Skip to main content

CDC-2001 Redo log gap

Code: CDC-2001 (LOG_GAP). Not retried: the task stops at once.

What the connector observed​

Before every mining step the connector lists the redo logs covering the range from its cursor to the database's current SCN, for every redo thread that is enabled. It prefers the archived copies at one archive destination (the one named in cdc.archive.destination, or the lowest valid local destination), and in the default online capture mode adds the online logs for the part of the range not yet archived. It then checks that the list is continuous, and it was not. The message names the thread, the SCN range and what is wrong:

  • no logs at all for the thread;
  • the earliest log starts after the start SCN;
  • a sequence is missing between two logs;
  • the latest log ends before the end SCN.

The connector reads V$ARCHIVED_LOG for its one destination only. A log archived only to another destination is not seen.

The steps below use Kafka Connect's REST API. oracle-cdc-admin wraps them: offsets set refuses an SCN whose redo is purged, and resnapshot reads chosen tables again.

Why it stopped rather than continued​

A missing sequence holds changes the connector has not read. Mining the logs on either side of it would publish everything except those changes, and nobody would know. The connector never chooses a later start point by itself.

Confirm the cause​

SELECT dest_id, dest_name, status, type, destination FROM v$archive_dest_status WHERE status <> 'INACTIVE';
SELECT thread#, sequence#, first_change#, next_change#, dest_id, name, deleted, status
FROM v$archived_log
WHERE thread# = :thread AND next_change# > :start_scn AND first_change# <= :end_scn
ORDER BY sequence#, dest_id;
SELECT thread#, status, enabled, sequence# FROM v$thread;

Look for the missing sequence: it may exist at another destination, be marked deleted, or be absent everywhere. oracle-cdc-doctor check (rule DOC-12) confirms which destination the connector reads.

Recover​

  1. If another destination holds the full sequence, set cdc.archive.destination to that destination's name (for example LOG_ARCHIVE_DEST_2) and restart the task.
  2. If a backup holds the missing log, restore it with RMAN so that V$ARCHIVED_LOG lists it at the destination the connector reads, confirm with the query above, and restart the task. Nothing is lost.
  3. If the log is gone for good, the changes in it cannot be captured. The only way on is a deliberate skip: move the offset to the first SCN after the gap (moving the offset by hand), then reload the captured tables with a snapshot signal. A snapshot republishes the rows that exist; rows deleted during the gap get no delete record, so downstream copies may need to be reconciled.

Restart the task with:

curl -s -X POST "$CONNECT/connectors/$NAME/restart?includeTasks=true&onlyFailed=true"

To prevent a recurrence, keep archived logs at the destination the connector reads for at least as long as the connector could be stopped plus its lag, and delete them with RMAN rather than with operating system commands.