A tool to migrate tables between two clusters, leveraging the DataStax Bulk Loader (DSBulk) to perform the actual data migration.
The tool provides the following main commands:
migrate-live
starts a live data migration using a pre-existing DSBulk installation, or
alternatively, the embedded DSBulk version. A "live" migration means that the data migration will
start immediately and will be performed by this migrator tool through the desired DSBulk
installation.
generate-script
generates a migration script that, once executed, will perform the desired data
migration, using a pre-existing DSBulk installation. Please note: this command does not actually
migrate the data; it only generates the migration script.
generate-ddl
reads the schema from the origin cluster and generates CQL files to recreate it in
the target cluster.
Build is done with Maven:
mvn clean package
The build produces two distributable fat jars:
dsbulk-migrator-<VERSION>-embedded-driver.jar
: contains an embedded Java driver; suitable for
live migrations using an external DSBulk, or for script generation. This jar is NOT suitable for
live migrations using an embedded DSBulk, since no DSBulk classes are present.
dsbulk-migrator-<VERSION>-embedded-dsbulk.jar
: contains an embedded DSBulk and an embedded Java
driver; suitable for all operations. Note that this jar is much bigger than the previous one, due
to the presence of DSBulk classes.
The project contains a few integration tests. Run them with:
mvn clean verify
The integration tests require Simulacron. Be sure to meet all the prerequisites before running the tests.
Launch the tool as follows:
java -jar /path/to/dsbulk-migrator.jar (migrate-live|generate-script|generate-ddl) [OPTIONS]
When doing a live migration, the options are used to effectively configure DSBulk and to connect to the clusters.
When generating a migration script, most options serve as default values in the generated scripts. Note however that, even when generating scripts, this tool still needs to access the origin cluster in order to gather metadata about the tables to migrate.
When generating a DDL file, only a few options are meaningful. Since DSBulk is not used, and the import cluster is never contacted, import options and DSBulk-related options are ignored. The tool still needs to access the origin cluster in order to gather metadata about the keyspaces and tables to generate DDL statements for.
The following options are available for the migrate-live
command; most have sensible default values and do not
need to be specified, unless you want to override the default value:
-c, --dsbulk-cmd=CMD The external DSBulk command to use. Ignored if the embedded DSBulk is
being used. The default is simply 'dsbulk', assuming that the
command is available through the PATH variable contents.
-d, --data-dir=PATH The directory where data will be exported to and imported from.The
default is a 'data' subdirectory in the current working directory.
The data directory will be created if it does not exist. Tables will
be exported and imported in subdirectories of the data directory
specified here; there will be one subdirectory per keyspace inside
the data directory, then one subdirectory per table inside each
keyspace directory.
-e, --dsbulk-use-embedded Use the embedded DSBulk version instead of an external one. The
default is to use an external DSBulk command.
--export-bundle=PATH The path to a secure connect bundle to connect to the origin cluster,
if that cluster is a DataStax Astra cluster. Options --export-host
and --export-bundle are mutually exclusive.
--export-consistency=CONSISTENCY
The consistency level to use when exporting data. The default is
LOCAL_QUORUM.
--export-dsbulk-option=OPT=VALUE
An extra DSBulk option to use when exporting. Any valid DSBulk option
can be specified here, and it will passed as is to the DSBulk
process. DSBulk options, including driver options, must be passed as
'--long.option.name=<value>'. Short options are not supported.
--export-host=HOST[:PORT]
The host name or IP and, optionally, the port of a node from the
origin cluster. If the port is not specified, it will default to
9042. This option can be specified multiple times. Options
--export-host and --export-bundle are mutually exclusive.
--export-max-concurrent-files=NUM|AUTO
The maximum number of concurrent files to write to. Must be a positive
number or the special value AUTO. The default is AUTO.
--export-max-concurrent-queries=NUM|AUTO
The maximum number of concurrent queries to execute. Must be a
positive number or the special value AUTO. The default is AUTO.
--export-max-records=NUM
The maximum number of records to export for each table. Must be a
positive number or -1. The default is -1 (export the entire table).
--export-password The password to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all. Omit the parameter value to be prompted for
the password interactively.
--export-splits=NUM|NC The maximum number of token range queries to generate. Use the NC
syntax to specify a multiple of the number of available cores, e.g.
