Upgrading from OpenNebula 5.0.x¶
This section describes the installation procedure for systems that are already running a 5.0.x OpenNebula. The upgrade to OpenNebula 5.10 can be done directly following this section, you don’t need to perform intermediate version upgrades. The upgrade will preserve all current users, hosts, resources and configurations; for both Sqlite and MySQL backends.
If you are using the vCenter drivers, there is a manual intervention required in order to upgrade to OpenNebula 5.4. Note that upgrading from OpenNebula < 5.2 to OpenNebula 5.4 is NOT supported. You need to upgrade first to OpenNebula 5.2, and then upgrade to OpenNebula 5.4.
Upgrading a Federation¶
If you have two or more 5.0.x OpenNebulas working as a Federation, you need to upgrade all of them. The upgrade does not have to be simultaneous, the slaves can be kept running while the master is upgraded.
The steps to follow are:
- Stop the MySQL replication in all the slaves
- Upgrade the master OpenNebula
- Upgrade each slave
- Resume the replication
During the time between steps 1 and 4 the slave OpenNebulas can be running, and users can keep accessing them if each zone has a local Sunstone instance. There is however an important limitation to note: all the shared database tables will not be updated in the slaves zones. This means that new user accounts, password changes, new ACL rules, etc. will not have any effect in the slaves. Read the federation architecture documentation for more details.
It is recommended to upgrade all the slave zones as soon as possible.
To perform the first step, pause the replication in each slave MySQL:
mysql> STOP SLAVE; mysql> SHOW SLAVE STATUS\G Slave_IO_Running: No Slave_SQL_Running: No
Then follow this section for the master zone. After the master has been updated to 5.10, upgrade each slave zone following this same section.
Upgrading from a High Availability deployment¶
The recommended procedure to upgrade two OpenNebulas configured in HA is to follow the upgrade procedure in a specific order. Some steps need to be executed in both servers, and others in just the active node. For the purpose of this section, we will still refer to the active node as such even after stopping the cluster, so we run the single node steps always in the same node:
- Preparation in the active node.
- Backup in the active node.
- Stop the cluster in the active node:
pcs cluster stop.
- Installation in both nodes. Before running
gem list > previous_gems.txtso we can go back to those specific
rackgems if the
pcsdrefuses to start.
- Configuration Files Upgrade in the active node.
- Database Upgrade in the active node.
- Check DB Consistency in the active node.
- Reload Start Scripts in CentOS 7 in both nodes.
- Start the cluster in the active node.
Before proceeding, make sure you don’t have any VMs in a transient state (prolog, migr, epil, save). Wait until these VMs get to a final state (runn, suspended, stopped, done). Check the Managing Virtual Machines guide for more information on the VM life-cycle.
Stop OpenNebula and any other related services you may have running: OneFlow, EC2, and Sunstone. Use preferably the system tools, like systemctl or service as root in order to stop the services.
Backup the configuration files located in /etc/one. You don’t need to do a manual backup of your database, the onedb command will perform one automatically.
cp -r /etc/one /etc/one.$(date +'%Y-%m-%d')
Make sure to run the
install_gems tool, as the new OpenNebula version may have different gem requirements.
install_gems you get a message asking to overwrite files for aws executables you can safely answer “yes”.
It is highly recommended not to keep your current
oned.conf, and update the
oned.conf file shipped with OpenNebula 5.10 to your setup. If for any reason you plan to preserve your current
oned.conf file, read the Compatibility Guide and the complete oned.conf reference for 5.0 and 5.10 versions.
Configuration Files Upgrade¶
If you haven’t modified any configuration files, the package managers will replace the configuration files with their newer versions and no manual intervention is required.
If you have customized any configuration files under
/etc/one we recommend you to follow these steps regardless of the platform/linux distribution.
- Install the new packages (already performed)
- Compare the old and new configuration files:
diff -ur /etc/one.YYYY-MM-DD /etc/one. Or you can use graphical diff-tools like
meldto compare both directories, which are very useful in this step.
- Edit the new files and port all the customizations from the previous version.
- You should never overwrite the configuration files with older versions.
