This documentation is for WSO2 Private PaaS 4.0.0. View documentation for the latest release.
Scale Down Workflow - WSO2 Private PaaS 4.0.0 - WSO2 Documentation
Skip to end of metadata
Go to start of metadata

 The workflow involved in scaling down WSO2 Private PaaS is illustrated in the diagram below:

The workflow involved in scaling down WSO2 Private PaaS is as follows:

  1. For the scale-down workflow based on the value of load average, the cartridge instance sends the health statistics and Load Balancer (LB) sends the count of the requests in-flight using the thrift protocol to Complex Event Processor (CEP).
  2. The CEP will aggregate the health statistics continuously and will publish the summarized health statistics to Message broker (MB).
  3. Auto-scaler (AS), which will be subscribed to the Summarized Health Stats topic on MB will obtain the values and using the Drools engine, and it will check the auto-scaling and deployment policies for the functionality required based on the count and the threshold values defined. 
  4. When it is required to scale down, AS will publish a graceful shutdown event to the Instance Notifier topic on MB and the Cartridge instance, which will be subscribed to MB will obtain it. 
  5. Thereafter the Cartridge instance will publish a start maintenance mode event to the Instance Status topic on MB.
  6. The Cloud Controller (CC), which will be subscribed to MB, will receive the event. 
  7. Next CC will publish an member maintenance mode event to the Topology topic to the MB which will be received by the LB, which is subscribed to the MB. 
  8. Then the LB will update the member list. 
  9. Next the Cartridge instance will send a Ready to graceful shutdown event to MB, so that the CC again will receive it. 
  10. Finally the CC will send the information about terminating the instance to the IaaS which will be sent to the Cartridge instance by the IaaS. 

Scale-down workflow will be the same as the above for all scale-down workflows based on the value of load average, in-flight requests and memory consumption.

  • No labels