8C = 8 times the number of available cores. The default is 8C. This
is an advanced setting; you should rarely need to modify the default
value.
--export-username=STRING
The username to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all.
-h, --help Displays this help message.
--import-bundle=PATH The path to a secure connect bundle to connect to the target cluster,
if that cluster is a DataStax Astra cluster. Options --import-host
and --import-bundle are mutually exclusive.
--import-consistency=CONSISTENCY
The consistency level to use when importing data. The default is
LOCAL_QUORUM.
--import-default-timestamp=<defaultTimestamp>
The default timestamp to use when importing data. Must be a valid
instant in ISO-8601 syntax. The default is 1970-01-01T00:00:00Z.
--import-dsbulk-option=OPT=VALUE
An extra DSBulk option to use when importing. Any valid DSBulk option
can be specified here, and it will passed as is to the DSBulk
process. DSBulk options, including driver options, must be passed as
'--long.option.name=<value>'. Short options are not supported.
--import-host=HOST[:PORT]
The host name or IP and, optionally, the port of a node from the
target cluster. If the port is not specified, it will default to
9042. This option can be specified multiple times. Options
--import-host and --import-bundle are mutually exclusive.
--import-max-concurrent-files=NUM|AUTO
The maximum number of concurrent files to read from. Must be a
positive number or the special value AUTO. The default is AUTO.
--import-max-concurrent-queries=NUM|AUTO
The maximum number of concurrent queries to execute. Must be a
positive number or the special value AUTO. The default is AUTO.
--import-max-errors=NUM
The maximum number of failed records to tolerate when importing data.
The default is 1000. Failed records will appear in a load.bad file
inside the DSBulk operation directory.
--import-password The password to use to authenticate against the target cluster.
Options --import-username and --import-password must be provided
together, or not at all. Omit the parameter value to be prompted for
the password interactively.
--import-username=STRING
The username to use to authenticate against the target cluster.
Options --import-username and --import-password must be provided
together, or not at all.
-k, --keyspaces=REGEX A regular expression to select keyspaces to migrate. The default is to
migrate all keyspaces except system keyspaces, DSE-specific
keyspaces, and the OpsCenter keyspace. Case-sensitive keyspace names
must be entered in their exact case.
-l, --dsbulk-log-dir=PATH The directory where DSBulk should store its logs. The default is a
'logs' subdirectory in the current working directory. This
subdirectory will be created if it does not exist. Each DSBulk
operation will create a subdirectory inside the log directory
specified here.
--max-concurrent-ops=NUM
The maximum number of concurrent operations (exports and imports) to
carry. Default is 1. Set this to higher values to allow exports and
imports to occur concurrently; e.g. with a value of 2, each table
will be imported as soon as it is exported, while the next table is
being exported.
--skip-truncate-confirmation
Skip truncate confirmation before actually truncating tables. Only
applicable when migrating counter tables, ignored otherwise.
-t, --tables=REGEX A regular expression to select tables to migrate.The default is to
migrate all tables inside the keyspaces that were selected for
migration with --keyspaces. Case-sensitive table names must be
entered in their exact case.
--table-types=regular|counter|all
The table types to migrate (regular, counter, or all). Default is all.
--truncate-before-export
Truncate tables before the export instead of after. Default is to
truncate after the export. Only applicable when migrating counter
tables, ignored otherwise.
-w, --dsbulk-working-dir=PATH
The directory where DSBulk should be executed. Ignored if the embedded
DSBulk is being used. If unspecified, it defaults to the current
working directory.
The following options are available for the generate-script
command; most have sensible default values and do not
need to be specified, unless you want to override the default value:
-c, --dsbulk-cmd=CMD The DSBulk command to use. The default is simply 'dsbulk', assuming
that the command is available through the PATH variable contents.
-d, --data-dir=PATH The directory where migration scripts will be generated. The default
is a 'data' subdirectory in the current working directory. The data
directory will be created if it does not exist.
--export-bundle=PATH The path to a secure connect bundle to connect to the origin cluster,
if that cluster is a DataStax Astra cluster. Options --export-host
and --export-bundle are mutually exclusive.
