A Squirro upgrade requires a downtime. This is typically quite short, but will still affect user experience for a moment. As a result, upgrades to production systems should only be run in a maintenance window.
Storage vs. Cluster Nodes
Squirro distinguishes between storage and cluster nodes. These can be installed on the same server or split up. Both can also be horizontally scaled and have multiple instances. For full details see How Squirro Scales.
Independent of the setup, in the instructions below always execute all the storage node updates first, and only then do the cluster node updates. This applies independent of whether cluster and storage node are on the same system, whether they are split into one server each, or if they have multiple instances each.
If you have used the offline installation, or have an internal mirror of the Squirro yum repository, please make sure this is updated prior to starting the Squirro update.
Run these instructions on every single server that has storage nodes installed before moving on to the cluster nodes.
Please note that user-provided custom assets which include a requirements.txt file (e.g., a custom data loader plugin) might need to re-uploaded after the upgrade if the packages they depend on (if any) do not work on Python 3.8.
# Clear all cached metadata
yum clean all
# Update Java JDK
yum update java-1.8.0-openjdk
# Update cluster node
yum update squirro-cluster-node-users
yum update squirro-cluster-node
mv /opt/squirro/virtualenv3 /opt/squirro/virtualenv3.old
# /opt/squirro/virtualenv3.old is not needed for Squirro platform operation
# Make sure that all the Squirro packages have been updated
yum update squirro-*
At this point you will have to check if any *.rpmnew files have been created in /etc/squirro/ or /etc/nginx/conf.d/. See Configuration Conflicts for how to handle those conflicts.
For example, the /etc/squirro/common.ini would need resolution if you are upgrading from a version that does not have the new pdfconversion service.
And validate if everything is coming up (this may have to be run a few times until everything stabilizes):
The yum autoremove command can cause the Squirro systemd services not to start during a reboot. We recommend that you issue the following commands to fix this issue and reboot the system to ensure it works. This only is an issue with upgrades from 3.4 and earlier.