuous generator:

databricks bundle run lakets_demo_stream_ticks -t dev \
  --var="lakebase_project=$PROJECT" -p $PROFILE

Within a minute or two, ticks flow and the downstream triggers populate the cache, the invalidation log, and the CDF shadow. Trigger a RollUp refresh on demand to see the cascade do its work:

databricks bundle run lakets_demo_rollup_refresh -t dev \
  --var="lakebase_project=$PROJECT" -p $PROFILE

Inspect the hot tier directly while it runs:

-- Partitions, with status
SELECT * FROM lakets.show_chunks('stock_ticks') ORDER BY range_start;

-- Latest price per symbol, served from the cache
SELECT symbol, price, volume FROM public._lvc_stock_ticks ORDER BY symbol;

-- Minute candles
SELECT * FROM public._rollup_ohlcv_1min ORDER BY bucket DESC LIMIT 20;

-- RollUp watermarks advancing
SELECT name, watermark FROM lakets._rollup_registry ORDER BY name;

-- Dirty buckets waiting for the next refresh
SELECT count(*) FROM lakets._rollup_invalidation_log;

-- Rows mirrored to the shadow (and onward to Unity Catalog via CDF)
SELECT count(*) FROM lakets_cdf._shadow_stock_ticks;

Retention drops a hot partition only after drop_after and the durability gate confirms the lakehouse has it, so plan to run the demo long enough to cross that threshold, or lower the thresholds in setup.sql.

Mid-demo knobs

Re-deploy with different values, or edit the stream_ticks job widgets in the workspace and re-run:

Knob	Effect
symbols_count	Widens the cache and the RollUp cardinality.
rows_per_sec	Changes the ingest rate.
burst_mode	Periodically pushes a large extra batch, so the invalidation log visibly spikes.

---