--export-consistency=CONSISTENCY
The consistency level to use when exporting data. The default is
LOCAL_QUORUM.
--export-dsbulk-option=OPT=VALUE
An extra DSBulk option to use when exporting. Any valid DSBulk option
can be specified here, and it will passed as is to the DSBulk
process. DSBulk options, including driver options, must be passed as
'--long.option.name=<value>'. Short options are not supported.
--export-host=HOST[:PORT]
The host name or IP and, optionally, the port of a node from the
origin cluster. If the port is not specified, it will default to
9042. This option can be specified multiple times. Options
--export-host and --export-bundle are mutually exclusive.
--export-max-concurrent-files=NUM|AUTO
The maximum number of concurrent files to write to. Must be a positive
number or the special value AUTO. The default is AUTO.
--export-max-concurrent-queries=NUM|AUTO
The maximum number of concurrent queries to execute. Must be a
positive number or the special value AUTO. The default is AUTO.
--export-max-records=NUM
The maximum number of records to export for each table. Must be a
positive number or -1. The default is -1 (export the entire table).
--export-password The password to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all. Omit the parameter value to be prompted for
the password interactively.
--export-splits=NUM|NC The maximum number of token range queries to generate. Use the NC
syntax to specify a multiple of the number of available cores, e.g.
8C = 8 times the number of available cores. The default is 8C. This
is an advanced setting; you should rarely need to modify the default
value.
--export-username=STRING
The username to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all.
-h, --help Displays this help message.
--import-bundle=PATH The path to a secure connect bundle to connect to the target cluster,
if that cluster is a DataStax Astra cluster. Options --import-host
and --import-bundle are mutually exclusive.
--import-consistency=CONSISTENCY
The consistency level to use when importing data. The default is
LOCAL_QUORUM.
--import-default-timestamp=<defaultTimestamp>
The default timestamp to use when importing data. Must be a valid
instant in ISO-8601 syntax. The default is 1970-01-01T00:00:00Z.
--import-dsbulk-option=OPT=VALUE
An extra DSBulk option to use when importing. Any valid DSBulk option
can be specified here, and it will passed as is to the DSBulk
process. DSBulk options, including driver options, must be passed as
'--long.option.name=<value>'. Short options are not supported.
--import-host=HOST[:PORT]
The host name or IP and, optionally, the port of a node from the
target cluster. If the port is not specified, it will default to
9042. This option can be specified multiple times. Options
--import-host and --import-bundle are mutually exclusive.
--import-max-concurrent-files=NUM|AUTO
The maximum number of concurrent files to read from. Must be a
positive number or the special value AUTO. The default is AUTO.
--import-max-concurrent-queries=NUM|AUTO
The maximum number of concurrent queries to execute. Must be a
positive number or the special value AUTO. The default is AUTO.
--import-max-errors=NUM
The maximum number of failed records to tolerate when importing data.
The default is 1000. Failed records will appear in a load.bad file
inside the DSBulk operation directory.
--import-password The password to use to authenticate against the target cluster.
Options --import-username and --import-password must be provided
together, or not at all. Omit the parameter value to be prompted for
the password interactively.
--import-username=STRING
The username to use to authenticate against the target cluster.
Options --import-username and --import-password must be provided
together, or not at all.
-k, --keyspaces=REGEX A regular expression to select keyspaces to migrate. The default is to
migrate all keyspaces except system keyspaces, DSE-specific
keyspaces, and the OpsCenter keyspace. Case-sensitive keyspace names
must be entered in their exact case.
-l, --dsbulk-log-dir=PATH The directory where DSBulk should store its logs. The default is a
'logs' subdirectory in the current working directory. This
subdirectory will be created if it does not exist. Each DSBulk
operation will create a subdirectory inside the log directory
specified here.
-t, --tables=REGEX A regular expression to select tables to migrate.The default is to
migrate all tables inside the keyspaces that were selected for
migration with --keyspaces. Case-sensitive table names must be
entered in their exact case.
--table-types=regular|counter|all
The table types to migrate (regular, counter, or all). Default is all.
The following options are available for the generate-ddl
command; most have sensible default values and do not
need to be specified, unless you want to override the default value:
-a, --optimize-for-astra Produce CQL scripts optimized for DataStax Astra. Astra does not allow
some options in DDL statements; using this option, forbidden options
will be omitted from the generated CQL files.
