Docker Management Design Patterns by Deepak Vohra

Author:Deepak Vohra
Language: eng
Format: epub
Publisher: Apress, Berkeley, CA
Published: 2017-11-26T16:00:00+00:00

Single point of failure —If all services are running on one node and that node has a problem, it results in downtime. Increasing redundancy across nodes obviates that problem.

The Solution

To overcome the issues discussed in the preceding section, service task scheduling in a Docker Swarm is based on a built-in scheduling policy. Docker Swarm mode uses the spread scheduling strategy to rank nodes for placement of a service task (replica). Node ranking is computed for scheduling of each task and a task is scheduled on the node with the highest computed ranking. The spread scheduling strategy computes node rank based on the node's available CPU, RAM, and the number of containers already running on the node. The spread strategy optimizes for the node with the least number of containers. Load sharing is the objective of the spread strategy and results in tasks (containers) spread thinly and evenly over several machines in the Swarm. The expected outcome of the spread strategy is that if a single node or a small subset of nodes go down or become available, only a few tasks are lost and a majority of tasks in the Swarm continue to be available.

Note

Because a container consumes resources during all states, including when it is exited, the spread strategy does not take into consideration the state of a container. It is recommended that a user remove stopped containers, because a node that would otherwise be eligible and suitable for scheduling a new task becomes unsuitable if it has several stopped containers.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.