Was this page helpful?
Caution
You're viewing documentation for a previous version of ScyllaDB Open Source. Switch to the latest stable version.
This document is a step by step procedure for upgrading from ScyllaDB 5.1 to ScyllaDB 5.2, and rollback to version 5.1 if required.
This guide covers upgrading Scylla on Red Hat Enterprise Linux (RHEL) 7/8, CentOS 7/8, Debian 10 and Ubuntu 20.04. It also applies when using ScyllaDB official image on EC2, GCP, or Azure; the image is based on Ubuntu 20.04.
See OS Support by Platform and Version for information about supported versions.
A ScyllaDB upgrade is a rolling procedure which does not require full cluster shutdown. For each of the nodes in the cluster, serially (i.e. one node at a time), you will:
Check that the cluster’s schema is synchronized
Drain the node and backup the data
Backup the configuration file
Stop ScyllaDB
Download and install new ScyllaDB packages
(Optional) Enable consistent cluster management in the configuration file
Start ScyllaDB
Validate that the upgrade was successful
Apply the following procedure serially on each node. Do not move to the next node before validating that the node you upgraded is up and running the new version.
During the rolling upgrade, it is highly recommended:
Not to use the new 5.2 features
Not to run administration functions, like repairs, refresh, rebuild or add or remove nodes. See sctool for suspending ScyllaDB Manager (only available for ScyllaDB Enterprise) scheduled or running repairs.
Not to apply schema changes
If you enabled consistent cluster management in each node’s configuration file, then as soon as every node has been upgraded to the new version, the cluster will start a procedure which initializes the Raft algorithm for consistent cluster metadata management. You must then verify that this procedure successfully finishes.
Note
If you use the ScyllaDB Monitoring Stack, we recommend upgrading the Monitoring Stack to the latest version before upgrading ScyllaDB.
For ScyllaDB 5.2, you MUST upgrade the Monitoring Stack to version 4.3 or later.
Make sure that all nodes have the schema synchronized before upgrade. The upgrade procedure will fail if there is a schema disagreement between nodes.
nodetool describecluster
Before any major procedure, like an upgrade, it is recommended to backup all the data to an external device. In Scylla, backup is done using the nodetool snapshot command. For each node in the cluster, run the following command:
nodetool drain
nodetool snapshot
Take note of the directory name that nodetool gives you, and copy all the directories having that name under /var/lib/scylla to a backup device.
When the upgrade is completed on all nodes, remove the snapshot with the nodetool clearsnapshot -t <snapshot> command to prevent running out of space.
Back up the scylla.yaml configuration file and the ScyllaDB packages
in case you need to rollback the upgrade.
sudo cp -a /etc/scylla/scylla.yaml /etc/scylla/scylla.yaml.backup
sudo cp /etc/apt/sources.list.d/scylla.list ~/scylla.list-backup
sudo cp -a /etc/scylla/scylla.yaml /etc/scylla/scylla.yaml.backup
sudo cp /etc/yum.repos.d/scylla.repo ~/scylla.repo-backup
sudo service scylla-server stop
Before upgrading, check what version you are running now using dpkg -s scylla-server. You should use the same version as this version in case you want to rollback the upgrade. If you are not running a 5.1.x version, stop right here! This guide only covers 5.1.x to 5.2.y upgrades.
To upgrade ScyllaDB:
Install the new ScyllaDB version:
sudo apt-get clean all sudo apt-get update sudo apt-get dist-upgrade scylla
Answer ‘y’ to the first two questions.
Before upgrading, check what version you are running now using rpm -qa | grep scylla-server. You should use the same version as this version in case you want to rollback the upgrade. If you are not running a 5.1.x version, stop right here! This guide only covers 5.1.x to 5.2.y upgrades.
To upgrade ScyllaDB:
Update the ScyllaDB rpm repo to 5.2.
Install the new ScyllaDB version:
sudo yum clean all sudo yum update scylla\* -y
Note
If you are running a ScyllaDB official image (for EC2 AMI, GCP, or Azure), you need to:
Download and install the new ScyllaDB release for Ubuntu; see the Debian/Ubuntu tab above for instructions.
Update underlying OS packages.
See Upgrade ScyllaDB Image for details.
If you enable this option on every node, this will cause the Scylla cluster to enable Raft and use it to consistently manage cluster-wide metadata as soon as you finish upgrading every node to the new version. Check the Raft in ScyllaDB document to learn more.
In 5.2, Raft-based consistent cluster management is disabled by default. In 5.3 it will be enabled by default, but you’ll be able to disable it explicitly during upgrade if needed (assuming you haven’t previously enabled it on every node). In further versions the option will be removed and consistent cluster management will be enabled unconditionally.
