Autoscaling nodes of a stateful application based on role-based autoscaling policies

A non-transitory machine readable medium comprising instructions that upon execution cause a controller host of a distributed processing environment to:

• receive information regarding a respective role of a plurality of roles of a stateful application performed by each respective application container of a plurality of application containers in a virtual cluster of the distributed processing environment, wherein a second role of the plurality of roles is dependent upon a first role of the plurality of roles, and wherein the plurality of application containers in the virtual cluster are distributed across multiple worker hosts of a plurality of worker hosts;
• maintain a set of role-based autoscaling policies defining conditions under which the plurality of roles are to be scaled, wherein a role-based autoscaling policy of the set of role-based autoscaling policies for the first role specifies a first condition that triggers scaling out of the first role by a first scaling factor and scaling out of the second role that is dependent upon the first role by a second scaling factor in tandem;
• receive, at the controller host from worker agents in the multiple worker hosts, load information collected from container agents in first application containers of the plurality of application containers in the virtual cluster for the first role;
• in response to the load information from the container agents in the first application containers for the first role meeting the first condition, direct, by the controller host, worker agents in the plurality of worker hosts to increase a number of the first application containers in the virtual cluster that perform the first role by the first scaling factor, and increase a number of second application containers in the virtual cluster that perform the second role that is dependent upon the first role by the second scaling factor, wherein the increased number of the first application containers includes a new first application container that includes a respective container agent to collect load information of the new first application container, and the increased number of the second application containers includes a new second application container that includes a respective container agent to collect load information of the new second application container;
• select, by the controller host based on loads of the plurality of worker hosts, at least one worker host of the plurality of worker hosts on which to deploy the new first application container and the new second application container, wherein the directing comprises sending a request from the controller host to at least one worker agent in the selected at least one worker host to deploy the new first application container and the new second application container;
• and
• cause deployment, based on the request, of the new first application container and the new second application container on the selected at least one worker host.

A system comprising:

• a controller host comprising:
• a processor;
• and
• a non-transitory computer-readable medium storing instructions executable on the processor to:
• for a plurality of application containers in a first virtual cluster of a distributed processing environment on which a stateful application is or will be deployed, receive information regarding a respective role of a plurality of roles of the stateful application performed by each respective application container of the plurality of application containers, wherein a second role of the plurality of roles is dependent upon a first role of the plurality of roles, wherein the distributed processing environment comprises a plurality of worker hosts containing multiple application containers for respective virtual clusters including the first virtual cluster, and wherein the plurality of application containers in the first virtual cluster are distributed across multiple worker hosts of the plurality of worker hosts;
• maintain a set of role-based autoscaling policies defining conditions under which the plurality of roles are to be scaled, wherein a role-based autoscaling policy of the set of role-based autoscaling policies for the first role specifies a first condition that triggers scaling out of the first role by a first scaling factor and scaling out of the second role by a second scaling factor in tandem;
• receive, at the controller host from worker agents in the multiple worker hosts, load information collected from container agents in first application containers of the plurality of application containers in the first virtual cluster for the first role;
• in response to the load information from the container agents in the first application containers for the first role meeting the first condition, direct, by the controller host, worker agents in the plurality of worker hosts to increase a number of the first application containers in the first virtual cluster that perform the first role by the first scaling factor, and increase a number of second application containers in the first virtual cluster that perform the second role by the second scaling factor, wherein the increased number of the first application containers includes a new first application container that includes a respective container agent to collect load information of the new first application container, and the increased number of the second application containers includes a new second application container that includes a respective container agent to collect load information of the new second application container;
• select, based on loads of the plurality of worker hosts, at least one worker host of the plurality of worker hosts on which to deploy the new first application container and the new second application container, wherein the directing comprises sending a request from the controller host to at least one worker agent in the selected at least one worker host to deploy the new first application container and the new second application container;
• and
• cause deployment, based on the request, of the new first application container and the new second application container on the selected at least one worker host.