After the Connections 8 CR1 had been released, we upgraded our production environment.
Upgrading the “Blue” stack (DB2/Websphere/IHS) has been done in-place.
- DB2 11.1 to 11.5.6
- WebSphere / IHS 22.214.171.124 to 126.96.36.199
- Connections 7.2212 to Connections 8 CR1
For the Component Pack upgrade we decided to do a side-by-side migration. An in-place update of the existing Component Pack servers would have meant to :
- update Kubernetes from 1.21. to 1.25/1.26
- migrate Kubernetes from Docker to containerd during the update process
- uninstall certain components from Component Pack 7
- install Component Pack 8
It’s doable, but I wanted to keep the downtime as low as possible, so I could squeeze it into a lunch break. Which means less than 1h.
And CP7 runs fine with Connections 8. At least for some time.
For the side-by-side migration I set up a small kubernetes 1.26 cluster (1 master- and 2 workernodes) with an HA-Proxy for the Loadbalancing in front of them. Installing the component pack in the new environment, while the old environment is still up, requires some changes in the first steps.
a) I’ve created new directories for the volumes in a different location on the nfs server.
b) for the bootstrap I had to set “skip_configure_redis: true”
After that, the installation followed the official installguide. I just skipped all the steps to configure connections. Just do not copy the version numbers from the documentation. They are outdated. Always use the helm command to get the latest version.
In the end I had a CP7 and CP8 at the same time. That allowed me to do a test run for the mongo and opensearch migration, and prepare the necessary commands for migration day.
On Migration day I had to do the configuration changes in gatekeeper – redis and elasticsearch url – and import the opensearch certificates for metrics and quickresults. After the restart everything worked fine.