ORA-01410 related to SGA change?

[jarviswabi]

I started getting ORA-01410 (Invalid ROWID) errors yesterday on my 10gR2 database, first time in my memory we've gotten them. They occurred on different tables at different times, and often the query that threw an error initially would work fine an hour or two later.

I initially thought it was a corrupt index on the first table that threw the error, so I rebuilt the index with no change to the error. Then I dropped and reloaded the table (this is partly a data warehouse) and reindexed it, and the query then worked. However, later errors were on tables that had no indices, so I don't think it was an index issue.

The one change that occurred yesterday was that I increased my Max SGA parameter based on a recommendation from the advisor. As I understand it, this can be done online without impact. However, the errors seemed to start shortly thereafter and no other changes had been made. Is it possible that the SGA change somehow invalidated information about the tables that produced those errors?

I restarted the database late last night to start with a fresh SGA and buffer cache so hopefully the errors would go away. I'm monitoring today to see if they return. I'm curious if anyone else has had these errors as a result of an SGA change, or can think of why that would occur?

Thanks for any input.

[PAVB]

My opinion is that an increase on max sga size is unlike of causing the issue you describe.

Probably the most common cause of ORA-01410 is an index being rebuilt at the same time a query is hitting the base table relying on the affected index. Query attempts to use the index as it was when query got loaded while the actual index is moved by the rebuild process -if the query hits one of the old blocks it usually fails with ORA-01410

Did you have other session rebuilding the offending index around the time of the failure?