)]}'
{"specs/kilo/merge-compute-and-central-agents.rst":[{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c27055be9efe29909295b6c7dd2c3878e8af4e0f","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"fa98f980_9558c258","line":21,"updated":"2014-09-19 12:12:27.000000000","message":"So is it actually the *pipeline* configuration that would need to differ, or the stevedore namespace used to load the pollster and discovery extensions?\n\nHaving the same pipeline.yaml config used for both roles should be fine (and more convenient) if the difference in roles is captured by loading different extensions.\n\nOn an all-in-one style deployment (e.g. a devstack or small PoC), would you envisage two processes running the same code but with differing config, or the *same* process doing double-duty handling both the compute and central agent roles?\n\nHow would you handle the fact that the compute uses an agent-level default discovery (local_instances) whereas the central agent relies on source-level and pollster-default discovery?","commit_id":"66fd715588254127b73febf562fef21b81bda280"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"be30d38c58fabca8ccf814cc2aa548021ec3bd9e","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"fa98f980_c0a9a625","line":21,"in_reply_to":"fa98f980_9558c258","updated":"2014-09-19 13:06:57.000000000","message":"All these agents will have just different meters to collect in the pipeline (with needed sink, of course), so like we\u0027ll list compute meters on the compute node and meters we use to set for the central agent for the controller node.\n\nFor all in one deployment I\u0027ll propose to have one agent collecting all the meters (like let\u0027s collect \"*\")\n\n\n\u003e\u003e\u003e How would you handle the fact that the compute uses an agent-level default discovery (local_instances) whereas the central agent relies on source-level and pollster-default discovery?\n\n - That\u0027s, actually, really good question and I\u0027d love to know your opinion here, folks. I would say that compute case should be somehow transformed to one of the source-level and pollster-default - source level looks better omho logically. Plus thinking about compute agent discovery, I don\u0027t actually like the idea of polling nova api to get this list, I guess communication directly with the hypervisor will be better in future.","commit_id":"66fd715588254127b73febf562fef21b81bda280"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c27055be9efe29909295b6c7dd2c3878e8af4e0f","unresolved":false,"context_lines":[{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":25,"context_line":""},{"line_number":26,"context_line":"We have logic mismatch with two separated agents, although they might be"},{"line_number":27,"context_line":"part of one peace of code, configured via pipeline."},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"Proposed change"},{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"fa98f980_f5bbce72","line":27,"updated":"2014-09-19 12:12:27.000000000","message":"Non-blocking nitpick:\n\n  s/one peace of code/one piece of code/","commit_id":"66fd715588254127b73febf562fef21b81bda280"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"be30d38c58fabca8ccf814cc2aa548021ec3bd9e","unresolved":false,"context_lines":[{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":25,"context_line":""},{"line_number":26,"context_line":"We have logic mismatch with two separated agents, although they might be"},{"line_number":27,"context_line":"part of one peace of code, configured via pipeline."},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"Proposed change"},{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"fa98f980_e0a4ea4c","line":27,"in_reply_to":"fa98f980_f5bbce72","updated":"2014-09-19 13:06:57.000000000","message":"will fix in new patch set","commit_id":"66fd715588254127b73febf562fef21b81bda280"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_ed52dedc","line":33,"updated":"2014-10-06 19:41:21.000000000","message":"\u0027unified wrappers\u0027 is ambiguous... can you clarify?","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_8bf050ff","line":33,"in_reply_to":"baa201ad_ed52dedc","updated":"2014-10-07 08:17:45.000000000","message":"Heh, I would say that\u0027s my wrong phrase. i meant not the wrappers, but unified mechanism, will change it in next patch set.","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_6d97ce3a","line":34,"updated":"2014-10-06 19:41:21.000000000","message":"can you also clarify how \u0027Pollsters will be selected? will we still blindly poll based on whatever service is configured in keystone? if i have an agent on compute node.. i want it only to do compute polling... how do i make sure it doesn\u0027t try to do any other polling...