Oracle Database / Storage, Undo & Redo
ORA-01555: correlate consistent-read failures with undo pressure
A long query can lose needed older block versions while other transactions generate undo. Capture query duration and the failing workload interval.
Environment and inputs: Oracle Database 19c baseline · SQL*Plus-compatible SQL unless RMAN, DGMGRL, ADRCI or a host shell is specified. Check release, patch level, edition and container first.
Start with this diagnostic
Use the stated SQL, RMAN, DGMGRL or host shell and replace sample names with approved targets. DBA and V$ views need authorized catalog access in the correct container. Do not grant broad privileges for an example.
SELECT begin_time, end_time, maxquerylen, tuned_undoretention, ssolderrcnt, nospaceerrcnt FROM v$undostat ORDER BY begin_time DESC FETCH FIRST 12 ROWS ONLY;Read the result
The interval statistics show query length, tuned retention and undo-related errors. They support diagnosis but do not identify every application cursor pattern causing the error.
Resolve and verify
Reduce unnecessary query duration, size undo for real write concurrency and review commit or fetch patterns. Reproduce the report under representative change volume.
Operational precautions
Increasing UNDO_RETENTION alone does not guarantee retention without sufficient capacity and appropriate configuration.