The Apache community developers have voted to stop developing the Apache Mesos cluster resource management platform and move existing developments to the Apache Attic legacy project repository. Enthusiasts interested in the further development of Mesos are invited to continue development by creating a fork of the project's git repository.
As the reason for the failure of the project, one of the key Mesos developers mentions the inability to compete with the Kubernetes platform, which was created later, generalized the experience of its predecessors and was created by Google, which has rich experience in creating large clusters. Unlike Kubernetes, the Mesos project was created by graduate students with little experience with clusters, who were then hired by Twitter. The project evolved through trial and error, and looking back, the developers admit that many things should have been done differently. Mesos is far from the "batteries included" principle, ie. does not provide a single set of components (for example, schedulers and services are developed in separate projects), which led to a strong fragmentation of the community, complicated deployment processes, and made the project unfriendly to beginners. Mistrust of users also gave rise to the actions of the startup Mesosphere, which is trying to develop commercial solutions based on Mesos.
Recall that Mesos was originally developed by Twitter and transferred to the Apache Foundation in 2010. Mesos-based clusters have been deployed at companies such as Netflix, Samsung, Twitter, IBM, PayPal, and Yelp. Mesos combines the functions of a cluster resource sharing system, container orchestration and a distributed kernel to organize the execution of jobs on a group of nodes. Mesos allows you to work with a cluster as a single set of resources, abstracting processors, GPUs, memory, storage systems and other computing resources on physical servers and virtual machines. When running distributed applications and frameworks, Mesos takes the job of dynamically allocating and isolating available resources.
Source: opennet.ru