\n\nie. two compute nodes, each will require an agent to poll locally... but what happens to non-compute pollsters? are they disabled?","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"cbdafc65f01b74859db613644c974b935e054744","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_bedf09d5","line":34,"in_reply_to":"baa201ad_6b2ecc44","updated":"2014-10-07 13:45:08.000000000","message":"cool cool. yeah that was my concern as well: how to make filtering logical and simple... i don\u0027t really have any suggestions right now.","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_6b2ecc44","line":34,"in_reply_to":"baa201ad_6d97ce3a","updated":"2014-10-07 08:17:45.000000000","message":"I was thinking about something like this: agent is using pipeline (as now) to understand what meters to poll, so in my mind everything will be defined by pipeline. The only moment I\u0027m kind of concerned, how to make the pipeline not so huge potentially - as now compute agent and central agent have different stevedore namespace, if they\u0027ll have one (how I did in the POC) if you\u0027ll say meters\u003d* in the pipeline it\u0027ll be all compute and central meters.. I don\u0027t know yet how to say in short words for now \u0027all non-compute metres in this case\u0027","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"0b29139c213d7470d078bcd12cfe896a63f60c89","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"wrappers we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"},{"line_number":37,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_070c3393","line":34,"in_reply_to":"baa201ad_bedf09d5","updated":"2014-10-08 07:43:52.000000000","message":"I have kind of idea to keep two stevedore namespaces for the polling agent, and check if they were set in the pipeline - although, this \u0027two namespaces idea\u0027 does not seem elegant tot me...","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":65,"context_line":"Performance/Scalability Impacts"},{"line_number":66,"context_line":"-------------------------------"},{"line_number":67,"context_line":""},{"line_number":68,"context_line":"None"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Other deployer impact"},{"line_number":71,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_6ddb4e4f","line":68,"updated":"2014-10-06 19:41:21.000000000","message":"possibly better memory usage? or at least not worse?","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":65,"context_line":"Performance/Scalability Impacts"},{"line_number":66,"context_line":"-------------------------------"},{"line_number":67,"context_line":""},{"line_number":68,"context_line":"None"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Other deployer impact"},{"line_number":71,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_eb21dc4d","line":68,"in_reply_to":"baa201ad_6ddb4e4f","updated":"2014-10-07 08:17:45.000000000","message":"Hmm, for now I\u0027m not sure it\u0027ll be true for real installations... Although for the all-in-one variant it\u0027ll work better, yeah. I\u0027ll add this infer here in next patch set. Thanks!","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":71,"context_line":"---------------------"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"* It\u0027ll be less services Ceilometer will run on node (if it\u0027ll be all-in-one"},{"line_number":74,"context_line":"  installation) and we\u0027ll have more unified way of Ceilometer installation."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"Developer impact"},{"line_number":77,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_edc57e5c","line":74,"updated":"2014-10-06 19:41:21.000000000","message":"is it still advisable to have two polling agents?","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":71,"context_line":"---------------------"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"* It\u0027ll be less services Ceilometer will run on node (if it\u0027ll be all-in-one"},{"line_number":74,"context_line":"  installation) and we\u0027ll have more unified way of Ceilometer installation."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"Developer impact"},{"line_number":77,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_8bc13056","line":74,"in_reply_to":"baa201ad_edc57e5c","updated":"2014-10-07 08:17:45.000000000","message":"Yeah, for sure - you will have the opportunity to run any number of agents you\u0027d love to (on one node or on the many ones) - the workload partitioning mechanism will be the same actually.","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":92,"context_line":"----------"},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Merge central agent code into the current compute one (probably, this new"},{"line_number":95,"context_line":"  agent will be called just *agent* or something)."