Description of problem: Currently we deploy with reserved_host_memory_mb=2048 on the compute nodes, however with DVR since routers are created on the compute node as well if it has a guest attached to the router via a subnet, and according to https://bugzilla.redhat.com/show_bug.cgi?id=1381610 the metadata proxy on each router is consuming a large amount of memory, we should be able to detect if the deployment is DVR and bump up that value. Version-Release number of selected component (if applicable): RHOP 10 How reproducible: 100% Steps to Reproduce: 1. Deploy with DVR 2. Create networks, subnets, and routers and launch instances 3. See metadata proxy growth on the compute node Actual results: Launching about 50 instances on a compute node (each connected to a different subnet and router), leads to metadata proxy growing to about 1.1GB. https://hurl.corp.redhat.com/dvr-metadata-proxy Expected results: Additional info: We should be able to bump up the reserved_host_memory_mb intelligently during deployment until we fix the metadata proxy growth problem.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2017:1778