- Stop all activities on the current farm
- Export data from MySQL to csv files on disk
- Upgrade the supervisor
- Import data into PostgreSQL from the csv files
- Enable activities on the new supervisor
The order of these steps are crucial! Note that there are steps to be taken before upgrading the supervisor.
Python 2.x (2.6 or above) is now a prerequisite for supervisor installation. It comes standard on Linux and Mac, but you will have to install it on Windows in advance. Also make sure that python is added to the System PATH. It's an opt-in feature for the Python.org MSI installer. https://www.python.org/downloads/
qbadmin supervisor --set stop_activity qbadmin supervisor --set reject_submit
You will also want stop workers from picking up new work. If you want to allow workers to finish up their current frames, do:
qblock --all --purge
Then wait for all activities to stop.
Step Two: Export Data from MySQL to to .csv (Comma-separated values) files
Make sure that the MySQL server is running, and that you can connect to it using the mysql client, and that your qube table version is at 37.
If you haven't changed the database administrator user and password, you should be able to do the following on a command prompt to confirm that the MySQL server is running:
Linux: /usr/bin/mysql -u root -e 'SELECT * FROM qube.tableversion' Mac: /usr/local/mysql/bin/mysql -u root -e 'SELECT * FROM qube.tableversion' Windows: "C:\Program Files\pfx\qube\mysql\bin\mysql" -u root -e "SELECT * FROM qube.tableversion"
Make sure that the above command works and returns:
+---------+ | version | +---------+ | 37 | +---------+
If you get something less than 37 returned by the above command, it means that your current Qube supervisor version is older than 6.10-0, and that you need to update your MySQL database tables first, before you can upgrade the database schema. To do so:
Download the "
upgrade_supervisor" program suitable for your supervisor platform from http://repo.pipelinefx.com/downloads/pub/db_migration_tools/
On a command prompt, run the
programthat you just downloaded.
- Check that there weren't any critical errors reported by by
- Check that the version is now indeed updated to 37, by running the
mysql -u root -e 'SELECT * FROM qube.tableversion'command command again
Choose a destination folder on your supervisor for the MySQL csv files. Make sure that your user and the mysql server process both have write permission to this folder and all its parent folders, and that the volume is sufficiently large. Also note that a faster disk, such as an SSD, will help speed up the export/import process.
Tip title On CentOS 7.x and possibly other Linux distros, create a working directory under /opt and do the export while running as the root user
Do NOT use /tmp, /var/tmp (/usr/tmp), or any subdirectories under them. These OSs give the MySQL service its own private /tmp and /var/tmp folders, which prevents the mysqldump command from running correctly. Creating a subdirectory under /root does not work either, nor will a subdirectory in any user's home directory, since non-root users home directories are usually mode 700, so the MariaDB server can't access it.
One approach that doeswork is creating a directory under /opt and opening up the permissions:
sudo mkdir -p /opt/mysql_dump sudo chmod 755 /opt/mysql_dump
Then, install the
export_data_from_mysql.pyscript from the next step into this directory as the root user, and run the export script as root.
- Download the
export_data_from_mysql.pyscript script from http://repo.pipelinefx.com/downloads/pub/db_migration_tools/ and copy it into the destination folder.
On a command prompt, go to the destination folder, and run
export_data_from_mysql.py. Running it without any argument will create a subfolder in the current directory named "
qube_mysqldump" and dump all files into it.
You may override the dump subfolder and DB username, password, and mysql install location. Run "
export_data_from_mysql.py -h" to see the list of options.
Sit back. This process can take a long time to complete, depending on how many jobs you have on the system.
Once the process completes, make sure there were no errors reported on the terminal. Also have a look a the dump directory to confirm that there is a subfolder "
qube" and a bunch of subfolders like "
- Take a note of the dump directory location, and proceed to the next step, "Upgrade the Supervisor".
Proceed with the upgrade of the supervisor software. Using the QubeInstaller is recommended, but you can also run the individual installer packages (RPMs, DEBs, MSIs, or PKGs), should you choose.
Step Three: Upgrade the Supervisor
See See .Upgrading Qube! v7.0-0 for for details, but come back here after upgrading the supervisor software.
Step Four: Import Data into PostgreSQL from the .csv files
Do this before making any change to your farm, or submitting new jobs.
We have found that the import takes roughly 1/4 of the time for the export. Importing 10,000 jobs with 100 frames on average on a Mac system with an SSD took about 33 seconds.
Importing the previously exported data
Once you upgrade the supervisor, you are ready to import data into the new PostgreSQL server.
Make sure that PostgreSQL server is running, and accepting connections:
Linux: /usr/local/pfx/pgsql/bin/psql -p 50055 -d pfx -U qube -c "SELECT * FROM qube.tableversion" Mac: /Applications/pfx/pgsql/bin/psql -p 50055 -d pfx -U qube -c "SELECT * FROM qube.tableversion" Windows: "C:\Program Files\pfx\pgsql\bin\psql" -p 50055 -d pfx -U qube -c "SELECT * FROM qube.tableversion"
Note that this should return:
version --------- 51 (1 row)
- On a command prompt, go to the folder where you ran the export script earlier. This should be the parent folder of the "qube_mysqldump" folder, by default.
Run the the
import_data_into_pgsql.pyscript script to import data from the csv files that were generated earlier.
Linux: python /usr/local/pfx/qube/utils/pgsql/import_data_into_pgsql.py Mac: python /Applications/pfx/qube/utils/pgsql/import_data_into_pgsql.py Windows: python "C:\Program Files\pfx\qube\utils\pgsql\import_data_into_pgsql.py
- Sit back. This process will also take some time to complete, although it should be significantly faster than the export.
- Make sure there weren't any errors reported on the terminal.
Step Five: Enable Activities on the New Supervisor
Run the following commands to enable the new supervisor to accept new jobs and start dispatching jobs to workers
qbadmin supervisor --unset stop_activity qbadmin supervisor --unset reject_submit
You'll also need to unlock the workers you want to start using again. If you'd like to unlock all workers, then do:
Congratulations, you are done. Enjoy the new ride!