Changes in MySQL Enterprise Backup 4.1.3 (2019-02-15)

• Functionality Added or Changed

• Bugs Fixed

Functionality Added or Changed

• In addition to the requirement that the target data directory for a restore specified by the --datadir option must be non-existent or empty, mysqlbackup now enforces the same rule for the --innodb_data_home_dir, --innodb_log_group_home_dir, and --innodb_undo_directory options (the --force option cannot be used to override the requirement on the three options). (WL #12665, WL #12407)

Bugs Fixed

• MySQL Enterprise Backup 4.1 failed to backup a MySQL Server that has been backed up by MySQL Enterprise Backup 4.0 before. It was due to the way the backup_history table was queried, which has now been fixed. (Bug #29208386)

• When working with a Group Replication cluster, mysqlbackup might quit unexpectedly near the end of a backup operation when, in order to write to the backup_history table, it tried to connect with an unencrypted connection to one of the nodes on which the backup user had not logged on before. It happened when the backup user was created with the caching_sha2_password plugin plugin, so that it must log on with an encrypted connection when it connected to the server for the first time; the attempt to log on thus failed, and mysqlbackup could not handle the failure. With this fix, at such failures, mysqlbackup quits gracefully with the warning that the backup operation is finished without updates to the backup history. (Bug #28893180)

• Restoring an incremental backup image using the copy-back-and-apply-log command failed with mysqlbackup complaining that the server repository configuration (including, for example, value of innodb_data_file_path) was unknown for the target server, unless the configuration is explicitly supplied to mysqlbackup. With this fix, mysqlbackup gets the required information from the backup-my.cnf file already restored with the base backup of the incremental backup. (Bug #28411028)

• A restore operation could corrupt a backup when, by mistake, a user specified the source directory to become the target directory for restoring some files (for example, specifying what was the backup's --backup_innodb_data_home_dir value as the restore's --innodb_data_home_dir value). This fix prevents the problem by having mysqlbackup throw an error when the command options make the source and target file paths the same for any file copying during a restore. (Bug #28376873)

• On FreeBSD platforms, using the --show-progress option did not make mysqlbackup print progress reports. (Bug #28350122)

• After encrypted InnoDB tables have been restored, sometimes the restored server could not be started, or the encrypted InnoDB tables could not be opened after the server had been started.

  This fix not only resolves the aforementioned issues, but also two other problems: the failure to restore a backup containing encrypted InnoDB tables that were row-compressed, and the failure to complete a backup when an encrypted InnoDB table was created in the middle of the backup operation. (Bug #28301281)

  References: See also: Bug #28360241, Bug #27168458.

• While MySQL Server interprets the system variable setting --innodb_checksum_algorithm=0 to mean --innodb_checksum_algorithm=crc32, a mysqlbackup operation (except for backup) failed when --innodb_checksum_algorithm=0 was set as a configuration option on the backed up server. With this fix, mysqlbackup now takes --innodb_checksum_algorithm=0 as valid and interprets it as --innodb_checksum_algorithm=crc32. (Bug #28295519)

• The Windows version of MySQL Enterprise Backup did not display its build ID when invoked. (Bug #27916702)

• mysqlbackup quit unexpectedly when backing up a MySQL Server of release 4.1.2 or later for the first time if the ALTER privilege on the mysql.backup_history_new table had not been granted to the MySQL user with which mysqlbackup connected to the server. With this fix, mysqlbackup quits gracefully in the situation after throwing the proper error.

  Also, the CREATE, INSERT, and DROP privileges on mysql.backup_history_old and CREATE, INSERT, DROP, and ALTER privileges on mysql.backup_history_new are now required only for backing up for the first time a MySQL Server that has been upgraded from 5.7.22 or earlier and has been backed up by MySQL Enterprise Backup before. (Bug #27879530, Bug #28546256)

• When the --show-progress=table option was used, mysqlbackup gave a warning in the error log on an aborted connection to the server near the end of the operation. It was because the connection to the server for writing to the backup_progress table had remained open. With this fix, the connection is properly closed after the mysqlbackup operation is finished. (Bug #27647283)

• A restore operation for an incremental backup failed when individual tablespaces with relative file paths were involved. (Bug #26442994)

• When the option --only-innodb-with-frm or --no-locking was used during a backup operation, the backup sometimes failed with mysqlbackup complaining that the highest LSN was larger in a copied page than on the backed-up server. It was because mysqlbackup did not perform a log flushing before copying the redo log when either of the mentioned options was used. With this fix, log flushing was always performed to prevent the error. (Bug #25412655)

• Partial backups sometimes failed because full-text index files had their file names matched by the regular expression provided by the --include-tables option, and the files were then handled as ordinary tablespace files by mysqlbackup. With this fix, mysqlbackup excludes any full-text index files from backups. (Bug #25044900)

• If, when a backup was in progress and mysqlbackup was reading the binary log (or the relay log) index file and the server tried to modify the index file (because, for example, a log flush or log purge just took place), the binary logging (or relay logging) stopped; the server also quit unexpected on Windows platforms. It was because mysqlbackup did not handle well the file sharing violation. With this fix, mysqlbackup now reads the index file using the local file system API, which handles the file sharing violation gracefully; also, mysqlbackup now copies the index file into its buffer and then closes it, instead of keeping it open for long, so the server can modify or delete the index file more easily. (Bug #22914974, Bug #26047119)