Project Lightning Talk: Five Minutes by CloudNativePG River - Gabriele Quaresima, Contributor

In this talk, Gabriel Quarisima, a staff engineer and contributor to the CloudNativePG project, explores how PostgreSQL workloads can be effectively managed on Kubernetes. He explains that CloudNativePG is an open-source operator designed to provide Kubernetes clusters with PostgreSQL capabilities, and it is currently a CNCF sandbox project. The speaker elaborates on the architecture of PostgreSQL clusters, which consists of a primary instance and multiple replicas, emphasizing a shared-nothing architecture spread across different data centers. Quarisima describes various features including high availability, declarative management, backup and recovery systems, and integration with Kubernetes APIs. He encourages participation in the CloudNativePG community and shares resources for further engagement.

Hello everyone, welcome. So I am Gabriel Quarisima and I'm a staff engineer inb and also a clonetpg contributor. And let's spend the next five minutes by the cloud nativep river. So let's start with the answer. Yes, you can run data on kubernetes. How many of you run that on kubernetes? Yeah, not a few. So nice. Do you run a cloud netPG or something else? Some some of

you cloud netpg, some of you other thing. Okay. Very very nice. So what is clown netpg? So it is uh an open-source operator uh designed to manage posgrql workloads on any supported kubernetes cluster. is also a CNCF sandbox project applied for also for incubation during the CubeCon uh America the last one. It is an active community of people uh but is also the umbrella of several repositories

you can contribute to. So in a nutshell uh CMPG define a a new Kubernetes CRD named cluster. We need another one another meaning for cluster. So uh we represent a posgrsql cluster uh made up of a single primary and an arbitrary number of instances of replicas. So you can see on the right the cluster CD with a fancy name and also three instances that is yeah the

pretty common way to deploy it and also yes the storage the dimension of the storage. So this is the suggested architecture. So basically we have uh a shared nothing architecture with uh three data centers. So basically three worker nodes in three different availability zone. Hopefully we have this distributed uh view topology and we have uh any Kubernetes worker node as its own instance. There will be a

primary and the replicas we define the sandbox we define and uh also each one with the with its own local uh persistent volumes and also uh with services we can connect each other. We have uh read only read write services for primary read only uh for replicas and primary read um for replicas sorry and read only read for primary and replicas services and yeah we have a

lot of features these are them most uh common so we have a direct integration with uh Kubernetes API it's our source of truth um we have an high availability capability scale up and down capabilities is uh definition of um an arbitrary number of replicas as we said um also uh we have the planets which uh a declarative management of posgressql configuration you you can add in the

spec of the CRD a lot of parameters to configure your own posgress and also uh we have backup and recovery system via volume snapshot CRD but also object store rely on Barman cloud currently and we have a cubectl plug-in for class management able to uh retrieve logs um performance with um a lot of stuff and uh we have also a clusteration refencing if we need to debug

the posgsql or if you need to switch off the kubernetes cluster to save money on the weekend and uh also we have a lot of most recent features like uh declarative databases and logical replication via declarative uh publication and subscription. So we try to pull everything in a declarative way and also we have the CMPGI interface. So we can you can plug your own backup system for

example uh you can write it you can plug it uh to the operator and we support the major posgressql upgrade and also we introduce the declarative foreign data management quorum basel fover and also the declarative loading of posgress extension via the OCI container. So yeah, please uh join the community. Uh we are on any kind of social and also on CNCF uh website, on GitHub, on Slack.

So there is uh a list of resources and yeah um meet us in the uh at the kiosk. Uh so on uh Wednesday uh you can reach us. Um we are there um contributor maintainers we are there and also there are uh two talks one is in five minutes so you can start to run but I add it anyway the one uh the one for sovereignity data

and also there is another uh talk uh by Gabriel Bolini and Lauren Parody on uh Wednesday. So yeah thanks and that's it. >> Awesome. Thank you very much.