The database schema and contents are incompatible between versions. The OpenNebula daemon checks the existing DB version, and will fail to start if the version found is not the one expected, with the message ‘Database version mismatch’.
You can upgrade the existing DB with the ‘onedb’ command. You can specify any Sqlite or MySQL database. Check the onedb reference for more information.
Make sure at this point that OpenNebula is not running. If you installed from packages, the service may have been started automatically.
For environments in a Federation: Before upgrading the master, make sure that all the slaves have the MySQL replication paused.
After you install the latest OpenNebula, and fix any possible conflicts in oned.conf, you can issue the ‘onedb upgrade -v’ command. The connection parameters have to be supplied with the command line options, see the onedb manpage for more information. Some examples:
onedb upgrade -v --sqlite /var/lib/one/one.db
onedb upgrade -v -S localhost -u oneadmin -p oneadmin -d opennebula
If everything goes well, you should get an output similar to this one:
$ onedb upgrade -v -u oneadmin -d opennebula MySQL Password: Version read: Shared tables 4.11.80 : OpenNebula 4.12.1 daemon bootstrap Local tables 4.11.80 : OpenNebula 4.12.1 daemon bootstrap MySQL dump stored in /var/lib/one/mysql_localhost_opennebula.sql Use 'onedb restore' or restore the DB using the mysql command: mysql -u user -h server -P port db_name < backup_file >>> Running migrators for shared tables ... >>> Running migrators for local tables ... > Done in 41.93s Database migrated from 4.11.80 to 4.13.80 (OpenNebula 4.13.80) by onedb command. Total time: 41.93s
Make sure you keep the backup file. If you face any issues, the onedb command can restore this backup, but it won’t downgrade databases to previous versions.
Check DB Consistency¶
After the upgrade is completed, you should run the command
First, move the 5.0.x backup file created by the upgrade command to a safe place.
mv /var/lib/one/mysql_localhost_opennebula.sql /path/for/one-backups/
Then execute the following command:
$ onedb fsck -S localhost -u oneadmin -p oneadmin -d opennebula MySQL dump stored in /var/lib/one/mysql_localhost_opennebula.sql Use 'onedb restore' or restore the DB using the mysql command: mysql -u user -h server -P port db_name < backup_file Total errors found: 0
Resume the Federation¶
This section applies only to environments working in a Federation.
For the master zone: This step is not necessary.
For a slave zone: The MySQL replication must be resumed now.
Do not copy the server-id from this example, each slave should already have a unique ID.
- Start the slave MySQL process and check its status. It may take a while to copy and apply all the pending commands.
mysql> START SLAVE; mysql> SHOW SLAVE STATUS\G
SHOW SLAVE STATUS output will provide detailed information, but to confirm that the slave is connected to the master MySQL, take a look at these columns:
Slave_IO_State: Waiting for master to send event Slave_IO_Running: Yes Slave_SQL_Running: Yes
Reload Start Scripts in CentOS 7¶
In order for the system to re-read the configuration files you should issue the following command after the installation of the new packages:
Update the Drivers¶
You should be able now to start OpenNebula as usual, running
service opennebula start as
root. At this point, as
oneadmin user, execute
onehost sync to update the new drivers in the hosts.
onehost sync is important. If the monitorization drivers are not updated, the hosts will behave erratically.
OpenNebula will continue the monitoring and management of your previous Hosts and VMs.
As a measure of caution, look for any error messages in oned.log, and check that all drivers are loaded successfully. After that, keep an eye on oned.log while you issue the onevm, onevnet, oneimage, oneuser, onehost list commands. Try also using the show subcommand for some resources.
Restoring the Previous Version¶
If for any reason you need to restore your previous OpenNebula, follow these steps:
- With OpenNebula 5.10 still installed, restore the DB backup using ‘onedb restore -f’
- Uninstall OpenNebula 5.10, and install again your previous version.
- Copy back the backup of /etc/one you did to restore your configuration.
If the MySQL database password contains special characters, such as
#, the onedb command will fail to connect to it.
The workaround is to temporarily change the oneadmin’s password to an ASCII string. The set password statement can be used for this:
$ mysql -u oneadmin -p mysql> SET PASSWORD = PASSWORD('newpass');