{"id":55712,"date":"2020-01-27T00:00:00","date_gmt":"2020-01-26T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/tri-urovnya-avtomasshtabirovaniya-v-kubernetes-kak-ih-effektivno-ispolzovat"},"modified":"2020-02-18T14:03:51","modified_gmt":"2020-02-18T11:03:51","slug":"tri-urovnya-avtomasshtabirovaniya-v-kubernetes-kak-ih-effektivno-ispolzovat","status":"publish","type":"post","link":"https:\/\/prohoster.info\/da\/blog\/administrirovanie\/tri-urovnya-avtomasshtabirovaniya-v-kubernetes-kak-ih-effektivno-ispolzovat","title":{"rendered":"Tre niveauer af automatisk skalering i Kubernetes: hvordan man effektivt bruger dem","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"Tre
\nFor a complete understanding of Kubernetes, one must know various ways to scale cluster resources: according to the words of the system developers<\/a><\/noindex>, this is one of the main tasks of Kubernetes. We have prepared a high-level overview of horizontal and vertical autoscaling mechanisms and resizing clusters, as well as recommendations on how to use them effectively.<\/p>\n

The article Kubernetes Autoscaling 101: Cluster Autoscaler, Horizontal Autoscaler, and Vertical Pod Autoscaler<\/a><\/noindex> was translated by the team that implemented autoscaling in Kubernetes aaS fra Mail.ru<\/a><\/noindex>.
\n<\/a><\/noindex><\/p>\n

Why it is important to think about scaling <\/h2>\n

\nKubernetes<\/a><\/noindex> \u2014 a tool for managing resources and orchestration. Of course, it is nice to dabble with cool features of deployment, monitoring, and pod management (a pod is a group of containers launched in response to a request). <\/p>\n

However, one should also consider the following questions:<\/p>\n

    \n
  1. How to scale modules and applications?\n<\/li>\n
  2. How to keep containers running and efficient?\n<\/li>\n
  3. How to respond to constant changes in code and user workloads?\n<\/li>\n<\/ol>\n

    \nConfiguring Kubernetes clusters for resource and performance balancing can be a complex task; it requires expert knowledge of how Kubernetes works internally. The workload on your application or services can fluctuate throughout the day or even within the hour, so it\u2019s best to view balancing as an ongoing process.<\/p>\n

    Kubernetes Autoscaling Levels<\/h2>\n

    \nEffective autoscaling requires coordination between two levels: <\/p>\n

      \n
    1. Niveau af pods, der inkluderer vandret (Horizontal Pod Autoscaler, HPA) og vertikal autoskalering (Vertical Pod Autoscaler, VPA). Dette er skalering af de eksisterende ressourcer til dine containere.\n<\/li>\n
    2. The cluster level, managed by the cluster autoscaling system (Cluster Autoscaler, CA), which increases or decreases the number of nodes within the cluster.\n<\/li>\n<\/ol>\n

      <\/p>\n

      Horizontal Autoscaler (HPA)<\/h2>\n

      \nSom navnet antyder, skalerer HPA antallet af repliker af pods. De fleste devops-brugere anvender CPU- og hukommelsessbelastning som triggere for at \u00e6ndre antallet af repliker. Dog kan systemet skaleres baseret p\u00e5 custom metrics<\/a><\/noindex>, their combination<\/a><\/noindex> eller endda external metrics<\/a><\/noindex>.<\/p>\n

      High-level scheme of how HPA works:<\/p>\n

        \n
      1. HPA overv\u00e5ger kontinuerligt de metrikv\u00e6rdier, der er angivet ved installationen, med en standardinterval p\u00e5 30 sekunder.\n<\/li>\n
      2. HPA fors\u00f8ger at \u00f8ge antallet af moduler, hvis den angivne gr\u00e6nse er n\u00e5et.\n<\/li>\n
      3. HPA opdaterer antallet af repliker inden for deployments-\/replikationskontrolleren.\n<\/li>\n
      4. Deployment-\/replikationskontrolleren deployer derefter alle n\u00f8dvendige ekstra moduler.\n<\/li>\n<\/ol>\n

        \n\"Tre
        \nHPA starter udrulningsprocessen for moduler, n\u00e5r metrikgr\u00e6nsen er n\u00e5et.<\/i><\/p>\n

        Ved brug af HPA, overvej f\u00f8lgende:<\/p>\n