},{"line_number":96,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":97,"context_line":"* Provide Devstack with new agents scheme support (for all-in-ane and multinode"},{"line_number":98,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_cdf7dad1","line":95,"updated":"2014-10-06 19:41:21.000000000","message":"can we try to propose a name for new/merged service in this bp rather than in code?\n\ni guess something like ceilometer-poll-agent? or something...","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"cbdafc65f01b74859db613644c974b935e054744","unresolved":false,"context_lines":[{"line_number":92,"context_line":"----------"},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Merge central agent code into the current compute one (probably, this new"},{"line_number":95,"context_line":"  agent will be called just *agent* or something)."},{"line_number":96,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":97,"context_line":"* Provide Devstack with new agents scheme support (for all-in-ane and multinode"},{"line_number":98,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_de63ed3f","line":95,"in_reply_to":"baa201ad_4bc3c856","updated":"2014-10-07 13:45:08.000000000","message":"i\u0027m cool with ceilometer-polling agent. :)","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":92,"context_line":"----------"},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"* Merge central agent code into the current compute one (probably, this new"},{"line_number":95,"context_line":"  agent will be called just *agent* or something)."},{"line_number":96,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":97,"context_line":"* Provide Devstack with new agents scheme support (for all-in-ane and multinode"},{"line_number":98,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_4bc3c856","line":95,"in_reply_to":"baa201ad_cdf7dad1","updated":"2014-10-07 08:17:45.000000000","message":"In my POC I\u0027ve named it ceilometer-polling-agent, but I can change this to any variant you\u0027d like to :)","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3fa7754d550f11e9a455ee4afe5f76ff0f06b273","unresolved":false,"context_lines":[{"line_number":94,"context_line":"* Merge central agent code into the current compute one (probably, this new"},{"line_number":95,"context_line":"  agent will be called just *agent* or something)."},{"line_number":96,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":97,"context_line":"* Provide Devstack with new agents scheme support (for all-in-ane and multinode"},{"line_number":98,"context_line":"  installations)."},{"line_number":99,"context_line":"* Investigate Grenade tests aspect."},{"line_number":100,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_2d2ea643","line":97,"updated":"2014-10-06 19:41:21.000000000","message":"s/all-in-ane/all-in-one/","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"1a38de261df1ffe518de1a3f6182762c74376e9d","unresolved":false,"context_lines":[{"line_number":94,"context_line":"* Merge central agent code into the current compute one (probably, this new"},{"line_number":95,"context_line":"  agent will be called just *agent* or something)."},{"line_number":96,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":97,"context_line":"* Provide Devstack with new agents scheme support (for all-in-ane and multinode"},{"line_number":98,"context_line":"  installations)."},{"line_number":99,"context_line":"* Investigate Grenade tests aspect."},{"line_number":100,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"baa201ad_8ba8508e","line":97,"in_reply_to":"baa201ad_2d2ea643","updated":"2014-10-07 08:17:45.000000000","message":"a-ha, will fix it! thanks!","commit_id":"ae3ac8134e074b0fc565e7f1452c4657bc245307"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"8eb4700347af81f72971d331fa57de21bbd489b5","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_d2b75d38","line":21,"updated":"2014-11-25 12:00:35.000000000","message":"I think I pointed out in a comment on an earlier patcheset that it\u0027s more the discovery and/or pollster loading that needs to differ between the two agents, rather than the pipeline.\n\ni.e. the ceilometer-polling-agent running on a compute node (in place of the old compute agent) and the ceilometer-polling-agent running on a controller node (in place of the old central agent) could both use the same pipeline.yaml, as long as they are somehow specialized to only use specific discoveries relevant to the role that the agent is playing (local_instances if replacing a comoute agent, versus the set of discoveries used by the central agent\u0027s pollsters)\n\nAlso, wouldn\u0027t we need some way of ensuring that the ceilometer-polling-agent only joins the appropriate partitioning groups?","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"794d701997c195e5e2cc861e6c494826a952b539","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_29274fcb","line":21,"in_reply_to":"5a890539_6949575f","updated":"2014-11-25 