Shulou(Shulou.com)06/02 Report--

Although most positions in IT departments will increase, career consultants say computer operators are expected to decrease. This role involves running Shell scripts, FTPing files, and handling exceptions, but because tools like cron can automate deletion operations, fewer and fewer people do the same.

This proved to be true. Too many people are running Kubernetes these days, and they spend too much time doing routing operations. These tasks need to be automated.

Google has long needed to hire systems administrators and reliability engineers for its growing cloud services, and simply cannot hire enough people. The problem is attracting enough people who are smart enough to become system administrators and move to Silicon Valley. So Google decided to build a common framework to manage clusters of virtual machines running any application. While Kubernetes is sometimes described as a full-stack system, it may actually require intervention and management, for example, to cope with scaling as workloads change.

Expertise on Kubernetes operations is summed up by experts as a "trio of stateful applications"-cluster management, load balancing, and replication. These capabilities are often associated with the need for high availability. According to experts, three jobs add significantly to the workload of managing applications. This is where people add operators or write automated tools to manage infrastructure.

The Kubernetes project describes five levels for operators, sometimes called maturity, which roughly correspond to the skill level of the operator. Level 1 enables basic installation of tools, including providing resources or negotiating with clusters to run the application. The second level provides seamless upgrades of patches and minor releases, while the third level includes backup and failover. At level four, you can handle alerts, log processing, and workload analysis, while at level five, you can solve some extended problems that Kuberenetes cannot solve, as well as advanced topics such as tuning configuration or scheduling. workload.

Once there is an automated tool for a particular tool (such as Redis, CouchDb or Kafka), someone shares it on github. Creating an automated operations tool may take a person/year, but it can save half a person/year for a cluster. Imagine deploying more than a thousand companies and a thousand clusters?

If you have an internal application that needs to focus on the trio (clustering, load balancing, and replication), it incurs ongoing maintenance costs. If you're having problems of Google's scale and don't want to hire a team of software reliability engineers to do work that can be programmed to some extent, writing your own automation tools is a good idea.

Automation tools consist of two parts: code for executing commands (written on Go, Ansible or simpler work, written on Helm) and custom resource definitions (CRDs). CRD maps operator code back to kubectl command. This makes it necessary to create a YAML file and pass it to Kubernetes via kubectl-apply when invoking operator functionality.

Again, someone on the team must know Ansible or Helm, and that code may need to be stored in version control and managed as a software development artifact. To make it more than a single point of failure, the team will want the second or third programmer to be proficient with Go or Ansible.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.