)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"11bd2b0587e0f7421384e274580c0c15a617cac7","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"aa5f668c_66b8d83f","updated":"2023-02-21 16:44:07.000000000","message":"content wise im ok with this i have one note i would like to highlgh to operators about what govenor to use but other then that im generally happy with this.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"972184f4aa3ea795fca6dca47666f0a4ecbe8fc9","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":2,"id":"c6c9aa47_237fdc16","updated":"2023-02-23 10:59:04.000000000","message":"im ok with this but there is one change i would make inline","commit_id":"9d70eb54ef1b77c04cdef725e5a5992df7dd3506"}],"doc/source/admin/cpu-topologies.rst":[{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"11bd2b0587e0f7421384e274580c0c15a617cac7","unresolved":true,"context_lines":[{"line_number":730,"context_line":""},{"line_number":731,"context_line":".. _configure-hyperv-numa:"},{"line_number":732,"context_line":""},{"line_number":733,"context_line":"Configuring CPU power management for dedicated cores"},{"line_number":734,"context_line":"----------------------------------------------------"},{"line_number":735,"context_line":""},{"line_number":736,"context_line":".. versionchanged:: 27.0.0"}],"source_content_type":"text/x-rst","patch_set":1,"id":"1c6f98c9_72d6fed8","line":733,"updated":"2023-02-21 16:44:07.000000000","message":"this is a bit different form the other content in this docs but its close enough for now. if we add other power management or monitoring features in the future i would prefer this to be moved to be located with that.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"972184f4aa3ea795fca6dca47666f0a4ecbe8fc9","unresolved":false,"context_lines":[{"line_number":730,"context_line":""},{"line_number":731,"context_line":".. _configure-hyperv-numa:"},{"line_number":732,"context_line":""},{"line_number":733,"context_line":"Configuring CPU power management for dedicated cores"},{"line_number":734,"context_line":"----------------------------------------------------"},{"line_number":735,"context_line":""},{"line_number":736,"context_line":".. versionchanged:: 27.0.0"}],"source_content_type":"text/x-rst","patch_set":1,"id":"c9a36bf9_a7c3a4af","line":733,"in_reply_to":"1c6f98c9_72d6fed8","updated":"2023-02-23 10:59:04.000000000","message":"Ack","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"eba74a54da9b852e1edf848c063274fd45eea15f","unresolved":true,"context_lines":[{"line_number":730,"context_line":""},{"line_number":731,"context_line":".. _configure-hyperv-numa:"},{"line_number":732,"context_line":""},{"line_number":733,"context_line":"Configuring CPU power management for dedicated cores"},{"line_number":734,"context_line":"----------------------------------------------------"},{"line_number":735,"context_line":""},{"line_number":736,"context_line":".. versionchanged:: 27.0.0"}],"source_content_type":"text/x-rst","patch_set":1,"id":"ee49f03a_6d418d5c","line":733,"in_reply_to":"1c6f98c9_72d6fed8","updated":"2023-02-23 10:59:47.000000000","message":"Yeah like I said to you on IRC, I think we can start with the docs here as all the power management is for the moment about pinned CPUs but say, if we merge something else quickly (let\u0027s say virtiofs-scaphandre for example) then we could have a larger top-level doc for this.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"cfd899a91aba9df348d5f976acfcaa878dd646e4","unresolved":false,"context_lines":[{"line_number":730,"context_line":""},{"line_number":731,"context_line":".. _configure-hyperv-numa:"},{"line_number":732,"context_line":""},{"line_number":733,"context_line":"Configuring CPU power management for dedicated cores"},{"line_number":734,"context_line":"----------------------------------------------------"},{"line_number":735,"context_line":""},{"line_number":736,"context_line":".. versionchanged:: 27.0.0"}],"source_content_type":"text/x-rst","patch_set":1,"id":"c5555530_9b9d13b5","line":733,"in_reply_to":"ee49f03a_6d418d5c","updated":"2023-02-23 11:06:08.000000000","message":"Done","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"11bd2b0587e0f7421384e274580c0c15a617cac7","unresolved":true,"context_lines":[{"line_number":743,"context_line":"   libvirt/KVM driver."