10 free sample questions from a bank of 56, with the correct answers and explanations. No signup required — start practising right now.
B is correct. B -- Oracle GoldenGate Application Adapters. JMS queues are a message-oriented destination, and the Application Adapters product is the part of the portfolio that moves data into and out of application queues.
B -- Oracle GoldenGate Application Adapters. JMS queues are a message-oriented destination, and the Application Adapters product is the part of the portfolio that moves data into and out of application queues.
The GoldenGate portfolio is split by destination. GoldenGate for Big Data covers Kafka, Hadoop, NoSQL and flat-file style targets. Veridata is a data comparison and repair tool, not a replication path. Oracle Data Integrator is a separate ETL and data movement product with its own adapters. Application Adapters sit alongside GoldenGate to bridge application constructs such as JMS queues, user-defined files and messaging systems.
B is correct because JMS is an application messaging construct rather than a database log, so replicating into a JMS queue needs the adapter layer. Application Adapters provide that: they are used with GoldenGate to deliver replicated data into JMS destinations, and the Oracle by-example that ships with the product is precisely a configuration that replicates from a Windows source into an ActiveMQ JMS destination. The other three products have no JMS capability at all.
Match the GoldenGate product to the destination type: databases and log-based capture for the classic engine, JMS and application constructs for Application Adapters, big-data stores for Big Data, and comparison for Veridata.
C is correct. C -- target objects can differ from source objects when you use SOURCEDEFS in the Replicat.
C -- target objects can differ from source objects when you use SOURCEDEFS in the Replicat.
GoldenGate replicates DDL by parsing the statement and re-issuing it against the target. By default the object names in the DDL are used unchanged, so source and target must match. The Replicat's SOURCEDEFS parameter points at a file of source definitions, which is the documented mechanism for reconciling differing object names. On the Extract side, WILDCARDRESOLVE and COLMAP are DML features and have no meaning for DDL, while user exits are available to DDL through the UEXIT facility.
C is correct because SOURCEDEFS exists precisely for the case where the target schema or object names differ from the source, and it is the only option in the list that describes a real, documented DDL-replication capability with that property. GoldenGate reads the source DDL and, using the sourcedefs, emits the equivalent statement against the correctly named target objects.
SOURCEDEFS is configured, so saying source and target objects simply can be different omits the one mechanism that actually makes it work and would be true of almost any statement.WILDCARDRESOLVE is an Extract parameter for resolving wildcard table specifications in DML, typically resolving them to a single object. It is not part of DDL replication at all.For DDL, the mapping question is object naming. SOURCEDEFS on the Replicat is the answer whenever the target names differ; anything from the DML wildcard-resolution family is a distractor.
A is correct. A -- the wrong build of GoldenGate was installed for the CPU architecture, so GGSCI cannot execute on that host.
A -- the wrong build of GoldenGate was installed for the CPU architecture, so GGSCI cannot execute on that host.
GGSCI is a compiled binary, so it has to match the machine it runs on. GoldenGate is published for both 32-bit and 64-bit platforms, and a build of the wrong width fails to load on the host's kernel rather than degrading gracefully. The same applies to the Extract, Replicat, Manager and logdump executables, because they are all built for the same platform as the product. On Linux the shared library path must also be right, but a wrong-architecture build fails earlier, before any library search can help.
A is correct because the error occurs at launch, before any GoldenGate configuration is consulted, which points at the binary itself rather than at its environment or its licensing. A GoldenGate install built for the other CPU architecture cannot be executed on the host, and the reported failure follows immediately from ggsci not starting at all.
ORACLE_HOME, and in any case a bad library path produces a loader error naming the missing library rather than a bare failure to start.A GGSCI that will not start at all is a binary problem before it is a configuration problem. Check the build matches the host architecture, then check the library path.
C — RMTTRAIL is for remote trails and EXTTRAIL is for local trails. It is the only true statement of the four.
