)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"0ccd77aae927fc401e85998e574b06f843537838","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Additional upgrade clarifications for cpu-resources"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Based on feedback from the mailing list."},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Change-Id: I468abe984d81c264a588f23d4b3804106339a597"},{"line_number":12,"context_line":"Signed-off-by: Stephen Finucane \u003cstephenfin@redhat.com\u003e"},{"line_number":13,"context_line":"Blueprint: cpu-resources"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"9fb8cfa7_e298671c","line":10,"updated":"2019-06-18 21:10:49.000000000","message":"fmr: http://lists.openstack.org/pipermail/openstack-discuss/2019-June/thread.html#7084","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"f8615917c07a2611f04a3e73f146f28891a09060","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Additional upgrade clarifications for cpu-resources"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Based on feedback from the mailing list."},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Change-Id: I468abe984d81c264a588f23d4b3804106339a597"},{"line_number":12,"context_line":"Signed-off-by: Stephen Finucane \u003cstephenfin@redhat.com\u003e"},{"line_number":13,"context_line":"Blueprint: cpu-resources"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"9fb8cfa7_e1cff726","line":10,"in_reply_to":"9fb8cfa7_e298671c","updated":"2019-06-19 15:28:31.000000000","message":"Done","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"}],"specs/train/approved/cpu-resources.rst":[{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"8f2a5d5dedf5b918dfc28f6245f64b64527f1b48","unresolved":false,"context_lines":[{"line_number":446,"context_line":"cpu_dedicated_set``."},{"line_number":447,"context_line":""},{"line_number":448,"context_line":"The meaning of ``[compute] cpu_shared_set`` will change with this feature, from"},{"line_number":449,"context_line":"being a list of host CPUs used for emulator threads to a list of host CPUs used"},{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_51ca5b4f","line":450,"range":{"start_line":449,"start_character":55,"end_line":450,"end_character":49},"updated":"2019-06-19 04:48:05.000000000","message":"A question, so we will explicitly consume VCPU resource for the emulator threads, right?\n\nHow can we refuse the non-pinned instance to consume those VCPU when we can mix the pinned and non-pinned instance on the same host?","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"f8615917c07a2611f04a3e73f146f28891a09060","unresolved":false,"context_lines":[{"line_number":446,"context_line":"cpu_dedicated_set``."},{"line_number":447,"context_line":""},{"line_number":448,"context_line":"The meaning of ``[compute] cpu_shared_set`` will change with this feature, from"},{"line_number":449,"context_line":"being a list of host CPUs used for emulator threads to a list of host CPUs used"},{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_a1b01f8d","line":450,"range":{"start_line":449,"start_character":55,"end_line":450,"end_character":49},"in_reply_to":"9fb8cfa7_51ca5b4f","updated":"2019-06-19 15:28:31.000000000","message":"\u003e A question, so we will explicitly consume VCPU resource for the\n \u003e emulator threads, right?\n\nIt\u0027s complicated, if we\u0027re using \u0027hw:emulator_threads_policy\u003disolate\u0027 then we need to consume an additional PCPU. If we\u0027re using \u0027hw:emulator_threads_policy\u003dshare\u0027 then we\u0027re going to place these on CPUs from the \u0027[compute] cpu_shared_set\u0027 but we don\u0027t necessarily need to track this. I personally wanted to drop the \u0027isolate\u0027 policy going forward or rather alias it to \u0027share\u0027, since I think everyone is using the \u0027share\u0027 behavior.\n\n \u003e How can we refuse the non-pinned instance to consume those VCPU\n \u003e when we can mix the pinned and non-pinned instance on the same\n \u003e host?\n\nAs noted, I don\u0027t think we need to track these when using the \u0027share\u0027 policy. I wouldn\u0027t consume a \u0027VCPU\u0027 per instance.