nbogojevic
About don't upgrade, recreate strategy
I saw your presentation on kubecon it was really good and inspiring.
You advise don't upgrade maybe recreate strategy.
Let's say I have 3 brokers and 3 of them haven't persistence volumes and also they are not stateful set. As an event sourcing architecture, I don't have to miss old data. After following your upgrade strategy, I re-create an updated broker.
Does it sync old data with old versions broker? If not what do you suggest for migrating strategy?
Hello @mcdemir Our strategy was not to migrate old data. Basically we create new cluster and feed both of them with same data. At a single point in time, we pause event-flow and switch services to new cluster and then tear down the old one down.
This worked for some of our workloads, but we also have others for which we needed to put in place upgrade in place due to need to persist data for long time. We are currently experimenting with strimzi project for such scenarios.
