Event_log shows the following message and then core dumps:
Reader: 11101 - failed to allocate memory for NodeC's col_seqno_array (from REP) [module ord]
Reader runs out of OS dynamic memory while building the tcache. This happens when there is a huge capture backlog that reader must process.
There are two solutions to prevent this issue from happening in the future:
1. Turning on supplemental logging on unique keys and primary keys so that read doesn't need to query the key information from Oracle.
2. Increase the OS dynamic memory datasize value:
ulimit -aS (datasize The maximum size of a process's heap in kilobytes.)