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"449c8179f789db98b6de9cac2600a4dc5cb74a32","unresolved":false,"context_lines":[{"line_number":446,"context_line":"cpu_dedicated_set``."},{"line_number":447,"context_line":""},{"line_number":448,"context_line":"The meaning of ``[compute] cpu_shared_set`` will change with this feature, from"},{"line_number":449,"context_line":"being a list of host CPUs used for emulator threads to a list of host CPUs used"},{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_51cfbb56","line":450,"range":{"start_line":449,"start_character":55,"end_line":450,"end_character":49},"in_reply_to":"9fb8cfa7_51ca5b4f","updated":"2019-06-19 04:50:41.000000000","message":"I guess we needn\u0027t refuse that. The operator still need to set aggregate for the hosts which is only running pinned instance. And in the future, the opeartor also needs a aggregate for hosts which accept the PCPU and VCPU mixed instance.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"8f2a5d5dedf5b918dfc28f6245f64b64527f1b48","unresolved":false,"context_lines":[{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"},{"line_number":454,"context_line":"cpu_shared_set`` to determine the number of ``VCPU`` resources if"},{"line_number":455,"context_line":"``vcpu_pin_set`` is unset. If ``vcpu_pin_set`` is set, a warning will be raised"},{"line_number":456,"context_line":"and ``vcpu_pin_set`` will continue to be used to calculate the number of"},{"line_number":457,"context_line":"``VCPU`` resource available while ``[compute] cpu_shared_set`` will continue to"},{"line_number":458,"context_line":"be used only for emulator threads."},{"line_number":459,"context_line":""},{"line_number":460,"context_line":"Finally, we will change documentation for the ``cpu_allocation_ratio`` config"},{"line_number":461,"context_line":"option to make it abundantly clear that this option ONLY applies to ``VCPU``"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_51725b7e","line":458,"range":{"start_line":453,"start_character":29,"end_line":458,"end_character":34},"updated":"2019-06-19 04:48:05.000000000","message":"If a host with 12 cpu threads which is used for pinned instance, and it only use `cpu_shared_set\u003d0-3` in stein release, the vcpu_pin_set is empty. In stein release, it reports the number of VCPU based on all available CPU on the host, and let us assume the host already allocated 12 VCPUs for the instances.\n\nAfter upgrade, we change the inventory of VCPU to the number vcpus in cpu_shared_set, it should be 4 VCPUs in this example, then the allocation for the existed instance (12 vCPUs) will exceed this inventory.\n\nWe should have problem for the host which is used for non-pinned instance also. The cpu_shared_set may shrink the existed inventory of VCPU, then leads to over allocation.\n\nOne method in my mind is that keeping report both VCPU and PCPU for this case, and asking the operator to update the cpu_shared_set to the all vCPUs for the host which is used for non-pinned instance. (actually, the operator shouldn\u0027t set this option for the non-pinned host.)","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"2353fcbe3ec1e2ccdb22f2c9a5c19d3838503320","unresolved":false,"context_lines":[{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"},{"line_number":454,"context_line":"cpu_shared_set`` to determine the number of ``VCPU`` resources if"},{"line_number":455,"context_line":"``vcpu_pin_set`` is unset. If ``vcpu_pin_set`` is set, a warning will be raised"},{"line_number":456,"context_line":"and ``vcpu_pin_set`` will continue to be used to calculate the number of"},{"line_number":457,"context_line":"``VCPU`` resource available while ``[compute] cpu_shared_set`` will continue to"},{"line_number":458,"context_line":"be used only for emulator threads."},{"line_number":459,"context_line":""},{"line_number":460,"context_line":"Finally, we will change documentation for the ``cpu_allocation_ratio`` config"},{"line_number":461,"context_line":"option to make it abundantly clear that this option ONLY applies to ``VCPU``"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7faddb67_17b15890","line":458,"range":{"start_line":453,"start_character":29,"end_line":458,"end_character":34},"in_reply_to":"9fb8cfa7_1475a798","updated":"2019-07-04 12:32:19.000000000","message":"I\u0027m ok with this assumption, it sounds make sense. Not sure we have enough explain for operator before. Probably we need those detail in the upgrade note.