I would like to use MySQL Cluster for catalog storage. MySQL
Cluster uses master/master replication and a shared nothing
architecture. The primary difference from regular MySQL is the
use of the NDB (Network DataBase) storage engine, which is what
provides the redundancy. NDB is an ACID compliant transaction
oriented storage engine supporting the READ COMMITTED isolation
level.
The advantages of using MySQL Cluster are that if I lose a
data node, I don't lose any data and everything continues as if
nothing had happened. A second advantage is that if I lose a server
node, Bacula can simply switch to a different server node to get
full access to the same data (right now, I'm using haproxy, installed
on the system running the director, to provide transparent failover
between server nodes). I didn't install MySQL Cluster just for
Bacula. I already had it running in production. Given how important
the catalog is to Bacula's operation, I figured it would be good
to use a database with built-in redundancy for it.
The only problem that I have detected so far is that NDB
doesn't support indexing on the BLOB or TEXT data types. The
comments in make_mysql_tables state that BLOB is used in order to
get case-insensitive sorts. The work around for this is to use
VARCHAR with an appropriate character set and/or collation order.
An example would be to take this:
CREATE TABLE Client (
ClientId INTEGER UNSIGNED NOT NULL AUTO_INCREMENT,
Name TINYBLOB NOT NULL,
Uname TINYBLOB NOT NULL, /* full uname -a of client */
AutoPrune TINYINT DEFAULT 0,
FileRetention BIGINT UNSIGNED DEFAULT 0,
JobRetention BIGINT UNSIGNED DEFAULT 0,
UNIQUE (Name(128)),
PRIMARY KEY(ClientId)
);
and replace it with this:
CREATE TABLE Client (
ClientId INTEGER UNSIGNED NOT NULL AUTO_INCREMENT,
Name VARCHAR(255) CHARACTER SET BINARY NOT NULL,
Uname TINYBLOB NOT NULL, /* full uname -a of client */
AutoPrune TINYINT DEFAULT 0,
FileRetention BIGINT UNSIGNED DEFAULT 0,
JobRetention BIGINT UNSIGNED DEFAULT 0,
UNIQUE (Name),
PRIMARY KEY(ClientId)
);
See: https://dev.mysql.com/doc/refman/5.7/en/globalization.html
and in particular:
https://dev.mysql.com/doc/refman/5.7/en/charset-binary-collations.html
I did my initial testing of Bacula using sqlite3. I ran into
this problem while switching databases in order to move to production.
Another comment in the script says that one is on their own if they
mess with the schema, which is a fair comment. So, I was wondering
if anybody had any comments on the idea of changing BLOB to VARCJHAR?
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Bacula-devel mailing list
Bacula-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-devel
