KubeCon 2019 - Barcelona Talks
https://www.youtube.com/watch?v=4jEASYCaVDo&list=PLj6h78yzYM2PpmMAnvpvsnR4c27wJePh3
Kubernetes is here to stay
One thing is for certain kube is turning into the primary platform for platforms.
The number of people and companies interested in and already using kubernetes, is quickly on the rise. Most people I spoke with are already using it although there were 1 or 2 were at the conference to assess the uptake and return (hopefully) with good impressions for their companies.
From a vendors perspective, the amount of competition in the niche has increased with a focus being on the, CI/CD, Security, APM, Tracing & Monitoring as well as General hosting and platforms.
Application Development
There is allot on offer in this niche, with open-tracing and open-census opting to merge (as with SMI - Service Mesh Interface a standardised interface being opted into by all of the players istio, gloo, linkerd etc) there is allot more cooperative work taking place which implies less of a liklihood to see browser-wars style events happening.
With the rise of the Service Mesh allot of energy is focused on bringing principles of observability into play as more and more projects move in this direction the need for gaining insight into what your applications bottlnecks are and its performance profile at scale is critical. There were 1 or 2 vendors there but the larger players like https://www.honeycomb.io/ were conspicuously absent.
There are a number of interesting perspectives (like telepresence and remote debugging and build capabilities) that seem an interesting idea but the more you think about it the less appealing they are (really you want to open a remote session to your production server and debug live pods? like for real?)
Some interesting new approaches like https://ballerina.io/ for writing microservices and https://github.com/storyscript/storyscript in the same abstracted space
In the provisioning space Terraform 0.12.0 was released (https://www.terraform.io/downloads.html) and https://www.pulumi.com/ were present; pulumi offers a more software like experience of provisioning infra and is the one to beat imho
Developing for kubernetes
Operators and gRPC are the name of the game now! Most systems in the future will be deployed with and operated by software on the k8s platform itself.
https://github.com/kubernetes-sigs/kubebuilder and https://github.com/operator-framework appear to be the best players in the space with kubebuilder being the lower level system
Systems like https://github.com/kubernetes-sigs/kind making writing e2e tests for kubernetes as a platform very simple and allot is on the go in this space
CI/CD
Quite allot going on in this space, with a good presentation on Jenkins-X and Argo-CD; a focus on the CI steps moves solidly towards “docker based build steps” which has been pioneered by systems such as gitlab and concourse we now see the rise of https://tekton.dev which appears to be a standardised approach to using k8s as your build tool.
Argo workflows for DAG/concurrent/serial based pipelines competes with older systems such as https://github.com/spotify/luigi from Spotify and https://airflow.apache.org/
Jenkins-x appears to be the system to beat, providing a large and qutie complete if opinionated way of doing kubernetes based CI/CD.
SaaS offering in the space are looking quite interesting too with vendors like https://harness.io/ offering ML based validation to allow another level of sanity checking on deployment.
Monitoring and Logging
APMs are critical to understanding your application at scale; like it or not vendors such as https://sysdig.com/ and https://www.instana.com/ will play a role in most production and certainly all non-prod environments as developers must know the individual profiles of each of their applications.
Grafana and Loki are major players with Grafana being the defacto visualisation tool, they have also released https://grafana.com/loki which is a move towards offering alternatives to other competitors in this space. Far simpler than the ELK stack loki seems to have the capability of becoming a major player in the space as it grows alongisde prometheus as being the metrics provider of choice.
Speaking of prometheus and grafana, https://github.com/cortexproject/cortex is grafanas contribution to prometheus at scale and appears to be more production ready than https://github.com/improbable-eng/thanos