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"f8615917c07a2611f04a3e73f146f28891a09060","unresolved":false,"context_lines":[{"line_number":450,"context_line":"for both emulator threads and ``VCPU`` resources. Note that because this option"},{"line_number":451,"context_line":"already exists, we can\u0027t rely on its presence to do things like ignore"},{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"},{"line_number":454,"context_line":"cpu_shared_set`` to determine the number of ``VCPU`` resources if"},{"line_number":455,"context_line":"``vcpu_pin_set`` is unset. If ``vcpu_pin_set`` is set, a warning will be raised"},{"line_number":456,"context_line":"and ``vcpu_pin_set`` will continue to be used to calculate the number of"},{"line_number":457,"context_line":"``VCPU`` resource available while ``[compute] cpu_shared_set`` will continue to"},{"line_number":458,"context_line":"be used only for emulator threads."},{"line_number":459,"context_line":""},{"line_number":460,"context_line":"Finally, we will change documentation for the ``cpu_allocation_ratio`` config"},{"line_number":461,"context_line":"option to make it abundantly clear that this option ONLY applies to ``VCPU``"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_1475a798","line":458,"range":{"start_line":453,"start_character":29,"end_line":458,"end_character":34},"in_reply_to":"9fb8cfa7_51725b7e","updated":"2019-06-19 15:28:31.000000000","message":"\u003e If a host with 12 cpu threads which is used for pinned instance, and it only use `cpu_shared_set\u003d0-3` in stein release, the vcpu_pin_set is empty. In stein release, it reports the number of VCPU based on all available CPU on the host, and let us assume the host already allocated 12 VCPUs for the instances.\n\nThis is _exceptionally_ unlikely and I\u0027ve chosen to ignore this possibility. The only reason to define \u0027cpu_shared_set\u0027 in Stein or before is to use emulator thread offloading. The idea behind emulator thread offloading is to isolate the additional work the emulator needs to do from the work the guest OS is doing, mainly for real-time. If you were using \u0027cpu_shared_set\u0027 without \u0027vcpu_pin_set\u0027 then your guest vCPUs can be pinned to any host core including those listed in \u0027cpu_shared_set\u0027. This would defeat the whole purpose of the feature so I\u0027m relying on people not having done this silly thing :) If we can\u0027t rely on this then we need to specify that \u0027cpu_dedicated_set\u0027 is always set, even if it\u0027s to an empty list/None, whenever \u0027cpu_shared_set\u0027 is set. That makes upgrade more difficult. See conversation starting here:\n\nhttp://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2019-06-18.log.html#t2019-06-18T10:11:47\n\nDo you think this is a reasonable assumption?","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"8f2a5d5dedf5b918dfc28f6245f64b64527f1b48","unresolved":false,"context_lines":[{"line_number":494,"context_line":""},{"line_number":495,"context_line":":Summary: We will provide a way to reshape inventory of existing instances using"},{"line_number":496,"context_line":"          pinned CPUs to use inventory of the ``PCPU`` resource class instead of"},{"line_number":497,"context_line":"          ``VCPU``. This will be a manual step that should be run once a"},{"line_number":498,"context_line":"          critical mass of hosts have been updated and start reporting ``PCPU``"},{"line_number":499,"context_line":"          resources."},{"line_number":500,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_71cbdf71","line":497,"range":{"start_line":497,"start_character":35,"end_line":497,"end_character":46},"updated":"2019-06-19 04:48:05.000000000","message":"this step should be before enabling the prefilter.\n\nIf we enable the prefilter first, then we may over-allocate the resource.