},{"line_number":744,"context_line":""},{"line_number":745,"context_line":"For power saving reasons, operators can decide to turn down the performance of"},{"line_number":746,"context_line":"CPU cores that are not currently in use. For obvious reasons, Nova only allows"},{"line_number":747,"context_line":"to change the CPU performance of a dedicated core and not a shared one."},{"line_number":748,"context_line":"Accordingly, usage of this feature relies on the reading of"},{"line_number":749,"context_line":":oslo.config:option:`compute.cpu_dedicated_set` config option to know which CPU"}],"source_content_type":"text/x-rst","patch_set":1,"id":"f57f1402_27246bfa","line":746,"range":{"start_line":746,"start_character":10,"end_line":746,"end_character":39},"updated":"2023-02-21 16:44:07.000000000","message":"you can also do it when they are in use.\n\ni would not recommend using performance as the governor in production.\nmore on that later","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"972184f4aa3ea795fca6dca47666f0a4ecbe8fc9","unresolved":false,"context_lines":[{"line_number":743,"context_line":"   libvirt/KVM driver."},{"line_number":744,"context_line":""},{"line_number":745,"context_line":"For power saving reasons, operators can decide to turn down the performance of"},{"line_number":746,"context_line":"CPU cores that are not currently in use. For obvious reasons, Nova only allows"},{"line_number":747,"context_line":"to change the CPU performance of a dedicated core and not a shared one."},{"line_number":748,"context_line":"Accordingly, usage of this feature relies on the reading of"},{"line_number":749,"context_line":":oslo.config:option:`compute.cpu_dedicated_set` config option to know which CPU"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9448d5a9_31a05fca","line":746,"range":{"start_line":746,"start_character":10,"end_line":746,"end_character":39},"in_reply_to":"f57f1402_27246bfa","updated":"2023-02-23 10:59:04.000000000","message":"Done","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"eba74a54da9b852e1edf848c063274fd45eea15f","unresolved":false,"context_lines":[{"line_number":743,"context_line":"   libvirt/KVM driver."},{"line_number":744,"context_line":""},{"line_number":745,"context_line":"For power saving reasons, operators can decide to turn down the performance of"},{"line_number":746,"context_line":"CPU cores that are not currently in use. For obvious reasons, Nova only allows"},{"line_number":747,"context_line":"to change the CPU performance of a dedicated core and not a shared one."},{"line_number":748,"context_line":"Accordingly, usage of this feature relies on the reading of"},{"line_number":749,"context_line":":oslo.config:option:`compute.cpu_dedicated_set` config option to know which CPU"}],"source_content_type":"text/x-rst","patch_set":1,"id":"69d4aef8_c536a082","line":746,"range":{"start_line":746,"start_character":10,"end_line":746,"end_character":39},"in_reply_to":"f57f1402_27246bfa","updated":"2023-02-23 10:59:47.000000000","message":"Good point, ack will clarify.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"11bd2b0587e0f7421384e274580c0c15a617cac7","unresolved":true,"context_lines":[{"line_number":766,"context_line":"The strategy can be chosen using"},{"line_number":767,"context_line":":oslo.config:option:`libvirt.cpu_power_management_strategy` config option."},{"line_number":768,"context_line":"``cpu_state`` supports the first online/offline strategy, while ``governor``"},{"line_number":769,"context_line":"sets the alternative strategy."},{"line_number":770,"context_line":"If the strategy is set to ``governor``, a couple of config options are provided"},{"line_number":771,"context_line":"to define which exact CPU govenor to use for each of the up and down states :"},{"line_number":772,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf7858c5_8430bc27","line":769,"updated":"2023-02-21 16:44:07.000000000","message":"we default to cpu_state because this give the best power savings and allow the operator to mange the govoner via other tools like tuned so you get the best of both worlds.\n\nthis is the correct strategy  to default too and is what most people shoudl use.