Downgrades of Riak KV are tested and supported for two feature release versions, with the general procedure being similar to that of a rolling upgrade.

Depending on the versions involved in the downgrade, there are additional steps to be performed before, during, and after the upgrade on on each node. These steps are related to changes or new features that are not present in the downgraded version.


For every node in the cluster:

  1. Stop Riak KV.
  2. Back up Riak’s etc and data directories.
  3. Downgrade the Riak KV.
  4. Reconfigure Solr cores.
  5. Start Riak KV.
  6. Monitor the reindex of the data.
  7. Finalize process and restart Riak KV.


  • Riak control should be disabled throughout the rolling downgrade process.
  • Configuration Files must be replaced with those of the version being downgraded to.

Components That Complicate Downgrades

Feature automatic required Notes
Active Anti-Entropy file format changes Can be opted out using a capability

When Downgrading is No Longer an Option

If you enabled LZ4 compression in LevelDB and/or enabled global expiration in LevelDB when you installed KV 3.0.8, you cannot downgrade.

General Process

While the cluster contains mixed version members, if you have not set the cluster to use the legacy AAE tree format, you will see the bad_version error emitted to the log any time nodes with differing versions attempt to exchange AAE data (including AAE fullsync).

This is benign and similar to the not_built and already_locked errors which can be seen during normal AAE operation. These events will stop once the downgrade is complete.

Stop Riak KV

1. Stop Riak KV:

riak stop

2. Back up your Riak KV /etc and /data directories:

sudo tar -czf riak_backup.tar.gz /var/lib/riak /etc/riak

3. Downgrade Riak KV:

sudo rpm -Uvh »riak_package_name«.rpm
sudo dpkg -i »riak_package_name«.deb

Start the node

6. Start Riak KV:

riak start

Monitor the reindex of the data

7. Monitor the build and exchange progress using the riak-admin aae-status command.

The All column shows how long it has been since a partition exchanged with all of its sibling replicas. Consult the riak-admin aae-status documentation for more information about the AAE status output.

Once riak-admin aae-status shows a value in the All column, the node will have successfully rebuilt all of the indexed data.

8. If you raised the concurrency AAE currency settings in riak.conf during Step 5, stop the node and remove the increased AAE thresholds.

13. Verify that transfers have completed:

riak-admin transfers