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"8f2a5d5dedf5b918dfc28f6245f64b64527f1b48","unresolved":false,"context_lines":[{"line_number":496,"context_line":"          pinned CPUs to use inventory of the ``PCPU`` resource class instead of"},{"line_number":497,"context_line":"          ``VCPU``. This will be a manual step that should be run once a"},{"line_number":498,"context_line":"          critical mass of hosts have been updated and start reporting ``PCPU``"},{"line_number":499,"context_line":"          resources."},{"line_number":500,"context_line":""},{"line_number":501,"context_line":"For existing compute nodes that have guests which use dedicated CPUs, the virt"},{"line_number":502,"context_line":"driver will need to move inventory of existing ``VCPU`` resources (which are"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_119dc365","line":499,"updated":"2019-06-19 04:48:05.000000000","message":"This resharp can be triggered by cpu_dedicated_set being set.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"f8615917c07a2611f04a3e73f146f28891a09060","unresolved":false,"context_lines":[{"line_number":496,"context_line":"          pinned CPUs to use inventory of the ``PCPU`` resource class instead of"},{"line_number":497,"context_line":"          ``VCPU``. This will be a manual step that should be run once a"},{"line_number":498,"context_line":"          critical mass of hosts have been updated and start reporting ``PCPU``"},{"line_number":499,"context_line":"          resources."},{"line_number":500,"context_line":""},{"line_number":501,"context_line":"For existing compute nodes that have guests which use dedicated CPUs, the virt"},{"line_number":502,"context_line":"driver will need to move inventory of existing ``VCPU`` resources (which are"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9fb8cfa7_2162ef17","line":499,"in_reply_to":"9fb8cfa7_119dc365","updated":"2019-06-19 15:28:31.000000000","message":"Oh, good idea.","commit_id":"484b15a140a269d6073e3f1f67d5a42d51b0ba66"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"81981e1289391e008841b73441e1fe7bf8c5b754","unresolved":false,"context_lines":[{"line_number":400,"context_line":"* `NUMA, CPU Pinning and \u0027vcpu_pin_set\u0027"},{"line_number":401,"context_line":"  \u003chttps://that.guru/blog/cpu-resources/\u003e`__"},{"line_number":402,"context_line":""},{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_f7e3ead7","line":403,"range":{"start_line":403,"start_character":64,"end_line":403,"end_character":69},"updated":"2019-09-23 22:10:06.000000000","message":"Train?","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":400,"context_line":"* `NUMA, CPU Pinning and \u0027vcpu_pin_set\u0027"},{"line_number":401,"context_line":"  \u003chttps://that.guru/blog/cpu-resources/\u003e`__"},{"line_number":402,"context_line":""},{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_4dc4660e","line":403,"range":{"start_line":403,"start_character":64,"end_line":403,"end_character":69},"in_reply_to":"3fa7e38b_e94d8ac4","updated":"2020-01-27 11:17:16.000000000","message":"Train (this code has already merged - this is a post-merge fixup)","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":400,"context_line":"* `NUMA, CPU Pinning and \u0027vcpu_pin_set\u0027"},{"line_number":401,"context_line":"  \u003chttps://that.guru/blog/cpu-resources/\u003e`__"},{"line_number":402,"context_line":""},{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_6dc7e212","line":403,"range":{"start_line":403,"start_character":64,"end_line":403,"end_character":69},"in_reply_to":"3fa7e38b_f7e3ead7","updated":"2020-01-27 11:17:16.000000000","message":"Done","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"d1c433f6f545a9ce0c5f8939a534c57ee3b73d0c","unresolved":false,"context_lines":[{"line_number":400,"context_line":"* `NUMA, CPU Pinning and \u0027vcpu_pin_set\u0027"},{"line_number":401,"context_line":"  \u003chttps://that.guru/blog/cpu-resources/\u003e`__"},{"line_number":402,"context_line":""},{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_e94d8ac4","line":403,"range":{"start_line":403,"start_character":64,"end_line":403,"end_character":69},"in_reply_to":"3fa7e38b_f7e3ead7","updated":"2019-10-23 00:12:20.000000000","message":"Ussuri :)","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"d1c433f6f545a9ce0c5f8939a534c57ee3b73d0c","unresolved":false,"context_lines":[{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."