Under high write load, backup fails with "log block numbers mismatch" error
General
Escalation
General
Escalation
Description
Since the migration to PS 8.0.34 and the new redo log implementation, many servers with a high write load are now failing with the error "log block numbers mismatch". The innodb_redo_log_capacity is set to 512GB and about 5MB/s is written to the redo. With these numbers, we can rule out a rewind of the redo log.
Here's the output of the command:
You will notice a few extra lines, I added a few extra outputs to get insights to what is going on. If that can help, the block number mismatch is always 128.
The modifications are in the file redo_log.cc at around line 300:
The backup looked only at 8 of the 32 redo log files and then failed to find the start_lsn value.
Since the migration to PS 8.0.34 and the new redo log implementation, many servers with a high write load are now failing with the error "log block numbers mismatch". The innodb_redo_log_capacity is set to 512GB and about 5MB/s is written to the redo. With these numbers, we can rule out a rewind of the redo log.
Here's the output of the command:
You will notice a few extra lines, I added a few extra outputs to get insights to what is going on. If that can help, the block number mismatch is always 128.
The modifications are in the file redo_log.cc at around line 300:
The backup looked only at 8 of the 32 redo log files and then failed to find the start_lsn value.