Upgrade Guide - Scylla Enterprise 2019.x.y to 2019.x.z for Red Hat Enterprise 7 or CentOS 7¶
This document is a step by step procedure for upgrading from Scylla Enterprise 2019.x.y to 2019.x.z.
This guide covers upgrading Scylla Enterprise from the following versions: 2019.x.y to 2019.x.z, on the following platforms:
Red Hat Enterprise Linux, version 7 and later
CentOS, version 7 and later
No longer provide packages for Fedora
Execute the following commands one node at the time, moving to the next node only after the upgrade procedure completed successfully.
If you are using CDC and upgrading Scylla 4.3 to 4.4, please review the API updates in querying CDC streams and CDC stream generations. In particular, you should update applications that use CDC according to CDC Upgrade notes before upgrading the cluster to 4.4.
If you are using CDC and upgrading from pre 4.3 version to 4.3, note the upgrading from experimental CDC.
If any of your instances are running Scylla Enterprise 2019.1.6 or earlier, and one of your Scylla nodes is up for more than a year, you might have been exposed to issue #6063. One way to check this is by comparing Generation No (from nodetool gossipinfo output) with the current time in Epoch format (date +%s), and check if the difference is higher than one year (31536000 seconds). See scylla-check-gossiper-generation for a script to do just that.
If this is the case, do not initiate the upgrade process before consulting with Scylla Support for further instructions.
Scylla Enterprise 2019.1.6 added a new configuration to restrict the memory usage cartesian product IN queries. If you are using IN in SELECT operations and hitting a “cartesian product size … is greater than maximum” error, you can either update the query (recommended) or bypass the warning temporarily by adding the following parameters to scylla.yaml:
The higher the values, the more likely you will hit an out of memory issue.
Scylla Enterprise 2019.1.8 added a new configuration to restrict the memory usage of reverse queries. If you are using reverse queries and hitting an error “Aborting reverse partition read because partition … is larger than the maximum safe size of … for reversible partitions” see the reverse queries FAQ section.
A Scylla upgrade is a rolling procedure that does not require a full cluster shutdown. For each of the nodes in the cluster, serially (i.e. one at a time), you will:
Drain node and backup the data
Check your current release
Backup configuration file
Download and install new Scylla packages
Validate that the upgrade was successful
Apply the following procedure serially on each node. Do not move to the next node before validating the node is up and running with the new version.
During the rolling upgrade, it is highly recommended:
Not to use new 2019.x.z features
Not to run administration functions, like repairs, refresh, rebuild or add or remove nodes
Not to apply schema changes
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 this name under
/var/lib/scylla to a backup device.
When the upgrade is complete (all nodes), the snapshot should be removed by
nodetool clearsnapshot -t <snapshot>, or you risk running out of space.
sudo cp -a /etc/scylla/scylla.yaml /etc/scylla/scylla.yaml.backup-2019.x.z
sudo systemctl stop scylla-server
Before upgrading, check what version you are running now using
rpm -qa | grep scylla-server. You should use the same version in case you want to rollback the upgrade. If you are not running a 2019.x.y version, stop right here! This guide only covers 2019.x.y to 2019.x.z upgrades.
Update the Scylla Enterprise RPM repo to 2019.x
sudo yum update scylla\* -y
sudo systemctl start scylla-server
Check cluster status with
nodetool statusand make sure all nodes, including the one you just upgraded, are in UN status.
curl -X GET "http://localhost:10000/storage_service/scylla_release_version"to check scylla version.
journalctl _COMM=scyllato check 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 is successful, move to the next node in the cluster.
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 Scylla Enterprise release 2019.x.z to 2019.x.y. Apply this procedure if an upgrade from 2019.x.y to 2019.x.z failed before completing on all nodes. Use this procedure only for nodes you upgraded to 2019.x.z
Scylla rollback is a rolling procedure that does not require a full cluster shutdown. For each of the nodes rollback to 2019.x.y, you will:
Drain the node and stop Scylla
Downgrade to previous release
Restore the configuration file
Validate the rollback success
nodetool drain sudo systemctl stop scylla-server
sudo yum downgrade scylla\*-2019.x.y -y
sudo rm -rf /etc/scylla/scylla.yaml sudo cp -a /etc/scylla/scylla.yaml.backup-2019.x.z /etc/scylla/scylla.yaml
sudo systemctl start scylla-server
Check the upgrade instruction above for validation. Once you are sure the node rollback is successful, move to the next node in the cluster.