14:36:17.000000000","message":"BTW the reason that the CLI command line args approach might be preferable is that it would allow the new polling agent to masquerade as any of the existing agents without any changes to the pipeline.yaml or ceilometer.conf (currently both of these config files can be shared between agents without any change).","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"b6a17224091d466fea3e8ba5c54900253be295e4","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_a3236103","line":21,"in_reply_to":"5a890539_a8d89eef","updated":"2014-11-25 13:27:23.000000000","message":"Eoghan, also (as said to Gordon) there is some variant with the pipeline module minimal modification, allowing to consume aliases on the pipeline like: \"meters: compute.*\" knowing that compute.* is equal to disk.*, cpu, cpu_util, network.incoming.*, network.outgoing.*, instance, instance_flavor and memory usage. The similar list of wildcards will be defined for the central.* as well. This will allow to set this thing still in the pipeline with future opportunity of dynamic configuration via REST api.","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"47ca8a4f04d6b775dee5fb5fe688939f917a8ecb","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_6949575f","line":21,"in_reply_to":"5a890539_a8d89eef","updated":"2014-11-25 14:27:36.000000000","message":"OK, thinking through the problem a bit more, I think part of the confusion is that we already have role-seperation for pollster loading, i.e. the distinction between the \u0027ceilometer.poll.compute\u0027 and \u0027ceilometer.poll.central\u0027 namespaces in the setup.cfg, whereas we just have a single \u0027ceilometer.discover\u0027 namespace for discovery extensions.\n\nLoading all discovery extensions into all agents works currently since:\n\n* a per-manager discovery extension is selected by the manager itself\n* a per-pollster discovery extension is selected by the corresponding pollster\n* a per-source discovery extension is selected in the pipeline.yaml\n\nSo in order to keep it close to the current concept, what if we introduce something like the following new per-agent CLI script config variable:\n\n  --pollster_namespace\u003d{compute|central|ipmi|compute,central,ipmi}\n\nthen this config setting would replace the corresponding init parameter in:\n\n  https://github.com/openstack/ceilometer/blob/master/ceilometer/agent.py#L139\n\nSo for example, a pure compute agent could be run as:\n\n  $ ceilometer-polling-agent --pollster_namespace\u003dcompute \\\n       --logfile /var/log/ceilometer/compute.log\n\nwhereas a pure central agent could be run via:\n\n  $ ceilometer-polling-agent --pollster_namespace\u003dcentral \\\n       --logfile /var/log/ceilometer/central.log\n\nand a single agent to rule them all:\n\n  $ ceilometer-polling-agent --pollster_namespace\u003dcentral,compute,ipmi \\\n       --logfile /var/log/ceilometer/polling.log\n\nThis assumes that we go with the approach to get of the need for the local_instances default-discovery, as mooted in:\n\n  https://review.openstack.org/127185","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"22c9488648ebbb821f4c85ba98afc8fd6716f0df","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_ed42b8a9","line":21,"in_reply_to":"5a890539_d2b75d38","updated":"2014-11-25 12:43:24.000000000","message":"a-ha, Eoghan, I finally got your point. I suppose I need some researches here once more. I thing this discovery thing should looks better here, thanks!","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"78602be8bee2d07f843b0f26d62605010e588924","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"It looks logical to merge the code of agents back, so they would differ only"},{"line_number":20,"context_line":"in the pipeline (like agent running on the compute nodes collect only compute"},{"line_number":21,"context_line":"meters, on cloud controller - any other set of meters)."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_a8d89eef","line":21,"in_reply_to":"5a890539_ed42b8a9","updated":"2014-11-25 13:09:23.000000000","message":"Eoghan, imho we can define this discovery specialisation only in several ways. The problem is that discovery is resource-based, so if we know we need to poll hardware.