},{"line_number":407,"context_line":""},{"line_number":408,"context_line":"Configuration options"},{"line_number":409,"context_line":"~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_49fdbe39","line":406,"updated":"2019-10-23 00:12:20.000000000","message":"It is not clear to me how this interacts or impacts NUMA. Could you either point me to the docs/specs, or state what the impact is (or that there is no impact)?","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":403,"context_line":"A key point here is that the new behavior must be opt-in during Stein. We"},{"line_number":404,"context_line":"recognize that operators may need time to upgrade a critical number of compute"},{"line_number":405,"context_line":"nodes so that they are reporting ``PCPU`` classes. This is reflected at"},{"line_number":406,"context_line":"numerous points below."},{"line_number":407,"context_line":""},{"line_number":408,"context_line":"Configuration options"},{"line_number":409,"context_line":"~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_edd2f2ce","line":406,"in_reply_to":"3fa7e38b_49fdbe39","updated":"2020-01-27 11:17:16.000000000","message":"No impact. We\u0027re not modelling NUMA in placement yet so things haven\u0027t changed there. This only affects our accounting","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"d1c433f6f545a9ce0c5f8939a534c57ee3b73d0c","unresolved":false,"context_lines":[{"line_number":412,"context_line":"          config options and set one or both of the existing ``[compute]"},{"line_number":413,"context_line":"          cpu_shared_set`` and new ``[compute] cpu_dedicated_set`` options."},{"line_number":414,"context_line":""},{"line_number":415,"context_line":"We will deprecate the ``vcpu_pin_set`` config option in Train. If both the"},{"line_number":416,"context_line":"``[compute] cpu_dedicated_set`` and ``[compute] cpu_shared_set`` config options"},{"line_number":417,"context_line":"are set in Train, this option will be ignored entirely and ``[compute]"},{"line_number":418,"context_line":"cpu_shared_set`` will be used in place of ``vcpu_pin_set`` to calculate the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_094906b6","line":415,"range":{"start_line":415,"start_character":56,"end_line":415,"end_character":61},"updated":"2019-10-23 00:12:20.000000000","message":"You could say \"this release\" instead of \u0027Train\u0027 . The release name is scattered throughout.","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"81981e1289391e008841b73441e1fe7bf8c5b754","unresolved":false,"context_lines":[{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"},{"line_number":454,"context_line":"cpu_shared_set`` to determine the number of ``VCPU`` resources if"},{"line_number":455,"context_line":"``vcpu_pin_set`` is unset. If ``vcpu_pin_set`` is set, a warning will be raised"},{"line_number":456,"context_line":"and ``vcpu_pin_set`` will continue to be used to calculate the number of"},{"line_number":457,"context_line":"``VCPU`` resource available while ``[compute] cpu_shared_set`` will continue to"},{"line_number":458,"context_line":"be used only for emulator threads."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_57d2fe01","line":455,"range":{"start_line":455,"start_character":73,"end_line":455,"end_character":79},"updated":"2019-09-23 22:10:06.000000000","message":"logged","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":452,"context_line":"``vcpu_pin_set``, as outlined previously, and must rely on ``[compute]"},{"line_number":453,"context_line":"cpu_dedicated_set`` instead. For this same reason, we will only use ``[compute]"},{"line_number":454,"context_line":"cpu_shared_set`` to determine the number of ``VCPU`` resources if"},{"line_number":455,"context_line":"``vcpu_pin_set`` is unset. If ``vcpu_pin_set`` is set, a warning will be raised"},{"line_number":456,"context_line":"and ``vcpu_pin_set`` will continue to be used to calculate the number of"},{"line_number":457,"context_line":"``VCPU`` resource available while ``[compute] cpu_shared_set`` will continue to"},{"line_number":458,"context_line":"be used only for emulator threads."