This tool can be used to check for and remove orphaned BLOB column parts from NDB tables, as well as to generate a file listing any orphaned parts. It is sometimes useful in diagnosing and repairing corrupted or damaged NDB tables containing BLOB or TEXT columns.

The basic syntax for ndb_blob_tool is shown here:

ndb_blob_tool [options] table [column, ...]

Unless you use the --help option, you must specify an action to be performed by including one or more of the options --check-orphans, --delete-orphans, or --dump-file. These options cause ndb_blob_tool to check for orphaned BLOB parts, remove any orphaned BLOB parts, and generate a dump file listing orphaned BLOB parts, respectively, and are described in more detail later in this section.

You must also specify the name of a table when invoking ndb_blob_tool. In addition, you can optionally follow the table name with the (comma-separated) names of one or more BLOB or TEXT columns from that table. If no columns are listed, the tool works on all of the table's BLOB and TEXT columns. If you need to specify a database, use the --database (-d) option.

The --verbose option provides additional information in the output about the tool's progress.

The following table includes options that are specific to ndb_blob_tool. Additional descriptions follow the table. For options common to most NDB Cluster programs (including ndb_blob_tool), see Section 18.4.27, “Options Common to NDB Cluster Programs — Options Common to NDB Cluster Programs”.

Example

First we create an NDB table in the test database, using the CREATE TABLE statement shown here:

USE test;

CREATE TABLE btest (
    c0 BIGINT UNSIGNED NOT NULL AUTO_INCREMENT PRIMARY KEY,
    c1 TEXT,
    c2 BLOB
)   ENGINE=NDB;

Then we insert a few rows into this table, using a series of statements similar to this one:

INSERT INTO btest VALUES (NULL, 'x', REPEAT('x', 1000));

When run with --check-orphans against this table, ndb_blob_tool generates the following output:

shell> ndb_blob_tool --check-orphans --verbose -d test btest
connected
processing 2 blobs
processing blob #0 c1 NDB$BLOB_19_1
NDB$BLOB_19_1: nextResult: res=1
total parts: 0
orphan parts: 0
processing blob #1 c2 NDB$BLOB_19_2
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=0
NDB$BLOB_19_2: nextResult: res=1
total parts: 10
orphan parts: 0
disconnected

NDBT_ProgramExit: 0 - OK

The tool reports that there are no NDB BLOB column parts associated with column c1, even though c1 is a TEXT column. This is due to the fact that, in an NDB table, only the first 256 bytes of a BLOB or TEXT column value are stored inline, and only the excess, if any, is stored separately; thus, if there are no values using more than 256 bytes in a given column of one of these types, no BLOB column parts are created by NDB for this column. See Section 11.7, “Data Type Storage Requirements”, for more information.