matiri

• Parallel MySql backup script storing backup info in Sqlite3 database.

Features:

• Multi-Server: Multiple MySQL servers are supported whether they are co-located on the same or separate physical servers.
• Parallel: Each database on the server to be backed up is done separately, in parallel (concurrency settable: default: 3)
• Compressed: Each database backup compressed
• Checksummed: SHA256 of each compressed backup file stored and the archive of all files
• Archived: All database backups tar'ed together into single file
• Recorded: Backup information stored in Sqlite3 database

MySql Credentials

matiri invokes a script called mysql.sh (must be in same directory as matiri script) to obtain MySQL server host, port number, user, password, and databases to include/exclude. A default implementation is supplied but should be modified to be more secure. For obvious security reasons, this file should be in mode 700. Note that only include or exclude are allowed, but not both.

Running

1. Alter the mysql.sh script to include one or more host, port, user, password, and included or excluded databases
2. Alter matiri_options to modify backup destinations, concurrency, naming conventions`
3. Start matiri

Directory Structure

All backup files are grouped by year and month, which can be modified within matiri_options $BASE_DESTINATION_DIR/YYYY/MM

Four files are produced:

1. mysql_backup_YYYY-MM-DD_ID.tar
   • tar of database backup files (see below)
2. mysql_backup_YYYY-MM-DD_ID.tar.sha256
   • SHA1 of #1
3. mysql_backup_YYYY-MM-DD_ID.meta
   • Info about the backup (redundant with Sqlite3 information)
4. mysql_backup_YYYY-MM-DD_ID.err
   • stderr output from backup process

Database Backup files

The above (#1) tar file is made up of:    For each database being backed up, two files are produced:

1. database__DBNAME.sql.gz
   • gzip of mysqldump output
2. database__DBNAME.sql.gz.sha1
   • SHA1 of #1

Dependencies

1. Sqlite3
2. mysqldump (Tested on mysqldump Ver 10.13 Distrib 5.5.34, for Linux (x86_64))
3. Standard Linux tools (tar, gzip, date, awk, xargs, sha1sum)

Sqlite3 Database

Default sqlite location: $BASE_DESTINATION_DIR/backups.sqlite3.db

• Each time matiri is run, an entry in the 'backup_event' table is created for each defined server (multiple MySQL servers = multiple backup events).
• The record is added, indicating a backup_event was started, with the 'completed' column set to -999 (not completed)
• For each of the databases to be backed up:
  • A database record is added before the database backup starts, with the backup_event id as the forign key 'backup_id'. The 'completed' column set to -999 (not completed).
  • If this database backup completes successfully, the record is updated with the 'completed' column set to 0 (completed), the end_time is set, the size ('bytes') and the SHA256 of the backup file are recorded.
• If the backup event has successfully executed, the backup_event is updated with the 'completed' column set to 0 (completed), the end_time is set, the size ('bytes') and the SHA256 of the tar file are recorded.

Database schema:

    CREATE TABLE backup_event (id INTEGER PRIMARY KEY, completed int NOT NULL, comments text, 
           host varchar(255) NOT NULL, port int NOT NULL, 
           start_time DATETIME not null, end_time DATETIME not null, 
           user varchar(64), bytes bigint NOT NULL, file text, sha256 char(64) NOT NULL, 
           error default NULL);


    CREATE TABLE database (id INTEGER PRIMARY KEY,  completed int NOT NULL, backup_id INTEGER, 
           database varchar(255) NOT NULL, file text, 
           start_time DATETIME not null, end_time DATETIME not null, 
           bytes bigint NOT NULL, sha256 char(64) NOT NULL, 
           error default NULL, FOREIGN KEY(backup_id) REFERENCES backup(id));

    CREATE INDEX backup_start_time on backup(start_time);

    CREATE INDEX database_start_time on database(start_time);

Perusing Sqlite3 Backup Database

Using the Sqlite3 command line tool

sqlite> select * from backup_event;
id|completed|comments|host|port|start_time|end_time|user|bytes|file|sha256|error
12|0||localhost|3306|2013-11-23 17:23:15|2013-11-23 17:23:15|backups|20480|/home/newtong/backups/2013/11/mysql_backup_2013-11-23_12.tar|6f216d2a4811382b66b25480328b385bab54e7531f73bf2aa5262b00b030017c|
13|0||localhost|3306|2013-11-23 17:23:26|2013-11-23 17:23:27|backups|20480|/home/newtong/backups/2013/11/mysql_backup_2013-11-23_13.tar|da5721440c8577a3b250232ba2e901350ea9a34876212312a5b9a28206ae6d33|
14|0||localhost|3306|2013-11-23 17:23:50|2013-11-23 17:23:51|backups|20480|/home/newtong/backups/2013/11/mysql_backup_2013-11-23_14.tar|f2a9b41e4157da803d79cf385db17dad1d273e48b352eba2cd0209eaf90fa2e9|
15|0||localhost|3306|2013-11-23 17:24:08|2013-11-23 17:24:15|backups|29399040|/home/newtong/backups/2013/11/mysql_backup_2013-11-23_15.tar|8ef6dbdb3537361e48bce1d3eeb3c114d25ebf8d7eb808312384035221f20e32|

sqlite> select * from database where backup_id = 15;
id|completed|backup_id|database|file|start_time|end_time|bytes|sha256|error
61|0|15|performance_schema|database__performance_schema.gz|2013-11-23 17:24:09|2013-11-23 17:24:15|1100|2e7ea55832e3fbb62ee1370a1f0b6ffef2415aba79a129b419181195588b6c27|
62|0|15|information_schema|database__information_schema.gz|2013-11-23 17:24:09|2013-11-23 17:24:15|395504|694688b16377916f31f9dbe2a8647928a6cbb4cd5419767b3335c5ca7e5e5f37|
63|0|15|mysql|database__mysql.gz|2013-11-23 17:24:09|2013-11-23 17:24:15|142109|a72789bedfdcc73ea419750fd4904fd1c859e175f23be208b58d1c262e45eae5|
64|0|15|specify_dao_live|database__specify_dao_live.gz|2013-11-23 17:24:09|2013-11-23 17:24:15|28842782|1d91517672bae1ae39294d3e4f819e79eb6d0c4325d9c30d2d78303c34e189cd|
sqlite> 

TODOs

0. Command line parameters for backup location and dynamically setting credentials (mysql.sh) script
1. Better docs here and in scripts
2. Fix scripts to better adhere to the Google bash style guide
3. Alternate compression: {p}bzip2 should be easy;
4. Scripts to both delete old backups files and remove entries from the Sqlite3 db (keep in sync)
5. Simple web server app to peruse backup information in Sqlite3 DB
6. Explanation of the Sqlite3 fields stored.
7. Implement support for passwordless backups

Name

Named after the Matiri River, New Zealand. I have been very close on several occasions but never have seen the river.

Acknowledgements

Partially developed at Agriculture and Agri-Food Canada, Ottawa, Ontario.