Success stories rarely attract as much attention as failures, because failures are where the real learning happens. Here are some incidents of my career progression told from a not-so-successful perspective.
Or let’s say this is my CV that hiring managers should not see.
slido.com
In my second month, I managed to delete all production Kafka topics. We identified it as a design flaw and built a more robust topic management system. So… you’re welcome. And somehow, I still survived the probation period.
kiwi.com
We were using preemptible instances on GKE (Google Kubernetes Engine), even on some production clusters. To optimize preemption handling, I deployed a third-party controller to do the job — but scaled it to zero to avoid introducing production changes outside of core hours. The next morning, I woke up every ops person after it wiped every single node from the cluster at once.
I sensed an opportunity to create an outage on our public production gateways after developers requested gRPC support. I provided my deep expertise in Istio and caused a protocol conflict. But I’m not sure why I’m blaming myself — my colleague deployed it.
After “proper” testing in staging, it was time to migrate our largest service to the service mesh. With my salary, I should probably know the difference between HTTP and TCP. Well… this was a big one.
One of our core teams was losing metrics due to UDP instability. I improved system reliability by switching metrics collection to a Unix socket on the Kubernetes nodes. It worked perfectly — but increased our metrics costs by 300%. What is the opposite of FinOps?
I gained deep expertise in kubelet after my colleague and I crashed a production bare-metal Kubernetes cluster three times in a row due to the same bug related to kubelet not playing well with the underlying cgroup driver. We were a dream team.
I’d also like to brag about my knowledge of Calico, which I gained after my actions completely nuked pod-to-pod networking on baremetal Kubernetes. I simply learn by setting production on fire.
I implemented a feature in one of our Kubernetes operators that caused finalizers to get stuck forever after resource deletion. Nobody could find the bug, which officially made me an amazing easter egg creator.
I had a simple task: show off my coding skills and implement Lambda functions to automate support for CPU overheating for one of our bare-metal providers. I still don’t know how they didn’t fire me during those THREE months.
vnet.sk
We developed a portal for sending marketing SMS messages. Occasionally, with my help, a few messages were sent on the wrong day due to my not-yet-perfect scheduling skills. I think those yelling customers helped shape my phlegmatic personality a bit.
We maintained all internal systems, and, unfortunately, everything relied on one giant monolithic database instance. Whenever the systems were down, everyone knew I was busy learning PostgreSQL in production.
And that’s all for now. I hope this summary won’t destroy my chances on the job market. Wish me luck — and stay tuned for more of my “achievements.”
By the way, I am still employed. Somehow…