Upgrading an Existing Oracle NoSQL Database Deployment

This article describes how to upgrade the software of your Oracle NoSQL Database deployment.

Upgrading a data store from an existing release to a new release can be accomplished one Storage Node at a time because Storage Nodes running a mix of two releases are permitted to run simultaneously in the same data store. This allows you to strategically upgrade Storage Nodes in the most efficient manner.

Installing new software requires that you restart each replication node ( referred to as node). Upgrading a data store while the store remains online and available to clients is called rolling upgrade. A rolling upgrade is useful in most cases, since downtime is undesirable in any system. Depending on your data store’s configuration, it is sometimes possible to perform a rolling upgrade. Upgrading a system after shutting down the data store and making it offline is called offline upgrade.

Even if your data store can support a rolling upgrade, you may sometimes want to perform an offline upgrade, which involves these steps:
  1. Shutting down all nodes.
  2. Installing new software on each Storage Node.
  3. Restarting each node.
While an offline upgrade is a simpler process in some ways, your data store is unavailable for the duration of the upgrade.

You can perform a rolling upgrade if the data store's replication factor is greater than two. With a replication factor greater than two, shards can maintain their majorities and continue reading and writing data on behalf of clients. Meanwhile, you can restart and upgrade software on each Storage Node, one at a time.

If the replication factor is 2, the shards cannot maintain their majorities through a node restart. This means that each shard will be unavailable while you shutdown a node in the shard, upgrade the software, and restart the node. This is only true if there are no arbiter nodes configured. An arbiter node is capable of supporting write availability in one of these scenarios:
  • When the primary replication factor is two and a single replication node becomes unavailable.
  • When only two replication nodes are available and they are unable to communicate to determine which one of them is the master.