Description:
There is NO need to maintain LOG_BLOCK_CHECKPOINT_NO in Redo Log Block.
We can get the end of the redo log in parse phase by LOG_BLOCK_HDR_NO.
Because Block_no of blocks in the same location of a redo file is not equal unless the file is larger than (512GB + 2k*2),and the maximum number of redo files is 512 GB(https://dev.mysql.com/doc/refman/8.0/en/innodb-parameters.html#sysvar_innodb_log_file_size)
Attached Block_NO calculation method:
inline uint32_t log_block_convert_lsn_to_no(lsn_t lsn) {
return ((uint32_t)(lsn / OS_FILE_LOG_BLOCK_SIZE) % LOG_BLOCK_MAX_NO + 1);
}
uint32_t LOG_BLOCK_MAX_NO = 0x3FFFFFFFUL + 1;
LOG_BLOCK_HDR_NO * 512/1024/1024/1024 = 512G
How to repeat:
just read the code
Suggested fix:
There is no need to maintain LOG_BLOCK_CHECKPOINT_NO
and no need to check LOG_BLOCK_CHECKPOINT_NO in recv_scan_log_recs
if (*scanned_checkpoint_no > 0 &&
log_block_get_checkpoint_no(log_block) < *scanned_checkpoint_no &&
*scanned_checkpoint_no - log_block_get_checkpoint_no(log_block) >
0x80000000UL) {
/* Garbage from a log buffer flush which was made
before the most recent database recovery */
ib::trace_2() << "Scanned cp no: " << *scanned_checkpoint_no
<< " block cp no "
<< log_block_get_checkpoint_no(log_block);
break;
}
