If InnoDB ‘s checksum algorithm reports that a page is corrupt, it tries to read it again; it retries 9 times.
If possible, it uses the following two POSIX fadvise() calls to optimize operating system cache usage, in the same fashion as xtrabackup:
posix_fadvise(filefd, 0, 0, POSIX_FADV_SEQUENTIAL); posix_fadvise(filefd, 0, 0, POSIX_FADV_DONTNEED);
- The tar4ibd binary produces archives that can be extracted with standard GNU tar, but you must use the -i option, or only part of your data will be extracted.
- It is used only for creating archives with backup files from XtraBackup, options -g (--listed-incremental) and -z (--gzip, --gunzip) are not supported and should not be used with tar4ibd.
- Compression at the moment of the backup should be achieved by streaming the backup (--compress) and piping through a compress utility (examples - with innobackupex - can be found here)
- Data from incremental backups is appended to the base backup at the preparation step, see the rationale of the process here.