Created attachment 1147617 [details] screenshot of top on an appliance managing QE's vsphere55, ~10 minutes after adding it. Description of problem: In QE, we have discovered that the 5.6 appliances develop excessive swapping with the same load as eg. 5.5 takes without problems. This has bad effects on our QE infrastructure. Version-Release number of selected component (if applicable): 5.6.0.0-beta2 How reproducible: always Steps to Reproduce: 1. Deploy the appliance 2. Add a provider (eg. QE's vsphere55) 3. Wait and watch eg. top Actual results: Some swapping (in hundreds of MB) develops quite fast (~5min). We have also seen appliances taking 6G of memory, those probably managed more providers Expected results: Nothing special happens. Additional info:
Created attachment 1147651 [details] evm.log file up to the point where the evmserverd service is stopped
Created attachment 1147652 [details] standard top reports showing pid count growth, memory and swap growth and cpu idle
Created attachment 1147653 [details] top output for detailed process analysis
Created attachment 1147655 [details] summary page showing counts of processes by worker type
So we believe this was all caused by appliances being sent to the wrong (smaller) osp flavor, appliance is running much better under 8GB of memory.