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_8ddd7edf","line":455,"range":{"start_line":455,"start_character":73,"end_line":455,"end_character":79},"in_reply_to":"3fa7e38b_57d2fe01","updated":"2020-01-27 11:17:16.000000000","message":"Done","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"81981e1289391e008841b73441e1fe7bf8c5b754","unresolved":false,"context_lines":[{"line_number":479,"context_line":"Flavor extra specs and image metadata properties"},{"line_number":480,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":481,"context_line":""},{"line_number":482,"context_line":":Summary: We will add a prefilter that we will rewrite legacy flavor extra"},{"line_number":483,"context_line":"          specs and image metadata properties. This should be enabled once a"},{"line_number":484,"context_line":"          critical mass of hosts have been updated and start reporting ``PCPU``"},{"line_number":485,"context_line":"          resources."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_17d8061c","line":482,"range":{"start_line":482,"start_character":39,"end_line":482,"end_character":41},"updated":"2019-09-23 22:10:06.000000000","message":"strike","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":479,"context_line":"Flavor extra specs and image metadata properties"},{"line_number":480,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":481,"context_line":""},{"line_number":482,"context_line":":Summary: We will add a prefilter that we will rewrite legacy flavor extra"},{"line_number":483,"context_line":"          specs and image metadata properties. This should be enabled once a"},{"line_number":484,"context_line":"          critical mass of hosts have been updated and start reporting ``PCPU``"},{"line_number":485,"context_line":"          resources."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_add87aec","line":482,"range":{"start_line":482,"start_character":39,"end_line":482,"end_character":41},"in_reply_to":"3fa7e38b_17d8061c","updated":"2020-01-27 11:17:16.000000000","message":"Done","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"81981e1289391e008841b73441e1fe7bf8c5b754","unresolved":false,"context_lines":[{"line_number":550,"context_line":"    unset, no action is necessary unless:"},{"line_number":551,"context_line":""},{"line_number":552,"context_line":"  - If set, unset ``reserved_host_cpus`` and set ``[compute] cpu_shared_set``"},{"line_number":553,"context_line":"    to the entire range of host cores minus a number of host cores you wish to"},{"line_number":554,"context_line":"    reserve."},{"line_number":555,"context_line":""},{"line_number":556,"context_line":"- (Once a sufficient number of compute nodes have been updated) Enable the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_97fdd66f","line":553,"range":{"start_line":553,"start_character":44,"end_line":553,"end_character":55},"updated":"2019-09-23 22:10:06.000000000","message":"whichever","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"5e34a5c90e185acfb2498ca20c2b1feb2ce039fb","unresolved":false,"context_lines":[{"line_number":550,"context_line":"    unset, no action is necessary unless:"},{"line_number":551,"context_line":""},{"line_number":552,"context_line":"  - If set, unset ``reserved_host_cpus`` and set ``[compute] cpu_shared_set``"},{"line_number":553,"context_line":"    to the entire range of host cores minus a number of host cores you wish to"},{"line_number":554,"context_line":"    reserve."