 Hello, and welcome. Next, I'm going to share how to achieve unified observability for cloud and edge with Flunbet. I'm Benjamin. I did the Conspiracy Observability and Edge Computing team. I'm also a member of Kuba Edge TSC. In this session, I'm going to give a brief introduction to a FlunT operator, and most importantly, how to use it and Flunbet to achieve unified cloud and edge observability. Actually, this project is an open source that is a Flunbet operator by Conspiracy in January 2019. We created this tool to address the problem of Flunbet configuration dynamic reloading without restarting the Flunbet pod. And it is donated to the Flun community in August last year. And in March this year, we finally add Flunbet support to this project. And we named the entire project to FlunT operator and released the FlunT operator 1.0. This is how FlunT operator looks like now. We have changed all the Flunbet CRDs to cluster-wide. And we also add names-based and cluster-wide CRDs for FlunD. And the user can use FlunT operator to manage and deploy Flunbet demo site, as well as FlunD stable site. If user want to process blogs from a specific name space, he can define a name space by FlunD CRD. And this CRD will allow user to only process logs from this name space and forward it to the final destinations. User can also use FlunT operator, can also deploy Flunbet demo site only, use Flunbet to collect logs and do some simple processing and forward to the final destination. A user can also deploy FlunD only. FlunD receive logs from a network through the HTTP forward or syslog plugin to some advanced processing and forward to the final destinations. And of course, a user can use Flunbet and FlunD together and deploy the Flunbet demo site together with FlunD stable site. And Flunbet will collect the cluster logs and forward it to FlunD. And FlunD do some advanced processing and forward to the final destinations. Since release 1.0, we have made some big improvement. For example, we have added the OpenSearch plugin for FlunD and Flunbet. We have added the Locate plugin. And in recent few months, we have noticed that Flunbet is starting to support a matrix plugins. So in release 1.5, we have added some matrix-related plugins. For example, they know the exposure matrix, the promises script matrix, the Flunbet matrix, and the promises remote write plugin together with OpenTelemetry plugin. Another important improvement in 1.5 is that we finally add the Flunbet custom plugin. While we add this, you may notice that Flunbet is having nearly 100 plugins now. So it's almost impossible or unreasonable to modify the code and release a new version of Flun operator to meet the requirement of a new plugin. So if user want to use an unsupported plugin, he can simply define a custom plugin section in existing input, filter, or output CRDs. So the configuration of a custom plugin is just the original Flunbet plugin configurations. So this way, user can use any Flunbet plugin without waiting for Flun operator release. This is a kind of a wicked type of the plugins. And the strong type plugins will be added continuously as we do before. We use Kubeh as our edge computing framework. So let's give a brief introduction to Kubeh. Kubeh is a CNCF incubating project for edge computing. Actually, there are several edge computing framework out there, for example, K3S. K3S actually will create a single entire cluster in the edge location. But Kubeh actually will set up several edge nodes in the edge location. And then it will connect this edge node to a cluster in the cloud through a security tunnel. So it's a bit different with K3S. Just like a regular Kubernetes node, edge node has a similar component like Kubeh called HD. HD is a lightweighted Kubeh. It trims the version Kubeh. Most importantly, HD also exports container matrices, just like Kubeh does. It's important to collect edge application matrices this week. Let's think of how to achieve unified cloud and edge observability with a blend bed. From this architecture, you can see that there is a cloud-side cluster deployed and also some edge nodes in the edge locations. For the cloud-side metrics, to collect cloud-side metrics, we use from-users agent to collect the metrics from notic spotter, Kubeh, Kubeh state metrics, et cetera. We also deploy a blend operator to manage and deploy the blend bed. Demonside to cloud and edge node. For the edge node, edge node actually have limited resources. And we shouldn't deploy many components to the edge node just to collect the observability data. So that's why we use a blend bed as an agent. We have changed the permissions agent to blend bed. And we have removed notic spotter and use blend bed. Notic spotter metrics plug-in instead. And we also use blend bed. Permissions script metrics plug-in to script metrics from the edge side Kubeh to get the container metrics. And of course, there is a blend bed. Demonside deployed to an edge node to collect edge logs for application and node. You can see the benefit here is we only need one component that is a blend bed for the edge node. You can use a blend bed to collect both logs and metrics for edge node and edge applications. It's a perfect fit for the limited resource edge node scenario. Let's see how we deploy a blend bed to the edge location. You can see here, we define a node affinity to deploy separate blend bed demonside to the edge node. And you can also see that we have a lot of few volumes of the edge node to the container to collect the node exporter metrics. To collect edge node container metrics, we define a permission script metrics plug-in to collect the edge node metrics. We deployed a node exporter blend bed metrics plug-in. And finally, we created a permission remote write plug-in for blend bed. And this plug-in will remove all the metrics to a cluster-side long-term storage. You can see here, this is how we achieve unified cloud and edge observability with a blend bed. It's very exciting to see blend bed is starting to support metrics and tracing because you will have a single tool to collect all the observability data. Blend bed operator, blend operator will support these new abilities whenever it's available. And you are welcome to participate in the blend operator product as well. So this is my sharing. Thank you for your time.