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"eba74a54da9b852e1edf848c063274fd45eea15f","unresolved":false,"context_lines":[{"line_number":766,"context_line":"The strategy can be chosen using"},{"line_number":767,"context_line":":oslo.config:option:`libvirt.cpu_power_management_strategy` config option."},{"line_number":768,"context_line":"``cpu_state`` supports the first online/offline strategy, while ``governor``"},{"line_number":769,"context_line":"sets the alternative strategy."},{"line_number":770,"context_line":"If the strategy is set to ``governor``, a couple of config options are provided"},{"line_number":771,"context_line":"to define which exact CPU govenor to use for each of the up and down states :"},{"line_number":772,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"643b579e_1c1d661f","line":769,"in_reply_to":"bf7858c5_8430bc27","updated":"2023-02-23 10:59:47.000000000","message":"Ack","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"972184f4aa3ea795fca6dca47666f0a4ecbe8fc9","unresolved":false,"context_lines":[{"line_number":766,"context_line":"The strategy can be chosen using"},{"line_number":767,"context_line":":oslo.config:option:`libvirt.cpu_power_management_strategy` config option."},{"line_number":768,"context_line":"``cpu_state`` supports the first online/offline strategy, while ``governor``"},{"line_number":769,"context_line":"sets the alternative strategy."},{"line_number":770,"context_line":"If the strategy is set to ``governor``, a couple of config options are provided"},{"line_number":771,"context_line":"to define which exact CPU govenor to use for each of the up and down states :"},{"line_number":772,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"8d6fdef4_12d604f3","line":769,"in_reply_to":"bf7858c5_8430bc27","updated":"2023-02-23 10:59:04.000000000","message":"Done","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"11bd2b0587e0f7421384e274580c0c15a617cac7","unresolved":true,"context_lines":[{"line_number":778,"context_line":".. important::"},{"line_number":779,"context_line":"   This is the responsibility of the operator to ensure that the govenors"},{"line_number":780,"context_line":"   defined by the configuration options are currently supported by the OS"},{"line_number":781,"context_line":"   underlying kernel that runs the compute service."},{"line_number":782,"context_line":""},{"line_number":783,"context_line":"As an example, a ``nova.conf`` part of configuration would look like::"},{"line_number":784,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7e3233ef_b6adedb3","line":781,"updated":"2023-02-21 16:44:07.000000000","message":"i dont know where to note this but it woudl be good to provode a recomendation fro a alternivie cpu_power_governor_high value.\n\n\nin order of perfernce you should use  \nschedutil\u003econservative\u003eondemand\u003eperformance\n\nperformance will have the worst overall power usage with miniumally better or equal perfroamce to the alternitives.\n\nschedutil using info form the kernel schduler based on stats related to the task to determin how to scale the frequency dynmically. \n\nconservative and ondemand but use time based huristics ondemand is just more agressive in increasing and decreasing the cpu frequency.\n\nperformance just sets the core to its max frequency permently which is very wasteful if the guest is not active.\n\nwe are using it as our default as it the only one we can be sure is alwasy there but \nschedutil woudl be a better default if we could rely on it.","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"eba74a54da9b852e1edf848c063274fd45eea15f","unresolved":false,"context_lines":[{"line_number":778,"context_line":".. important::"},{"line_number":779,"context_line":"   This is the responsibility of the operator to ensure that the govenors"},{"line_number":780,"context_line":"   defined by the configuration options are currently supported by the OS"},{"line_number":781,"context_line":"   underlying kernel that runs the compute service."