For installation where tungsten-installer was used to perform the original installation and deployment, the update tool must be used. This includes all installations where the original deployment was in a release of Tungsten Replicator 2.1.0 or earlier; any installation where tungsten-installer was used with Tungsten Replicator 2.1.1, or where an installation originally took place using tungsten-installer that has been updated to Tungsten Replicator 2.1.1 or later.
To perform the upgrade:
Download the latest Tungsten Replicator package to your staging server.
Stop the replicator service on the host:
The replicator service must be switched off on each machine before the upgrade process is started. Multiple machines can be updated at the same time, but each datasource must have been stopped before the upgrade process is started. Failing to shutdown the replicator before running the upgrade process will generate an error:
ERROR >> host1 >> The replicator in /opt/continuent is still running. » You must stop it before installation can continue. (HostReplicatorServiceRunningCheck)
Run the ./tools/update command.
To update a local installation, you
must supply the
parameter to specify the installation location of your service.
./tools/update --release-directory=/opt/continuentINFO >> host1 >> Getting services list INFO >> host1 >> . Processing services command... NAME VALUE ---- ----- appliedLastSeqno: 5243 appliedLatency : 0.405 role : master serviceName : firstrep serviceType : local started : true state : ONLINE Finished services command... NOTE >> host1 >> Deployment finished
The replicator will be upgraded to the latest version. If your installation has only a single service, the service will be restarted automatically. If you have multiple services, the replicator will need to be restarted manually.
To update a remote installation, you must have SSH installed and configured to support password-less access to the remote host. The host (and optional username) must be supplied on the command-line:
./tools/update --host=host2 --release-directory=/opt/continuent
When upgrading a cluster, you should upgrade slaves first and then update the master. You can avoid replication downtime by switching the master to an upgraded slave, upgrading the old master, and then switching back again.