Migration operations

Migrating the base port

It is possible to change the base port of an existing Data Science Studio instance, by editing the installation configuration file:

  • Stop DSS

    DATADIR/bin/dss stop

  • Edit the DATADIR/install.ini installation configuration file

    [server]
port = NEW_BASE_PORT

  • Regenerate DSS configuration

    DATADIR/bin/dssadmin regenerate-config

  • Restart DSS

    DATADIR/bin/dss start

Migrating the installation directory

It is possible to change the installation directory of an existing Data Science Studio instance, by replaying the installer in “upgrade” mode:

  • Stop DSS

    DATADIR/bin/dss stop

  • Move the installed kit to its new location (or unpack the .tar.gz distribution archive to a new location)

    mv OLD_DIR/dataiku-dss-VERSION NEW_DIR/
# or
cd NEW_DIR
tar xf /PATH/TO/dataiku-dss-VERSION.tar.gz

  • Run the installer in upgrade mode

    NEW_DIR/dataiku-dss-VERSION/installer.sh -d DATA_DIR -u

  • If you have configured DSS with User Isolation Framework, run the impersonation installation step as root from DATADIR

    ./bin/dssadmin install-impersonation DSSUSER

  • Restart DSS

    DATADIR/bin/dss start

Migrating the data directory

It is possible to change the path of the data directory of an existing Data Science Studio instance, by replaying the installer in “upgrade” mode. Note that the Python virtual environment has to be rebuilt after migration, as described here. This is because Python virtual environments embed their installation path in various places.

  • Stop DSS

    DATADIR/bin/dss stop

  • Save the list of locally-installed Python packages

    DATADIR/bin/pip freeze -l >local-python-packages.txt

  • Move the data directory to its new location

    mv DATADIR NEWDATADIR

  • Remove the Python virtualenv, keeping a backup copy

    mv NEWDATADIR/pyenv NEWDATADIR/pyenv.backup

  • Run the installer in upgrade mode. This recreates the Python virtualenv

    dataiku-dss-VERSION/installer.sh -d NEWDATADIR -u

  • Reinstall locally-installed Python packages (if any)

    NEWDATADIR/bin/pip install -r local-python-packages.txt

  • If you have configured DSS with User Isolation Framework:

    • As root, edit the file /etc/dataiku-security/INSTALL_ID/security/security-config.ini to update, in section dirs, the path under dss_data_dir. For information on finding INSTALL_ID, see Initial Setup.
    • Still as root, update the paths in sudoers file /etc/sudoers.d/dataiku-dss-<dss_user>-<dss_instance_id> to point to NEWDATADIR
    • Still as root, either:
      • Run NEWDATADIR/bin/dssadmin install-impersonation DSSUSER to update NEWDATADIR/security/execwrapper.sh
      • Or, edit NEWDATADIR/security/execwrapper.sh and update the path to CONFIGDIR to point to NEWDATADIR

  • Restart DSS

    NEWDATADIR/bin/dss start

  • If you have configured DSS to start automatically on server boot:

    • Re-run the install-boot.sh command (as root), or
    • Directly adjust the data directory path in the service configuration file, which you can locate with: ls -l /etc/default/dataiku* /etc/sysconfig/dataiku*

  • When everything is considered stable, remove the backup

    rm -rf NEWDATADIR/pyenv.backup

Note

If you could not save the list of locally-installed Python packages before migration (step 2 above), it is possible to reconstitute it by looking at the package installation directory:

ls -l NEWDATADIR/pyenv.backup/lib/python?.?/site-packages