},{"line_number":555,"context_line":""},{"line_number":556,"context_line":"- (Once a sufficient number of compute nodes have been updated) Enable the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3fa7e38b_0de3cea1","line":553,"range":{"start_line":553,"start_character":44,"end_line":553,"end_character":55},"in_reply_to":"3fa7e38b_97fdd66f","updated":"2020-01-27 11:17:16.000000000","message":"I wasn\u0027t sure what this was referring to and it reads okay to me as is, so I\u0027ve left this one","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"2353fcbe3ec1e2ccdb22f2c9a5c19d3838503320","unresolved":false,"context_lines":[{"line_number":531,"context_line":"Summary"},{"line_number":532,"context_line":"~~~~~~~"},{"line_number":533,"context_line":""},{"line_number":534,"context_line":"The final upgrade process will look like similar to standard upgrades, though"},{"line_number":535,"context_line":"there are some slight changes necessary:"},{"line_number":536,"context_line":""},{"line_number":537,"context_line":"- Upgrade controllers"},{"line_number":538,"context_line":""},{"line_number":539,"context_line":"- Update compute nodes in batches"},{"line_number":540,"context_line":""},{"line_number":541,"context_line":"  For compute nodes hosting pinned instances:"},{"line_number":542,"context_line":""},{"line_number":543,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_dedicated_set``. If"},{"line_number":544,"context_line":"    unset, set ``[compute] cpu_dedicated_set`` to the entire range of host"},{"line_number":545,"context_line":"    CPUs."},{"line_number":546,"context_line":""},{"line_number":547,"context_line":"  For compute nodes hosting unpinned instances:"},{"line_number":548,"context_line":""},{"line_number":549,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_shared_set``. If"},{"line_number":550,"context_line":"    unset, no action is necessary unless:"},{"line_number":551,"context_line":""},{"line_number":552,"context_line":"  - If set, unset ``reserved_host_cpus`` and set ``[compute] cpu_shared_set``"},{"line_number":553,"context_line":"    to the entire range of host cores minus a number of host cores you wish to"},{"line_number":554,"context_line":"    reserve."},{"line_number":555,"context_line":""},{"line_number":556,"context_line":"- (Once a sufficient number of compute nodes have been updated) Enable the"},{"line_number":557,"context_line":"  prefilter"},{"line_number":558,"context_line":""},{"line_number":559,"context_line":"- Upgrade remaining compute nodes"},{"line_number":560,"context_line":""},{"line_number":561,"context_line":"Implementation"},{"line_number":562,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7faddb67_17271815","line":559,"range":{"start_line":534,"start_character":0,"end_line":559,"end_character":33},"updated":"2019-07-04 12:32:19.000000000","message":"++.\n\nI may also want to add those two rules:\n\n1. The operator doesn\u0027t allow to change a different value for dedicated_cpu_set with vcpu_pin_set when any instance is running on the host. \n2. The operator doesn\u0027t allow to change the value of dedicated_cpu_set and shared_cpu_set when any instance is running on the host.\n\nBut I\u0027m already ok with this.","commit_id":"0fc90beec57d5f0001ef2c85b4c414b4c41b28c9"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"803f510acc9a919436a567f647d386622c188b46","unresolved":false,"context_lines":[{"line_number":540,"context_line":"  For compute nodes hosting pinned instances:"},{"line_number":541,"context_line":""},{"line_number":542,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_dedicated_set``. If"},{"line_number":543,"context_line":"    unset, set ``[compute] cpu_dedicated_set`` to the entire range of host"},{"line_number":544,"context_line":"    CPUs."},{"line_number":545,"context_line":""},{"line_number":546,"context_line":"  For compute nodes hosting unpinned instances:"},{"line_number":547,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3fa7e38b_26c3243b","line":544,"range":{"start_line":543,"start_character":47,"end_line":544,"end_character":9},"updated":"2020-01-29 11:15:15.000000000","message":"nit: this is not really correct. it should be set to the range of host cpus - cpus reseved for the host use.\nthis is to account for the lack of reserved_host_cpus and the fact that you normally want to reserve core 0 for host use for the agnent. i would optionally be advise able for at lest\n1 core to be set in the cpu_shared_set so that you could move the emulator threads to run on that core.