*, we know we need to use one of the ceilometer.hardware.pollsters, and this pollster contains the discovery to use. Basically we know about what to poll (via pipeline - and it might be even \u0027*\u0027) before knowing how and where to poll from. That\u0027s why it\u0027s much easier to have the separation of the pipeline layer, not later. If we do not have this pipeline-level separation, we need some other trigger to use. It might be configuration option in the ceilometer.conf, but this means it\u0027ll be completely static thing (comparing with the fact we\u0027ll probably have dynamic pipeline in future). Probably (as I answered Gordon) there is the way of possibility of having some pipeline field like \u0027roles \u003d compute\u0027 or \u0027roles \u003d central\u0027 or without it (meaning both) separated from sources and sinks. Both of these solutions (roles in the pipeline or in the conf) look kind of ugly to me.... Do you have some ideas here? Any help will be really appreciated :)","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"28d2f231f410ab18e6088877e9d167b8a4325b3a","unresolved":false,"context_lines":[{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_0bdf503c","line":33,"updated":"2014-11-13 09:40:52.000000000","message":"i think we should explicitly add how Pollster selection is going to work (see discussion in prevoius patch).\n\ni think we should just say the polling agent will poll everything it can discover... if only compute meters are required, it should be configured as such in the pipeline for that specific agent.","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"4ace290816c344545b27d00e252dbf5ef8fa6ce2","unresolved":false,"context_lines":[{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"We need to merge carefully code of compute and central agents, using unified"},{"line_number":33,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Alternatives"},{"line_number":36,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_6d008866","line":33,"in_reply_to":"5a890539_0bdf503c","updated":"2014-11-25 12:41:44.000000000","message":"Gordon, finally after some experiments, I see no way except the following one:\ndefine if agent is compute or central, or both in the ceilometer.conf... Or there is some variant with the pipeline module modification, allowing to consume aliases on the pipeline like: \"meters: compute.*\" knowing that compute.* is equal to disk.*, cpu, cpu_util, network.incoming.*, network.outgoing.*, instance, instance_flavor and memory usage. The similar list of wildcards will be defined for the central.* as well. This will allow to set this thing still in the pipeline with future opportunity of dynamic configuration via REST api. For me this tricky variant with pipeline aliases looks better than conf file, as this will allow in future to modify this.","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"28d2f231f410ab18e6088877e9d167b8a4325b3a","unresolved":false,"context_lines":[{"line_number":121,"context_line":"Documentation Impact"},{"line_number":122,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":123,"context_line":""},{"line_number":124,"context_line":"We\u0027ll need to rewrite our installation guide and common documentation parts"},{"line_number":125,"context_line":"with the information about one agent instead of two."},{"line_number":126,"context_line":""},{"line_number":127,"context_line":""}],"source_content_type":"text/x-rst","patch_set":9,"id":"5a890539_eb8cc437","line":124,"updated":"2014-11-13 09:40:52.000000000","message":"add explicitly we need to give examples on how to configure discovery so we don\u0027t poll all when we don\u0027t want to.","commit_id":"40f22fb3c5c25a558309c497b02319864b871c63"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"e9fb0a7c5c61c2836be6d69c5dac5d58fa61acda","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Also we need to save the opportunity of setting up agent to poll only compute"},{"line_number":39,"context_line":"or only meters used to be polled by central agent. Basically all-in-one"},{"line_number":40,"context_line":"installation will be satisfied with agent polling all available meters, but"},{"line_number":41,"context_line":"production installations need this separation. This will be achieved by adding"},{"line_number":42,"context_line":"additional per-agent CLI script parameter:"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_5586ae17","line":39,"updated":"2014-11-27 08:38:37.000000000","message":"what happens if we run an all-in-one polling agent which turns on the coordination.backend_url feature? Does this means the compute agent resources(VM instances) will be partitioned? I don\u0027t think the current compute pollsters can polling remote vm instances.