Autoscaling takes the following factors into account:
- Memory usage
- CPU usage
- Number of request in flight (RIF).
Memory and CPU usages are sent periodically (in 15 second time intervals) by the Python Cartridge Agent to the Complex Event Processor (CEP). This information is then summarized, i.e average, gradient and second derivative, by CEP and sent to the Autoscaler. Based on the received values the autoscaler is able to predict the memory load of the CPU and RIF respectively. If the load is higher than the expected threshold, scaling up will happen and if the load is less than the expected threshold, scaling down will happen.
Troubleshoot autoscaling in Private PaaS
The below steps help you troubleshoot autoscaling when running Private PaaS on a virtual machine.
If you are running Private PaaS on your machine, you can troubleshoot autoscaling through the
wso2carbon.log file, which is in the
<PRIVATE_PAAS_HOME>/repository/logs directory or by scanning through the carbon log on the console.
SSH into the Private PaaS instance using the ssh-client.
keynamerefers to the file path when you are using the EC2 or OpenStack IaaS.
iprefers to the IP address of your virtual machine.
Navigate to the
carbon.logfile, which is in the
Scan the log file to identify any issues with regard to autoscaling.
You can find out the current number of instances in PPaaS by logging into the PPaaS console, navigating to the applications page and viewing the runtime information of a specific application.
For more information see Getting the Runtime Topology of an Application.