The option can also be enabled after the cluster is upgraded to 5.2 (see Enabling Raft in existing cluster).
To enable the option, modify the scylla.yaml configuration file in /etc/scylla/ and add the following:
consistent_cluster_management: true
Note
Once you finish upgrading every node with consistent_cluster_management enabled, it won’t be possible to turn the option back off.
sudo service scylla-server start
Check cluster status with nodetool status and make sure all nodes, including the one you just upgraded, are in UN status.
Use curl -X GET "http://localhost:10000/storage_service/scylla_release_version" to check the ScyllaDB version. Validate that the version matches the one you upgraded to.
Check scylla-server log (by journalctl _COMM=scylla) and /var/log/syslog to validate there are no new errors in the log.
Check again after two minutes, to validate no new issues are introduced.
Once you are sure the node upgrade was successful, move to the next node in the cluster.
See Scylla Metrics Update - Scylla 5.1 to 5.2 for more information..
The following section applies only if you enabled the consistent_cluster_management option on every node when upgrading the cluster.
Enabling consistent_cluster_management on every node during upgrade will cause the Scylla cluster to start an additional internal procedure as soon as every node is upgraded to the new version.
The goal of this procedure is to initialize data structures used by the Raft algorithm to consistently manage cluster-wide metadata such as table schemas.
Assuming you performed the rolling upgrade procedure correctly, in particular ensuring that schema is synchronized on every step, and if there are no problems with cluster connectivity, then this follow-up internal procedure should take no longer than a few seconds to finish. However, the procedure requires full cluster availability. If an unlucky accident (e.g. a hardware problem) causes one of your nodes to fail before this procedure finishes, the procedure may get stuck. This may cause the cluster to end up in a state where schema change operations are unavailable.
Therefore, following the rolling upgrade, you must verify that this internal procedure has finished successfully by checking the logs of every Scylla node. If the procedure gets stuck, manual intervention is required.
Refer to the following document for instructions on how to verify that the procedure was successful and how to proceed if it gets stuck: Verifying that the internal Raft upgrade procedure finished successfully.
Note
Execute the following commands one node at the time, moving to the next node only after the rollback procedure completed successfully.
The following procedure describes a rollback from ScyllaDB 5.2.x to 5.1.y. Apply this procedure if an upgrade from 5.1 to 5.2 failed before completing on all nodes. Use this procedure only for nodes you upgraded to 5.2.
Warning
The rollback procedure can be applied only if some nodes have not been upgraded to 5.2 yet. As soon as the last node in the rolling upgrade procedure is started with 5.2, rollback becomes impossible. At that point, the only way to restore a cluster to 5.1 is by restoring it from backup.
ScyllaDB rollback is a rolling procedure which does not require full cluster shutdown. For each of the nodes you rollback to 5.1, serially (i.e. one node at a time), you will:
Drain the node and stop Scylla
Retrieve the old ScyllaDB packages
Restore the configuration file
Reload systemd configuration
Restart ScyllaDB
Validate the rollback success
Apply the following procedure serially on each node. Do not move to the next node before validating that the rollback was successful and the node is up and running the old version.
nodetool drain
nodetool snapshot
sudo service scylla-server stop
Restore the 5.1 packages backed up during the upgrade.
sudo cp ~/scylla.list-backup /etc/apt/sources.list.d/scylla.list sudo chown root.root /etc/apt/sources.list.d/scylla.list sudo chmod 644 /etc/apt/sources.list.d/scylla.list
Install:
sudo apt-get update sudo apt-get remove scylla\* -y sudo apt-get install scylla
Answer ‘y’ to the first two questions.
Restore the 5.1 packages backed up during the upgrade.
sudo cp ~/scylla.repo-backup /etc/yum.repos.d/scylla.repo sudo chown root.root /etc/yum.repos.d/scylla.repo sudo chmod 644 /etc/yum.repos.d/scylla.repo
Install:
sudo yum clean all sudo rm -rf /var/cache/yum sudo yum downgrade scylla-\*cqlsh -y sudo yum remove scylla-\*cqlsh -y sudo yum downgrade scylla\* -y sudo yum install scylla -y
sudo rm -rf /etc/scylla/scylla.yaml
sudo cp /etc/scylla/scylla.yaml-backup /etc/scylla/scylla.yaml
You must reload the unit file if the systemd unit file is changed.
sudo systemctl daemon-reload
sudo service scylla-server start
Check the upgrade instructions above for validation. Once you are sure the node rollback is successful, move to the next node in the cluster.
Was this page helpful?