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"50d420f6e310b69a97bd4e254f420fbb7a7def1e","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Also we need to save the opportunity of setting up agent to poll only compute"},{"line_number":39,"context_line":"or only meters used to be polled by central agent. Basically all-in-one"},{"line_number":40,"context_line":"installation will be satisfied with agent polling all available meters, but"},{"line_number":41,"context_line":"production installations need this separation. This will be achieved by adding"},{"line_number":42,"context_line":"additional per-agent CLI script parameter:"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_da29e638","line":39,"in_reply_to":"5a890539_5586ae17","updated":"2014-11-27 10:41:57.000000000","message":"actually it can if hypervisor uri can be set to the remote one - libvirt, for instance, support this feature for sure (basically I\u0027m not sure about others)","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":1669,"name":"Julien Danjou","display_name":"jd","email":"julien@danjou.info","username":"jdanjou"},"change_message_id":"09b9644c316201dc4d422b07902d6e428c45db9a","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Also we need to save the opportunity of setting up agent to poll only compute"},{"line_number":39,"context_line":"or only meters used to be polled by central agent. Basically all-in-one"},{"line_number":40,"context_line":"installation will be satisfied with agent polling all available meters, but"},{"line_number":41,"context_line":"production installations need this separation. This will be achieved by adding"},{"line_number":42,"context_line":"additional per-agent CLI script parameter:"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_987cc6ed","line":39,"in_reply_to":"5a890539_da29e638","updated":"2014-11-27 11:43:50.000000000","message":"+1 with Dina on this.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"51f37ccbf44bbd170803d3c36a940bd8f2f5f343","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Also we need to save the opportunity of setting up agent to poll only compute"},{"line_number":39,"context_line":"or only meters used to be polled by central agent. Basically all-in-one"},{"line_number":40,"context_line":"installation will be satisfied with agent polling all available meters, but"},{"line_number":41,"context_line":"production installations need this separation. This will be achieved by adding"},{"line_number":42,"context_line":"additional per-agent CLI script parameter:"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_fac859aa","line":39,"in_reply_to":"5a890539_da29e638","updated":"2014-11-27 15:12:20.000000000","message":"yes, it supports that, but that means we need to modify the compute pollster and virt driver. Better to emphasize that here.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"dd06177297c73028ef437c81dad65a6cac8c23b3","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mechanism we already have to process discovery and polling processes."},{"line_number":37,"context_line":""},{"line_number":38,"context_line":"Also we need to save the opportunity of setting up agent to poll only compute"},{"line_number":39,"context_line":"or only meters used to be polled by central agent. Basically all-in-one"},{"line_number":40,"context_line":"installation will be satisfied with agent polling all available meters, but"},{"line_number":41,"context_line":"production installations need this separation. This will be achieved by adding"},{"line_number":42,"context_line":"additional per-agent CLI script parameter:"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_df338260","line":39,"in_reply_to":"5a890539_fac859aa","updated":"2014-11-28 07:47:28.000000000","message":"Lanhao, how should we change compute pollster? It\u0027s just other type of URI to be set in conf file, nothing more imho... Do I miss something?","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"e9fb0a7c5c61c2836be6d69c5dac5d58fa61acda","unresolved":false,"context_lines":[{"line_number":42,"context_line":"additional per-agent CLI script parameter:"},{"line_number":43,"context_line":"*--pollster-namespace\u003d{compute|central|compute,central}*"},{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_956c96d2","line":45,"updated":"2014-11-27 08:38:37.000000000","message":"yes, ipmi agent is very like compute agent in deployment perspective, could be added too.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"50d420f6e310b69a97bd4e254f420fbb7a7def1e","unresolved":false,"context_lines":[{"line_number":42,"context_line":"additional per-agent CLI script parameter:"},{"line_number":43,"context_line":"*--pollster-namespace\u003d{compute|central|compute,central}*"},{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_badb3a4e","line":45,"in_reply_to":"5a890539_956c96d2","updated":"2014-11-27 10:41:57.000000000","message":"yes, I just thought it\u0027ll be good to start with most difficult part of central+compute agents merge","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"e9fb0a7c5c61c2836be6d69c5dac5d58fa61acda","unresolved":false,"context_lines":[{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_7502ca0a","line":47,"updated":"2014-11-27 08:38:37.000000000","message":"Maybe name it ceilometer-agent-polling? Just following the existing naming convention.