 Navarone, you're up. Thank you, Vibhul. Hi, folks. I'll go in just a bit faster here. So I talk on what you should not miss when deploying Python applications to Kubernetes. And I hope my audio is fine, and you are able to see the slides. I'll not waste time on introducing myself. I'm just a random person in the dev community. So first advice that I have when running Python applications on Kubernetes is never run containers as root. There are a lot of security issues that you may encounter when you run containers as root. People may do exploits. In Kubernetes, you can enforce the same using a context called pod security context. And then you can specify which user to run is. You can specify which group to run is. And then you can also say, what file system mask do you want to provide to the container? So any volume which is mounted to the container will be masked according to that. The next advice is always put resource requests and limits for your application. This is very important because if your application somewhere behaves weirdly and then you float your quota of resources, this may result in preemption, which you may not want. And that preemption may be very unpredictable. To do that, in Kubernetes, while defining a deployment, you can say that, hey, my workload takes only 512 Mb of memory at max. And 100 Mb is 0.1 of a CPU. CPU frictions are weird, but they are existent. You can also say that, hey, it is expected that my application would take 1.8 Mb of memory, but it can take like 512. But when it goes up to 512, just kill it. It will just kill the process. Next advice is, always perform life-checking checks using probes. So Kubernetes has two beautiful things called health checks, lightningless probes, readiness probes. You can think of them as similar to watchdogs that you can define. In Kubernetes, you can specify several kinds of probe types. So you have a TCP probe, which actually pings the port and sees whether any traffic goes to the port or not. You can also do an history picket request, which actually queries that path which you specify and checks whether the request return is 200. It checks the response code. If it's not 200, it fails the probe. And then you can specify some more tuning parameters, like how frequently the probe should work, and then how many probe failures should actually result in a failure that's called failure threshold. You can also say that, hey, my application takes a certain amount of time to boot up. Let's not do health checks before that. So you can specify an initial delay. So here is an example. Again, you have to put it in your deployment, manifest. The next advice is, always handle signals properly. Kubernetes expects your applications to actually handle all OS signals. So when you try to kill a Kubernetes workload, I'm talking in terms of workloads, because there are several kinds of workloads, and I want to abstract it out for the short session. When you want to handle signals, you have to handle two specific signals, signal and signal. You don't need to check for signal. It's just like, hey, you're gone. No more chances. You do have to check for signal when it is always advised that you trap that signal somehow and finish all the work that the app is doing. Kubernetes provides something called termination grace period seconds. What this means is it may look like really big chunk of word, but then what it means is it is the time difference between when the sick term is sent, when the process is asked to, hey, can you gracefully terminate, and sick kill, which is like, hey, I can't run you anymore. So if you have a very important application, you should actually trap the signal and then finish all your pending work that you have and save state to database. Sorry for the capital D. Probably that's the type on my side. It should not happen in Python. Beautiful code. Any questions? I'll take in the chat, and that's all on my presentation. You can also DM me on Twitter, the only number on. Thank you. Also a shameless plug. Tomorrow I have a workshop on Kubernetes in the second time slot, and it revolves around majorly this kind of topics. How do you resiliently deploy Python applications to Kubernetes? Thank you. Thank you, Vipul.