In Integrated Capture, the GoldenGate Extract process (client) communicates directly with the Oracle database's LogMiner (server) via XStreams.
GGSCI> VIEW GGSEVT
Recent database patching (e.g., 19.14 Merge Patch ) can sometimes cause issues where integrated extracts hang and subsequently detach, as noted in recent MOS (My Oracle Support) documents. Troubleshooting and Resolution Steps
Review the database alert log around the exact timestamp of the detachment. Look for corresponding errors such as: Look for corresponding errors such as: Seeing this
Seeing this message alone is not a problem. However, if your replication has stopped unexpectedly, follow this checklist.
| Error | Meaning | DBA Action | | --------------------------------- | ------------------------------------------------- | ---------------------------------------------------------------------- | | OGG-02912 (Insufficient SGA) | Database’s streams_pool_size too small. | Increase streams_pool_size or sga_target . Restart extract. | | OGG-02902 (Timeout) | LogMiner server didn’t respond. | Check database alert log for bottlenecks or deadlocks. | | OGG-06439 (No capture resources) | Too many concurrent LogMining servers. | Stop other extracts or increase MAX_SERVERS in the database config. |
For Integrated Capture, relying entirely on Automatic Memory Management (AMM) can sometimes result in laggy allocations. It is best practice to set a minimum floor for the Streams Pool. ALTER SYSTEM SET STREAMS_POOL_SIZE = 2G SCOPE=BOTH; Use code with caution. | Increase streams_pool_size or sga_target
Check the Oracle Database alert log for corresponding Logminer or Streams pool events. Look for parallel execution messages, memory resizing operations, or ORA-errors occurring at the exact same timestamp. Resolution and Prevention Strategies
If the Extract process encounters a data error (e.g., fetching from an un-logged table, encountering an unsupported data type, or hitting a bug) it will "abend" (abnormally end). This sudden death forces the database to log the detachment message. Step-by-Step Troubleshooting Workflow
This implies the detachment was involuntary. Something—like a database crash, a memory exhaustion issue (SGA/Streams Pool), or a permissions change—forced the client off. Troubleshooting Forced Detachments ATTACH CAPTURE <
If the state is "WAITING FOR TRANSACTION" or "ERROR," that is likely the root cause. Step 3: Handle Hung Sessions/Registering Again
oggadmin> ATTACH CAPTURE <capture_name> CLIENT <client_name>
If replication is stalled but the process isn't abended, check for lag: