)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"76eec48edf55bcbda16d49eb3ff7f4ce061390fb","unresolved":true,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Signed-off-by: Tyler Stachecki \u003ctstachecki@bloomberg.net\u003e"},{"line_number":27,"context_line":"Change-Id: I9e4973e2b88af9fe29378878e00873ab4004adcc"},{"line_number":28,"context_line":"Closes-Bug: 2011127"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"9dac382b_96beed5a","line":28,"updated":"2023-03-10 22:20:00.000000000","message":"this would be a feature not a bug which is the main reason for the -1\n\nthe current behavior was intoduced by\nhttps://github.com/openstack/nova/commit/e5ba8494374a1b049eae257fe05b10c5804049ae\n\nwe could change how we do the normlaistaion potitelly but we already do normalistaion on a per wigher basis\n\nhttps://github.com/openstack/nova/blob/373be3db5b7b058767ddac50ff1367725c932a84/nova/weights.py#L126-L172\n\nyou want to change from doing the normalisation on a per weigher basis to returnign prenormalised values.\n\neither could be done but they will have very differnt behaivors.\nwe likely woudl have to have to make this a configurable behavior\nif we were to change this as people have come to rely on the current behavior.\n \ngiven this is a feature and not a bug this would also requrie a design spec or at teh minium a specless blueprint and would need wider agreement on the design.\n \na specless blueprint can be approvced a the nove team meeting but i would suggest adding this as a topic to the upcominig nova ptg to get wider attention.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":33910,"name":"Tyler Stachecki","display_name":"Tyler Stachecki","email":"tstachecki@bloomberg.net","username":"tjstachecki"},"change_message_id":"8b16c0a964d8d70e30114fd303375d3898ee056f","unresolved":true,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Signed-off-by: Tyler Stachecki \u003ctstachecki@bloomberg.net\u003e"},{"line_number":27,"context_line":"Change-Id: I9e4973e2b88af9fe29378878e00873ab4004adcc"},{"line_number":28,"context_line":"Closes-Bug: 2011127"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"fd02f913_87c2180a","line":28,"in_reply_to":"244409d8_5c4dfe77","updated":"2023-03-10 22:45:12.000000000","message":"Interesting suggestion - especially in the case of CPUs, that is a good idea from a power/performance tuning prospective.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":33910,"name":"Tyler Stachecki","display_name":"Tyler Stachecki","email":"tstachecki@bloomberg.net","username":"tjstachecki"},"change_message_id":"0bdc250f83631100c4bc1190e622487d8c1f0f51","unresolved":true,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Signed-off-by: Tyler Stachecki \u003ctstachecki@bloomberg.net\u003e"},{"line_number":27,"context_line":"Change-Id: I9e4973e2b88af9fe29378878e00873ab4004adcc"},{"line_number":28,"context_line":"Closes-Bug: 2011127"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"a9484142_d452f209","line":28,"in_reply_to":"9dac382b_96beed5a","updated":"2023-03-10 22:25:58.000000000","message":"OK - this all makes a lot more sense in that context and I\u0027ll take that as an action item. Apologies for my unfamiliarity with the process around how something like this should first get tabled.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"d9438e0eda0e23f7e7a2413537821ece562e6e72","unresolved":true,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Signed-off-by: Tyler Stachecki \u003ctstachecki@bloomberg.net\u003e"},{"line_number":27,"context_line":"Change-Id: I9e4973e2b88af9fe29378878e00873ab4004adcc"},{"line_number":28,"context_line":"Closes-Bug: 2011127"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"244409d8_5c4dfe77","line":28,"in_reply_to":"a9484142_d452f209","updated":"2023-03-10 22:38:20.000000000","message":"dont worry about it we dont expect everyone to knwo what process to follow.\n\none other thing that occured to me is instead of changing the exsitng ram/cpu/disk\nweighers we coudl add a second set of ram/cpu/disk weither that have your per host normalised behvioer.\n\nthat woudl enabel you to have both the current (Absolute) ram weigher and new (relitive) ram weigher enabeld so that if you have 2 host at 50% ram usage you could prefer to then select teh host with the most overall ram.\n\ni think i personally preer that approch as its more flexible then a config value to choose between each behavior.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"}],"/PATCHSET_LEVEL":[{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"225bebd8c4bee9c15f9f9c2f399cc9972ee56f00","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"a1d62846_3d8498c8","updated":"2023-03-10 22:57:42.000000000","message":"Agree we should not just change this default, release boundary or not. Some people want to stack and some want to spread (most seem to be in the former camp, IME, due to power and cooling requirements). It\u0027s fine to opt into spreading, but it\u0027s not okay to just push this opinion on everyone by default this late in the game.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"58bd9f172456b356eef503d7de2d3695616eb906","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"cb49d255_4ae8c97a","updated":"2023-03-10 22:06:54.000000000","message":"this would be a non trivial behavior change\n\nalso you can invert the behavior and choose stacking vs spread already uing the multiplpers\n\nhttps://docs.openstack.org/nova/latest/configuration/config.html#filter_scheduler.ram_weight_multiplier\n\n\nRAM weight multipler ratio.