Oracle documentation defines the local trail versus remote trail exactly this way: a trail on the local system is an Extract (local) trail written via EXTTRAIL, while a trail on the target system is a remote trail written via RMTTRAIL, added on the source and linked to the data-pump Extract.
EXTTRAIL equals local, RMTTRAIL equals remote — and any option claiming Extract cannot transform data is always false.
BE is correct. B and E -- run the chkpt_db_create.sql script, or issue ADD CHECKPOINTTABLE in GGSCI. Both create the checkpoint table.
B and E -- run the chkpt_db_create.sql script, or issue ADD CHECKPOINTTABLE in GGSCI. Both create the checkpoint table.
A Replicat records its read position so it can resume where it stopped. Oracle ships chkpt_db_create.sql to create the table directly, and ADD CHECKPOINTTABLE does the same thing through GGSCI so the table name is registered for future use. A checkpoint table is a global temporary table owned by the GoldenGate schema and is unrelated to the trail files themselves.
B is correct because chkpt_db_create.sql is the documented script for creating a checkpoint table directly in the target database. E is correct because ADD CHECKPOINTTABLE performs the same creation from GGSCI and additionally remembers the name so later ADD REPLICAT statements can inherit it. Both are genuine ways to satisfy the requirement, which is why the question asks for two.
chkpt_db_extracttrail.sql is a control script for managing Extract trail files on disk, not a script that creates a checkpoint table in a database.ADD EXTTRAIL is the GGSCI command that defines a trail file on disk. Checkpoints are stored in a database table, so a trail-file command cannot create one.INFO REPLICAT ... , SHOWCH is a reporting command. It reads and displays checkpoint positions; it never creates the underlying table.Two routes to a checkpoint table: the chkpt_db_create.sql script, or ADD CHECKPOINTTABLE from GGSCI. Trail-file commands and INFO commands are about reading and writing trail files, never about creating tables.
C is correct. C -- configure the Extract with FORMATASCII SQLLOADER. Only Extract writes the initial-load file, so only Extract chooses its format.
C -- configure the Extract with FORMATASCII SQLLOADER. Only Extract writes the initial-load file, so only Extract chooses its format.
During initial load the Extract group runs in a special mode and writes source records to an output file in a chosen format rather than to a trail. FORMATASCII SQLLOADER produces the delimited text layout that SQL*Loader expects, which lets a fast bulk loader consume the data. FORMATXML produces tag-delimited output for a user-defined format, and the trailing OPTION keyword selects which of those formats is written.
C is correct because the file is produced by Extract, and the format keyword is an Extract parameter. FORMATASCII SQLLOADER makes Extract emit exactly the control-character-delimited layout that SQL*Loader's conventional path parses, so the two tools meet without a conversion step. The Replicat has nothing to do with writing the file at all.
FORMATASCII SQLLOADER parameter.FORMATSQL is not a GoldenGate format keyword. The valid forms are FORMATASCII, FORMATXML and FORMATBINARY, so this names a parameter that does not exist.FORMATXML on the Replicat. Even if Replicat could format output, XML tag-delimited format is not what a SQL*Loader initial load consumes.File format is chosen by the writer, and during initial load the writer is the Extract. FORMATASCII SQLLOADER for SQL*Loader, FORMATXML for a user-defined parser -- both on Extract.
C is correct. C -- the BR parameter. Bounded recovery is what lets Extract restart cleanly with long-running open transactions still in flight.
C -- the BR parameter. Bounded recovery is what lets Extract restart cleanly with long-running open transactions still in flight.
Extract caches an entire transaction in memory from its first statement until it reaches the commit record, so that the trail always contains whole transactions. If the process stops while transactions are open, that cached data is written to disk and recovered on restart -- which is bounded recovery. The BR parameter controls it, with sub-parameters for start and stop age and for interval. BLOBMEMORY is a separate setting that caps how much memory large objects may consume before spilling.