-d, --data-dir=PATH The directory where CQL files will be generated. The default is a
'data' subdirectory in the current working directory. The data
directory will be created if it does not exist.
--export-bundle=PATH The path to a secure connect bundle to connect to the origin cluster,
if that cluster is a DataStax Astra cluster. Options --export-host
and --export-bundle are mutually exclusive.
--export-host=HOST[:PORT]
The host name or IP and, optionally, the port of a node from the
origin cluster. If the port is not specified, it will default to
9042. This option can be specified multiple times. Options
--export-host and --export-bundle are mutually exclusive.
--export-password The password to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all. Omit the parameter value to be prompted for
the password interactively.
--export-username=STRING
The username to use to authenticate against the origin cluster.
Options --export-username and --export-password must be provided
together, or not at all.
-h, --help Displays this help message.
-k, --keyspaces=REGEX A regular expression to select keyspaces to export. The default is to
export all keyspaces except system keyspaces, DSE-specific
keyspaces, and the OpsCenter keyspace. Case-sensitive keyspace names
must be entered in their exact case.
-t, --tables=REGEX A regular expression to select tables to export.The default is to
export all tables inside the keyspaces that were selected for export
with --keyspaces. Case-sensitive table names must be entered in
their exact case.
Global help is available as follows:
java -jar /path/to/dsbulk-migrator-embedded-dsbulk.jar --help
This will print help about the available commands.
Per-command help is available as follows:
java -jar /path/to/dsbulk-migrator-embedded-dsbulk.jar COMMAND --help
This will print detailed help about the selected command along with all the available options for the specified command.
Generate a migration script to migrate from an existing cluster to an Astra cluster:
java -jar target/dsbulk-migrator-<VERSION>-embedded-driver.jar migrate-live \
--data-dir=/path/to/data/dir \
--dsbulk-cmd=${DSBULK_ROOT}/bin/dsbulk \
--dsbulk-log-dir=/path/to/log/dir \
--export-host=my-origin-cluster.com \
--export-username=user1 \
--export-password=s3cr3t \
--import-bundle=/path/to/bundle \
--import-username=user1 \
--import-password=s3cr3t
Migrate live from an existing cluster to an Astra cluster using an external DSBulk installation; passwords will be prompted interactively:
java -jar target/dsbulk-migrator-<VERSION>-embedded-driver.jar migrate-live \
--data-dir=/path/to/data/dir \
--dsbulk-cmd=${DSBULK_ROOT}/bin/dsbulk \
--dsbulk-log-dir=/path/to/log/dir \
--export-host=my-origin-cluster.com \
--export-username=user1 \
--export-password # password will be prompted \
--import-bundle=/path/to/bundle \
--import-username=user1 \
--import-password # password will be prompted
Migrate live from an existing cluster to an Astra cluster using the embedded DSBulk installation; passwords will be prompted interactively; extra DSBulk options are passed:
java -jar target/dsbulk-migrator-<VERSION>-embedded-dsbulk.jar migrate-live \
--data-dir=/path/to/data/dir \
--dsbulk-use-embedded \
--dsbulk-log-dir=/path/to/log/dir \
--export-host=my-origin-cluster.com \
--export-username=user1 \
--export-password # password will be prompted \
--export-dsbulk-option "--connector.csv.maxCharsPerColumn=65536" \
--export-dsbulk-option "--executor.maxPerSecond=1000" \
--import-bundle=/path/to/bundle \
--import-username=user1 \
--import-password # password will be prompted \
--import-dsbulk-option "--connector.csv.maxCharsPerColumn=65536" \
--import-dsbulk-option "--executor.maxPerSecond=1000"
Note that for the last example, you must use the dsbulk-migrator-<VERSION>-embedded-dsbulk.jar
fat
jar, otherwise, an error will be raised because no embedded DSBulk can be found.
Generate DDL files to re-create the origin schema in an Astra target cluster:
java -jar target/dsbulk-migrator-<VERSION>-embedded-driver.jar generate-ddl \
--data-dir=/path/to/data/dir \
--export-host=my-origin-cluster.com \
--export-username=user1 \
--export-password=s3cr3t \
--optimize-for-astra