},{"line_number":782,"context_line":""},{"line_number":783,"context_line":"As an example, a ``nova.conf`` part of configuration would look like::"},{"line_number":784,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"c024eceb_e0ced8c6","line":781,"in_reply_to":"7e3233ef_b6adedb3","updated":"2023-02-23 10:59:47.000000000","message":"Ack","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"4f9bd5cfdbdee56b3432f359e10d5285bae93c86","unresolved":false,"context_lines":[{"line_number":778,"context_line":".. important::"},{"line_number":779,"context_line":"   This is the responsibility of the operator to ensure that the govenors"},{"line_number":780,"context_line":"   defined by the configuration options are currently supported by the OS"},{"line_number":781,"context_line":"   underlying kernel that runs the compute service."},{"line_number":782,"context_line":""},{"line_number":783,"context_line":"As an example, a ``nova.conf`` part of configuration would look like::"},{"line_number":784,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"45be3d2c_0b974fd5","line":781,"in_reply_to":"7e3233ef_b6adedb3","updated":"2023-02-23 10:59:42.000000000","message":"Done","commit_id":"a328da921b274b1d1930606319c5a2438419b8c1"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"972184f4aa3ea795fca6dca47666f0a4ecbe8fc9","unresolved":true,"context_lines":[{"line_number":789,"context_line":"   the high-power state (if the kernel supports it) as the CPU frequency is"},{"line_number":790,"context_line":"   dynamically set based on CPU task states. Other governors may be worth to"},{"line_number":791,"context_line":"   be tested, including ``conservative`` and ``ondemand`` which are quite a bit"},{"line_number":792,"context_line":"   more power consuming. See `Linux kernel docs`_ for further explanations."},{"line_number":793,"context_line":""},{"line_number":794,"context_line":".. _`Linux kernel docs`: https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt"},{"line_number":795,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"18c45892_85b1d582","line":792,"range":{"start_line":792,"start_character":8,"end_line":792,"end_character":23},"updated":"2023-02-23 10:59:04.000000000","message":"more power consuming then schedutil but more efficient then performance","commit_id":"9d70eb54ef1b77c04cdef725e5a5992df7dd3506"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"bfc1f486e232a3a5180edc2d8191634d4e1c686d","unresolved":true,"context_lines":[{"line_number":789,"context_line":"   the high-power state (if the kernel supports it) as the CPU frequency is"},{"line_number":790,"context_line":"   dynamically set based on CPU task states. Other governors may be worth to"},{"line_number":791,"context_line":"   be tested, including ``conservative`` and ``ondemand`` which are quite a bit"},{"line_number":792,"context_line":"   more power consuming. See `Linux kernel docs`_ for further explanations."},{"line_number":793,"context_line":""},{"line_number":794,"context_line":".. _`Linux kernel docs`: https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt"},{"line_number":795,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"2d4f5267_68c2065e","line":792,"range":{"start_line":792,"start_character":8,"end_line":792,"end_character":23},"in_reply_to":"18c45892_85b1d582","updated":"2023-02-23 11:03:18.000000000","message":"Good call, respining.","commit_id":"9d70eb54ef1b77c04cdef725e5a5992df7dd3506"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"cfd899a91aba9df348d5f976acfcaa878dd646e4","unresolved":false,"context_lines":[{"line_number":789,"context_line":"   the high-power state (if the kernel supports it) as the CPU frequency is"},{"line_number":790,"context_line":"   dynamically set based on CPU task states. Other governors may be worth to"},{"line_number":791,"context_line":"   be tested, including ``conservative`` and ``ondemand`` which are quite a bit"},{"line_number":792,"context_line":"   more power consuming. See `Linux kernel docs`_ for further explanations."},{"line_number":793,"context_line":""},{"line_number":794,"context_line":".. _`Linux kernel docs`: https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt"},{"line_number":795,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"f61abb11_9d3416f6","line":792,"range":{"start_line":792,"start_character":8,"end_line":792,"end_character":23},"in_reply_to":"2d4f5267_68c2065e","updated":"2023-02-23 11:06:08.000000000","message":"Done","commit_id":"9d70eb54ef1b77c04cdef725e5a5992df7dd3506"}]}