The following guides provide information for troubleshooting and addressing problems with Oracle deployments.
Extractor Slow-down on Single Service
If when replicating from Oracle, a significant increase in the latency for the extractor within a single service, it may be due to the size of changes and the data not being automatically purged correctly by Oracle.
The CDC capture tables grow over time, and are automatically purged by
Oracle by performing a split on the table partition and releasing the
change data from the previous day. In some situations, the purge
process is unable to acquire the lock required to partition the table.
By default, the purge job does not wait to acquire the lock. To change
this behavior, the
DDL_LOCK_TIMEOUT parameter can
be set so that the partition operation waits for the lock to be
available. For more information on setting this value, see
Last Updated: 2016-04-20
Condition or Error
It is possible for the Oracle server to get into a state where Tungsten Cluster is online, and with no other errors showing in the log. However, when logging into the Oracle server an error is returned:
ORA-00257: ARCHIVER ERROR. CONNECT INTERNAL ONLY, UNTIL FREED
This is a lack of resources within the Oracle server, and not an issue with Tungsten Cluster.
The issue can be addressed by increasing the logical size of the recovery area, by connecting to the Oracle database as the system user and running the fol\ lowing command:
sqlplus sys/oracle as sysdbaSQL>
ALTER SYSTEM SET db_recovery_file_dest_size = 80G;