\n\nif you followed the current advice it would not be wrong exactly but i would would not consider it to be the best practice either.","commit_id":"4b506d0e9f75200940aa41de419cc626b7dfbbbf"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"35b7dcf62251c524cf03bac7ffd53cb268e67ec9","unresolved":false,"context_lines":[{"line_number":540,"context_line":"  For compute nodes hosting pinned instances:"},{"line_number":541,"context_line":""},{"line_number":542,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_dedicated_set``. If"},{"line_number":543,"context_line":"    unset, set ``[compute] cpu_dedicated_set`` to the entire range of host"},{"line_number":544,"context_line":"    CPUs."},{"line_number":545,"context_line":""},{"line_number":546,"context_line":"  For compute nodes hosting unpinned instances:"},{"line_number":547,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3fa7e38b_81fbe212","line":544,"range":{"start_line":543,"start_character":47,"end_line":544,"end_character":9},"in_reply_to":"3fa7e38b_067388e9","updated":"2020-01-29 11:45:09.000000000","message":"if you did not have the vcpu_pin_set defiend yes\nif you did then you should copy the value into cpu_dedicated_set to maintian the same behavior.","commit_id":"4b506d0e9f75200940aa41de419cc626b7dfbbbf"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"d30fc830a730c65dfc5a501a16d13a2727f12fe9","unresolved":false,"context_lines":[{"line_number":540,"context_line":"  For compute nodes hosting pinned instances:"},{"line_number":541,"context_line":""},{"line_number":542,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_dedicated_set``. If"},{"line_number":543,"context_line":"    unset, set ``[compute] cpu_dedicated_set`` to the entire range of host"},{"line_number":544,"context_line":"    CPUs."},{"line_number":545,"context_line":""},{"line_number":546,"context_line":"  For compute nodes hosting unpinned instances:"},{"line_number":547,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3fa7e38b_e10f3629","line":544,"range":{"start_line":543,"start_character":47,"end_line":544,"end_character":9},"in_reply_to":"3fa7e38b_26c3243b","updated":"2020-01-29 11:46:03.000000000","message":"\u003e nit: this is not really correct. it should be set to the range of\n \u003e host cpus - cpus reseved for the host use.\n \u003e this is to account for the lack of reserved_host_cpus and the fact\n \u003e that you normally want to reserve core 0 for host use for the\n \u003e agnent. i would optionally be advise able for at lest\n \u003e 1 core to be set in the cpu_shared_set so that you could move the\n \u003e emulator threads to run on that core.\n\nI don\u0027t think we care about \u0027reserved_host_cpus\u0027 for pinned instances though, right? What I\u0027m describing is how one would get the exact same behavior they had before. Realistically no one should be in this situation and they should have defined \u0027vcpu_pin_set\u0027.","commit_id":"4b506d0e9f75200940aa41de419cc626b7dfbbbf"},{"author":{"_account_id":9708,"name":"Balazs Gibizer","display_name":"gibi","email":"gibizer@gmail.com","username":"gibi"},"change_message_id":"8a51c19ec56d40b632ee17164a5f416a1e3e8ba1","unresolved":false,"context_lines":[{"line_number":540,"context_line":"  For compute nodes hosting pinned instances:"},{"line_number":541,"context_line":""},{"line_number":542,"context_line":"  - If set, unset ``vcpu_pin_set`` and set ``[compute] cpu_dedicated_set``. If"},{"line_number":543,"context_line":"    unset, set ``[compute] cpu_dedicated_set`` to the entire range of host"},{"line_number":544,"context_line":"    CPUs."},{"line_number":545,"context_line":""},{"line_number":546,"context_line":"  For compute nodes hosting unpinned instances:"},{"line_number":547,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3fa7e38b_067388e9","line":544,"range":{"start_line":543,"start_character":47,"end_line":544,"end_character":9},"in_reply_to":"3fa7e38b_26c3243b","updated":"2020-01-29 11:38:15.000000000","message":"I think Stephen says that if before the upgrade use had no vcpu_pin_set configured then you used every physical cpu from the host, so if you want to keep doing that then set cp_dedicated_set accordingly. I think this is the correct instruction to keep the same behavior (which might not an optimal behavior)","commit_id":"4b506d0e9f75200940aa41de419cc626b7dfbbbf"}]}