Suspect a contributing factor is the growth of the MIQ Server, once a worker(s) have exceeded memory threshold, resulting in large memory footprints of new workers. Data in the logs on this appliance start with the Mig Server being extremely large 1+Gb. To validate the current understanding / theory, can the tests be restarted and full logs captured after a few worker memory threshold exceptions have occurred? What is the size of the environment under management, expecting large as comments indicate 100 tenants in the environment?
Assigning this to the Provider team to get parallel investigation. I will continue looking from the Miq Server being a possible impact to the worker size (additional data requested in https://bugzilla.redhat.com/show_bug.cgi?id=1511877#c11), and have assigned to the provider team so OpenStack dev can investigate the refresh worker itself.
Himanshu, Do you have a full log set from a test run showing this issue? Was the growth in worker size shown in the attached graphs from a single worker instance or are the composed of several instances spawned over time?
Has similar growth been seen with other providers in testing?
Please re-open this ticket if the problem is reproduced.