RAC Capture goes idle when database has been inactive and Capture consumes too much CPU (4282527)

The problem is two-fold:

When there is no database activity for a while, capture goes idle and it cannot be woken up. Generating DML and switching logs does not make the capture move ahead. Show capture detail shows the same info, as if nothing is happening on the database.

The second problem is that capture consumes too much CPU with one instance down an no activity on the other instance.

Capture idle is due to system call usleep not returning specific to Sun. The high CPU is caused by capture's data sequencer looking for new data too quickly, when it found no data from the RAC instances the previous time it looked.

WORKAROUND

Bounce capture, which will cause the message previously added to get replicated.

STATUS

Waiting for fix in future release of SharePlex 5.3

RAC with one node down