You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently if an error occurs during restore job, the job is stopped and the error is printed to stderr. It's not convenient in case when a large database is being restored, because an error makes it necessary to start the job from the very beginning.
A following flow is suggested: If an error occurs, it's printed to the stderr/log (like described here #10541), the problem row is skipped, and the job continues restoring the following rows.
Additionally, after an object is restored, the number of restored rows vs the number of total rows in the dump should be printed.
This should not be default behavior, because there is a risk of consistency loss if some rows are skipped.
Output example:
Restore started. Total number of objects to restore: 3
Restored cluster/Db1/table1: 100 rows out of 100
Restored cluster/Db1/table2: 200 rows out of 200
Error restoring cluster/Db1/table3, row 234, backup_20241021T122353/table3/data_01.csv
Error: Cannot write data #1
Error restoring cluster/Db1/table3, row 235, backup_20241021T122353/table3/data_01.csv
Error: Cannot write data #1
Restored cluster/Db1/table2: 298 rows out of 300
Job completed with 2 errors.
The text was updated successfully, but these errors were encountered:
Currently if an error occurs during restore job, the job is stopped and the error is printed to stderr. It's not convenient in case when a large database is being restored, because an error makes it necessary to start the job from the very beginning.
A following flow is suggested: If an error occurs, it's printed to the stderr/log (like described here #10541), the problem row is skipped, and the job continues restoring the following rows.
Additionally, after an object is restored, the number of restored rows vs the number of total rows in the dump should be printed.
This should not be default behavior, because there is a risk of consistency loss if some rows are skipped.
Output example:
The text was updated successfully, but these errors were encountered: