Broadcom Tanzu Greenplum Technical Specialist (250-608) Certification Sample Questions

01. A land-registry DBA runs CREATE INDEX parcels_gix ON parcels USING GIST (geom); on a table of parcel polygons.
How does the PostGIS spatial index represent each parcel?
a) By the full vertex list of each polygon, so matching rows need no further geometry test
b) By a hash of each polygon's well-known text, grouped by the segment that stores the row
c) By its bounding box, a lossy proxy that narrows the candidates before the exact geometry test
d) By the H3 cell that holds each polygon's centroid, at one fixed resolution

02. A solutions architect is confirming what the Greenplum PostGIS extension offers before a geospatial project begins.
Which of the following does Greenplum PostGIS support?
Select all that apply.
a) Topology data through postgis_topology
b) Spatial indexes created USING GIST
c) 3D geoprocessing through postgis_sfcgal
d) Raster data in columns of the raster type

03. A business team has used gpMLBot to train a model on their development Greenplum cluster. They now want the model to run inside a production data pipeline that lives outside that cluster.
What does the gpmlbot deploy phase do?
a) It exports the trained model as CSV and SQL files for loading into another database
b) It retrains every candidate algorithm on the full table and keeps the most precise
c) It publishes the saved model as a REST endpoint served from the coordinator host
d) It scores a new table with a saved model and reports the results

04. An ETL developer is writing a first gpload job. The INPUT section of the control file names source files on the ETL host where the job will run.
How is the gpfdist file server provided for this load?
a) An administrator must start gpfdist on the coordinator before every run
b) The segments start gpfdist on their own hosts when the external table is read
c) gpload starts one or more gpfdist instances on the current host itself
d) No file server is involved; gpload streams the files through COPY on the coordinator

05. A retail chain stores sales-district polygons and store locations as geometry columns in the same spatial reference system. The point for store 118 lies exactly on the boundary line that District North shares with its neighbor; it is not in the interior of either district.
Which predicate returns FALSE for store 118 (s.geom) and the District North polygon (d.geom)?
a) ST_Intersects(d.geom, s.geom)
b) ST_Covers(d.geom, s.geom)
c) ST_Touches(d.geom, s.geom)
d) ST_Within(s.geom, d.geom)

06. A database is restored with gprestore onto a test cluster. A dashboard that reads the materialized view mv_daily_revenue then fails because the view holds no data, although its definition is present.
What should the administrator do?
a) Run REFRESH MATERIALIZED VIEW on it, because a backup holds only a materialized view's definition
b) Run ANALYZE on the view, since restored objects have no statistics until they are analyzed
c) Restore again with --data-only, since the first restore left out view data by default
d) Back up the view again with --include-table, and then restore that backup

07. A support application looks up one customer at a time with WHERE lower(email) = lower($1). The customers table already has a B-tree index on email, but EXPLAIN shows that the lookup does not use it.
Which change lets the lookup use an index?
a) Create a bitmap index on email so that values differing only in case share one bitmap
b) Run ANALYZE on customers so the optimizer recognizes the existing index
c) Create an index on the expression lower(email)
d) List-partition customers on email so each lookup scans one partition

08. An administrator needs to check a production Greenplum system for catalog inconsistencies. No maintenance window is available, so the system stays in normal mode with user workload running. The administrator wants results that reflect real catalog problems rather than work in flight.
How should gpcheckcat be run?
a) With no options, since the utility pauses user sessions during each test
b) With -A, so every database is checked while users remain connected
c) With -g, so generated repair scripts absorb differences from running work
d) With -O, which limits the run to the tests that are safe to perform while the system stays online

09. A DBA runs EXPLAIN twice on the same query: once with GPORCA, giving a top-node cost of about 296, and once with optimizer set to off, giving about 3,519 from the Postgres-based planner. The DBA concludes that the GPORCA plan is roughly twelve times cheaper to run.
How should this conclusion be judged?
a) It is sound, because both optimizers express cost in the same unit of disk page fetches
b) It is unsupported, because costs from different optimizers cannot be compared
c) It is sound only if ANALYZE ran just before both EXPLAIN commands
d) It is unsupported, because both cost figures include the time spent sending result rows to the client

10. Every night a CRM system writes a file containing new customers and changed details for existing customers. The Greenplum customer table must gain the new rows and have the changed attributes of existing rows overwritten, matched on customer_id. Customers missing from the file must stay as they are.
Which gpload control-file setting meets this requirement?
a) MODE: MERGE with MATCH_COLUMNS set to customer_id and the attributes listed in UPDATE_COLUMNS
b) MODE: INSERT with PRELOAD and TRUNCATE: true, so each night's file fully replaces the table contents
c) MODE: UPDATE with MATCH_COLUMNS set to customer_id and the changed attributes listed in UPDATE_COLUMNS
d) MODE: INSERT with ERROR_LIMIT and LOG_ERRORS, so rows for existing customers are rejected and logged