Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.5.9, 10.2(EOL), 10.3(EOL), 10.4(EOL), 10.5, 10.6
-
CentOS 7
Description
Hi,
I do a daily "snapshot" restore of my daily backup performed by MariaBackup.
I've just noticed that I see *LOADS* of messages like:
Ignoring data file 'database1/table1.ibd' with space ID 25508, since the redo log references database1/table1.ibd with space ID 25506.
On tables that some jobs perform a "RENAME TABLE", to swap with a new version of them.
I've been seeing this since very long ago, but became more noticeable now that there is a job that does "RENAME TABLE" every 5 minutes.
Is this normal?
Does this mean anything bad to the database or backups? Or to the 'restored' database?
Is there any risk of corruption because of this, or loss of data/tables?
Thank you.
Attachments
Issue Links
- relates to
-
MDEV-25909 Unnecessary calls to fil_ibd_load() slow down recovery
-
- Open
-
Yes, the messages are normal and nothing to worry about. The messages were originally introduced by me in 2014 in MySQL 5.7 when I changed the redo log format and the recovery logic to rely on the redo log explicitly identifying the modified .ibd files. Before that, recovery used to read every .ibd file in the data directory in order to determine the numeric tablespace identifier.
Because I do not remember us observing any problems with that logic in the past 3 years, I think that it is safe to remove the message output.
Note: DDL operations are not expected to be fully crash-safe before
MDEV-17567has been implemented. In Mariabackup, there is some extra locking and logging, so that a correct backup should be created even if the server is concurrently executing DDL.If we ignore mismatch between .frm files and InnoDB (until
MDEV-17567fixes that), then RENAME TABLE inside InnoDB should be crash-safe ever sinceMDEV-14717.