General upgrade procedure
The upgrade process for DataStax Enterprise is aimed at providing minimal downtime (ideally zero). With a few exceptions, the cluster will continue to work as though it were on the older version until all of the nodes in the cluster have been upgraded.
The order in which you upgrade the nodes does matter. Use the following guidelines when planning an upgrade:
- In multiple data center clusters upgrade all the nodes within one data center before moving on to another data center.
- Upgrade any Analytics nodes or data centers first, then Cassandra nodes or data centers, and finally Search nodes or data centers.
- If you have any Analytics nodes, upgrade the jobtracker in the data center node first.
- Upgrade the seed nodes within a data center first.
To perform an upgrade with zero downtime, we recommend performing the upgrade as a rolling restart. The procedure for the rolling upgrade is:
- Make a backup of the data by taking a snapshot of the node to be upgraded.
- Stop the node.
- Install the new product.
- Configure the new product.
- Start the node.
- Check the logs for warnings, errors and exceptions.
- Repeat these steps for each node in the cluster.
While the cluster is in a partially upgraded state, please observe the general limitations.