)]}'
{"nova/values.yaml":[{"author":{"_account_id":21420,"name":"Gage Hugo","email":"gagehugo@gmail.com","username":"ghugo"},"change_message_id":"6eaf7dec8da9e8852bb207f944e93ccbbd9431ed","unresolved":true,"context_lines":[{"line_number":1690,"context_line":"      # NOTE(portdirect): the bind port should not be defined, and is manipulated"},{"line_number":1691,"context_line":"      # via the endpoints section."},{"line_number":1692,"context_line":"      osapi_compute_listen_port: null"},{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"6ca979c5_af0e74be","line":1693,"range":{"start_line":1693,"start_character":66,"end_line":1693,"end_character":79},"updated":"2021-07-16 16:04:52.000000000","message":"Where does this formula come from? What if someone wants to do cpu_count / 2?","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"},{"author":{"_account_id":33130,"name":"Susanta gautam","email":"xuxant02@gmail.com","username":"sgautam"},"change_message_id":"a708eef6c488f755546a11e8a93bb86105b35cdd","unresolved":true,"context_lines":[{"line_number":1690,"context_line":"      # NOTE(portdirect): the bind port should not be defined, and is manipulated"},{"line_number":1691,"context_line":"      # via the endpoints section."},{"line_number":1692,"context_line":"      osapi_compute_listen_port: null"},{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"a60debb1_cee97a06","line":1693,"range":{"start_line":1693,"start_character":66,"end_line":1693,"end_character":79},"in_reply_to":"6ca979c5_af0e74be","updated":"2021-07-16 16:23:01.000000000","message":"About the formula, inspiration is taken from https://github.com/openstack/puppet-openstacklib/blob/master/lib/facter/os_workers.rb. May be we should make the formula for calculation changeable based on values.yaml? Your suggestion will be appreciate.","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"},{"author":{"_account_id":18256,"name":"Hemachandra Reddy","email":"hemachandra.reddy@att.com","username":"hr858f"},"change_message_id":"653d5d9804c023c07bdec0649d2c2060be56d24f","unresolved":true,"context_lines":[{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"},{"line_number":1697,"context_line":"      metadata_workers: auto"},{"line_number":1698,"context_line":"      use_neutron: true"},{"line_number":1699,"context_line":"      firewall_driver: nova.virt.firewall.NoopFirewallDriver"},{"line_number":1700,"context_line":"      linuxnet_interface_driver: openvswitch"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"61b15c4c_e73f6e49","line":1697,"range":{"start_line":1696,"start_character":0,"end_line":1697,"end_character":28},"updated":"2021-07-16 16:44:43.000000000","message":"I\u0027d go with just worker per pod and if it is required the pods can be scaled. I think managing manage workers this way fits better to a non Kubernetes environment.\n\nMoreover, nproc inside a pod reports number of processors on the host. On a typical production system with 80 processors makes it 20 workers per pod as default. That might just be too high.","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"},{"author":{"_account_id":33130,"name":"Susanta gautam","email":"xuxant02@gmail.com","username":"sgautam"},"change_message_id":"782844db4f1027e4a6036cfd0d07ad004761d3e8","unresolved":true,"context_lines":[{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"},{"line_number":1697,"context_line":"      metadata_workers: auto"},{"line_number":1698,"context_line":"      use_neutron: true"},{"line_number":1699,"context_line":"      firewall_driver: nova.virt.firewall.NoopFirewallDriver"},{"line_number":1700,"context_line":"      linuxnet_interface_driver: openvswitch"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"72ae147e_b5f3cead","line":1697,"range":{"start_line":1696,"start_character":0,"end_line":1697,"end_character":28},"in_reply_to":"34c204ce_b75a0485","updated":"2021-07-20 08:17:04.000000000","message":"About the nproc part, thank you for pointing it out. But in default settings in values.yaml, right now there is no resources quota configured in default and here if we set the values in metadata_workers and osapi_compute_workers to Null, by default the workers count will be set to the number of processors which is not good options for machine like one having large number of cpu but less memory. So we thought maybe we can use some kind of formula to calculate the workers process. Scaling is the good options but increasing the worker count based on the machine type using some kind of formula might help while deploying in different environment with similar values file.","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"},{"author":{"_account_id":18256,"name":"Hemachandra Reddy","email":"hemachandra.reddy@att.com","username":"hr858f"},"change_message_id":"25686a7a5c9902d683f2b2c35c4a95cd088b9b1b","unresolved":true,"context_lines":[{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"},{"line_number":1697,"context_line":"      metadata_workers: auto"},{"line_number":1698,"context_line":"      use_neutron: true"},{"line_number":1699,"context_line":"      firewall_driver: nova.virt.firewall.NoopFirewallDriver"},{"line_number":1700,"context_line":"      linuxnet_interface_driver: openvswitch"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"34c204ce_b75a0485","line":1697,"range":{"start_line":1696,"start_character":0,"end_line":1697,"end_character":28},"in_reply_to":"61b15c4c_e73f6e49","updated":"2021-07-16 16:51:00.000000000","message":"Sorry for the typos...\n\nI\u0027d go with just one worker per pod and if it is required the pods can be scaled. I think managing workers this way fits better to a non Kubernetes environment.\n\nMoreover, nproc inside a pod reports number of processors on the host. On a typical production system with 80 processors makes it 20 workers per pod as default. That might just be too high.\nGoing by nproc, it also ignores any CPU resource limits sets on the pod.","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"},{"author":{"_account_id":18250,"name":"Roy Tang","email":"roy.s.tang@att.com","username":"rt7380"},"change_message_id":"a00ed7a88948117b3ab4e1088b536e4b78b0ca6a","unresolved":true,"context_lines":[{"line_number":1693,"context_line":"      # NOTE(workers): number of workers will be calculates with (cpu_count / 4) if"},{"line_number":1694,"context_line":"      # values for osapi_compute_workers and metadata_workers is set to auto."},{"line_number":1695,"context_line":"      # Int number can be provided if any other values needed to be set."},{"line_number":1696,"context_line":"      osapi_compute_workers: auto"},{"line_number":1697,"context_line":"      metadata_workers: auto"},{"line_number":1698,"context_line":"      use_neutron: true"},{"line_number":1699,"context_line":"      firewall_driver: nova.virt.firewall.NoopFirewallDriver"},{"line_number":1700,"context_line":"      linuxnet_interface_driver: openvswitch"}],"source_content_type":"text/x-yaml","patch_set":2,"id":"3960dd65_59e6f717","line":1697,"range":{"start_line":1696,"start_character":0,"end_line":1697,"end_character":28},"in_reply_to":"72ae147e_b5f3cead","updated":"2021-07-27 17:22:41.000000000","message":"I find myself agreeing with Reddy on this, and to sum up my concerns:\n- we are introducing a \"new\" value here to a Nova config parameter that is not in their documentation (integers or null), so this may add confusion to some people.\n- the formula chosen to calculate the number of workers process per pod is not easy to understand, and not easy to change (not configurable)\n- and the concept of workers is really not needed in a containerized env, where anyone can easily control the resources (CPU/RAM/etc) on a pod to optimize it for 1 worker, and to scale it as needed (with proper affinity rules preferably)\n\nIf the issue is we are not optimizing the pod resources, let\u0027s update the limit to make it so, then we can just stick with the simple approach of scaling up and down as needed, instead of relying on some formula that is dependent on resource allocation.","commit_id":"513e7866693574259ec37ac85f2e1671849dae51"}]}