\n\nThis option determines how hosts with more or less available RAM are weighed. A positive value will result in the scheduler preferring hosts with more available RAM, and a negative number will result in the scheduler preferring hosts with less available RAM. Another way to look at it is that positive values for this option will tend to spread instances across many hosts, while negative values will tend to fill up (stack) hosts as much as possible before scheduling to a less-used host. The absolute value, whether positive or negative, controls how strong the RAM weigher is relative to other weighers.\n\nNote that this setting only affects scheduling if the RAMWeigher weigher is enabled.\n\nPossible values:\n\n    An integer or float value, where the value corresponds to the multipler ratio for this weigher.\n\n\nthe ram wiegher already prefers host with more aviable ram and spreads by default\n","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"bd6130451df0ed50409f8b203d1168b80901be82","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"5f80c7af_a12adaa2","in_reply_to":"3058d79c_2ca61d61","updated":"2023-03-10 22:29:33.000000000","message":"we did that because it allows upgrading without any change in bahviaor and gave use a release to get operator feedback and then we made the new spreading behavior the default.\n\ni think we likely shoudl do the same here with a per_host_mormalistion config option \nfor and we can either default to the old or new behviaor on code review.\n\nfor power savign reason there those that want to fully pack host before they move on to the next host for those that want better performance your method woudl be better.\n\nwe have peopel in both camps.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":33910,"name":"Tyler Stachecki","display_name":"Tyler Stachecki","email":"tstachecki@bloomberg.net","username":"tjstachecki"},"change_message_id":"bc64a303935ccf1606cb8fb4ef1ac36b7999e3c9","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"c4880569_645254e4","in_reply_to":"5a365f7a_5fe7a739","updated":"2023-03-10 22:16:57.000000000","message":"Sorry, should read:\n\n\"The problem this commit is solving is how the allocations for a *specific* weigher are spread across hypervisors, not how the weighers are considered in relation to each other.\"","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"0e7dcc1beead32c07b76dcaeed371510b7428729","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"3058d79c_2ca61d61","in_reply_to":"c4880569_645254e4","updated":"2023-03-10 22:25:28.000000000","message":"yep looking at the code change you are doing i understand the problem you are trying to solve.\n\nright now we intentioally spread based on the absultoe amount fo ram.\ne.g we aim to bring all host to the same amount fo free ram rahter then bring all host to the same amoutn of relitive free ram.\n\n\nim not nessisarly agaisnt that change but given we have maintained the same behavior for the last 10 years i dont think we can make the change form absolut to relivite spreading/packing without makeing it configurable however we likely can default to the relitve behavior if other agree that is preferable.\n\ni suspect we will want a spec to descirbe the motivation and explain why the new behavior should become the default. if peopel think there is not a need to maintain the old behavior then we might eb able to make this chagne without a config option but i think the upgrade impact will be a concern and we will want to treat this like\n\nhttps://docs.openstack.org/nova/latest/configuration/config.html#compute.packing_host_numa_cells_allocation_strategy\n\nwhich we intoduced maintianing the old behavior and fliped the value to the spreading behavior in the next release.","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"},{"author":{"_account_id":33910,"name":"Tyler Stachecki","display_name":"Tyler Stachecki","email":"tstachecki@bloomberg.net","username":"tjstachecki"},"change_message_id":"0ebd0b4fef61fcfe07d8c8ed502b469e4ec75caf","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"5a365f7a_5fe7a739","in_reply_to":"cb49d255_4ae8c97a","updated":"2023-03-10 22:14:41.000000000","message":"Thanks for the super-fast review!\n\nI understand this is not a trivial change and that it\u0027s going to impact scheduling considerably. If it needs to be gated by a switch or wait until a release bump, that\u0027s OK -- though I think the change is very much warranted.\n\nre: the multipler -- I\u0027m aware of its presence... this commit is doing something different. The problem this commit is not solving is how the allocations for a *specific* weigher are spread across hypervisors, not how the weighers are considered in relation to each other.\n\nIOW your suggestion (changing the ratios) impacts the w*_multiplier values here:\n\"weight \u003d w1_multiplier * norm(w1) + w2_multiplier * norm(w2) + ...\"\n\nwhereas this commit is changing how the normalization function works.\n\nThere\u0027s a detailed writeup in the LP with a very specific example:\nhttps://bugs.launchpad.net/nova/+bug/2011127","commit_id":"1a5c0bad095392a0654037c5f6bd1d22aceda9e6"}]}