Work

Leading a migration of our Kafka clusters from Confluent Cloud to self-managed on bare metal. The cost case was clear; the operational case is taking longer to prove. We're in week six, running both environments in parallel. Rebalancing behavior on the self-managed side has been the main source of surprises.

Also working on improving our incident review process. We do reviews, but they're inconsistent — some are thorough, some are a paragraph. Trying to find a lightweight template that gets better output without adding an hour to every post-mortem.

Reading

Designing Data-Intensive Applications for the second time. The first read was fast. The second is slow, with notes. The chapter on replication is the one I find myself explaining to colleagues most often, so understanding it precisely seems worth the time.

Also working through the Kafka documentation properly — the sections on exactly-once semantics and transaction handling specifically.

Running

Training for a 25km trail race in the Harz in September. Currently at 45km per week, building toward 60. The Grunewald loops are getting familiar enough that I'm starting to get bored of them, which probably means I should enter something in the interim.

Home lab

Added a fourth node to my ThinkCentre cluster. Running a small Kafka setup there to test migration configs before I propose them for production. Also experimenting with Loki for log aggregation — the resource consumption is more reasonable than I expected.


Home · About