C is correct because bounded recovery is exactly the facility described: Extract stopped with open transactions, and the mechanism that guarantees a clean and efficient restart is BR. Tuning the bounded recovery behaviour with BR controls how old an open transaction may be before it is treated as bounded, and how often recovery data is examined, so recovery does not have to rescan the whole log.
FORCETRANS and PURGEORPHANEDTRANSACTIONS do not control bounded recovery. FORCETRANS forces a transaction to be written to the trail even if it contains no captured data, and the other option relates to cleaning up records that were captured but never applied.WARNLONGTRANS only controls the warning threshold for long-running transactions. It reports; it does not provide or tune any recovery mechanism.PURGEORPHANEDTRANSACTIONS concerns discarding trail records that were extracted but never replicated, which is about trail hygiene rather than about restarting a stopped Extract.If the question is about restarting after a stop with transactions still open, the keyword is BR. BLOBMEMORY caps LOB memory, and WARNLONGTRANS only warns.
The answer is A, ENABLE_GOLDENGATE_REPLICATION=true. From 11.2.0.4 a database must be told that GoldenGate will replicate from it, and the flag lives in the compatible parameter, for example compatible='11.2.0.4'. It is a deliberate opt-in: without it Extract cannot register and replication fails before any trail is written.
A is correct. A -- REPERROR in the Replicat parameter file. It is the parameter that decides what happens when an error is hit while applying a transaction.
A -- REPERROR in the Replicat parameter file. It is the parameter that decides what happens when an error is hit while applying a transaction.
Error handling in one-way replication is configured on the Replicat. REPERROR takes a number, IGNORE or REPERRORDELAY: a number skips the failed transaction that many times and then writes it to the discard file, IGNORE skips it indefinitely, and REPERRORDELAY defers it so later transactions can pass. HANDLECOLLISIONS is the older mechanism for the bidirectional case, and MAP/MAPERR write rejected records to a file.
A is correct because REPERROR is the documented control for error handling in a one-way configuration. Setting it to a count, IGNORE or REPERRORDELAY tells the Replicat whether to abandon the transaction, keep retrying it, or set it aside, and any transaction that is finally abandoned lands in the discard file for later inspection.
HANDLECOLLISIONS belongs to bidirectional replication, where a rollback caused by a conflicting row has to be retried automatically. In one-way replication there is no concurrent writer to collide with.MAPEXCEPTION is not a GoldenGate Replicat parameter. Error routing is done with MAPERR to a file and REPERROR to control the behaviour.One-way error handling is REPERROR on the Replicat. HANDLECOLLISIONS and CDR are bidirectional features and do not belong in a one-way configuration.
C is correct. C -- tables that have conflicts. Conflict resolution only earns its keep in a bidirectional setup, and the profiling output tells you which tables are contended.
C -- tables that have conflicts. Conflict resolution only earns its keep in a bidirectional setup, and the profiling output tells you which tables are contended.
In a master-to-master topology both databases accept writes for the same tables, so the same row can be updated on both sides. GoldenGate detects this by comparing before and after images and resolves it with a configured rule. The profiling scripts that run before a bidirectional implementation are the tool that finds out which tables are actually shared, because a table with no overlap never needs a resolution rule.
C is correct because a conflict is by definition a disagreement about the same row between two writers, which requires two masters. If the profiling output shows no conflicting tables, no conflict resolution configuration is needed and the implementation is simpler; if it shows conflicting tables, those are exactly the tables that need a rule.
NOLOGGING tables are a performance and recoverability concern because their changes are not written to the redo log in the normal way. It says nothing about two writers contending for a row.In a bidirectional build, the profiling step exists to answer one question: which tables have two writers? Only those tables need a conflict resolution rule.
Want the full bank of 56 questions for Oracle GoldenGate 12c Implementation Essentials (1Z0-447)? See all practice exams.