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"50d420f6e310b69a97bd4e254f420fbb7a7def1e","unresolved":false,"context_lines":[{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_ba97ba39","line":47,"in_reply_to":"5a890539_7502ca0a","updated":"2014-11-27 10:41:57.000000000","message":"I was thinking about it, but this word order sound too strange for me. If we\u0027ll merge ipmi agent as well here, it\u0027ll be left only ceilometer-agent-notification here, so it won\u0027t look like a convention. Basically I\u0027m not really against your variant, I joust wonder if we really need this wrong order anymore :) Please let me know if you think I really should change this name, I\u0027m ok with it.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":1669,"name":"Julien Danjou","display_name":"jd","email":"julien@danjou.info","username":"jdanjou"},"change_message_id":"09b9644c316201dc4d422b07902d6e428c45db9a","unresolved":false,"context_lines":[{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_d85d6e7b","line":47,"in_reply_to":"5a890539_ba97ba39","updated":"2014-11-27 11:43:50.000000000","message":"Let\u0027s go with some without agent at all, like \"ceilometer-polling\"?\n\nI think reusing pollster would be confusing with the plugins themselves.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"dd06177297c73028ef437c81dad65a6cac8c23b3","unresolved":false,"context_lines":[{"line_number":44,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":45,"context_line":"case running of polling agent will look like following:"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"    ceilometer-polling-agent --pollster-namespace\u003dcentral,compute --logfile /var/log/ceilometer/polling.log"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_5ff07291","line":47,"in_reply_to":"5a890539_d85d6e7b","updated":"2014-11-28 07:47:28.000000000","message":"it might be called \u0027polling-namespace\u0027, for instance. I\u0027ll change it if we\u0027ll solve next comment","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"e9fb0a7c5c61c2836be6d69c5dac5d58fa61acda","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_2a537d3a","line":51,"updated":"2014-11-27 08:38:37.000000000","message":"One question, how to use stevedore to load extensions from multiple namespaces, use multiple ExtensionManager?","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"50d420f6e310b69a97bd4e254f420fbb7a7def1e","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_5a1b36ca","line":51,"in_reply_to":"5a890539_2a537d3a","updated":"2014-11-27 10:41:57.000000000","message":"actually we need just to load extensions themselves, not the managers. We use self.pollster_manager in only in https://github.com/openstack/ceilometer/blob/master/ceilometer/agent.py mode and only to grab extensions here: https://github.com/openstack/ceilometer/blob/master/ceilometer/agent.py#L180 - in my POC I removed storing polling managers, but store only iterator created from all extensions that could be found from different namespaces, and that\u0027s it.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":1669,"name":"Julien Danjou","display_name":"jd","email":"julien@danjou.info","username":"jdanjou"},"change_message_id":"09b9644c316201dc4d422b07902d6e428c45db9a","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_98bf6660","line":51,"in_reply_to":"5a890539_5a1b36ca","updated":"2014-11-27 11:43:50.000000000","message":"I don\u0027t think it\u0027s a good idea to keep different namespaces. If you have to pass something, it\u0027s a subset/a list of pollster, not a namespace.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":1669,"name":"Julien Danjou","display_name":"jd","email":"julien@danjou.info","username":"jdanjou"},"change_message_id":"4c100b4dc7feb1be9b21bda1129acc6bdff8f9ef","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_a8fd993d","line":51,"in_reply_to":"5a890539_5f75d2f8","updated":"2014-11-28 11:02:50.000000000","message":"I\u0027d say AND.\n\nI don\u0027t think we should keep separate namespaces, _but_ I\u0027m inclined to say that it\u0027s not a problem to solve in this spec. So just amend this spec saying you could use both a list of pollster or a (deprecated) pollster namespace and we\u0027re good.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"dd06177297c73028ef437c81dad65a6cac8c23b3","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_bf079680","line":51,"in_reply_to":"5a890539_98bf6660","updated":"2014-11-28 07:47:28.000000000","message":"Julien, thanks for the review! This solution was something agreed with Eoghan because of following question: how to separate effectively compute and central metrics on the polling level to keep some simple opportunity to run polling agent behaving lie a central and polling agent behaving like a compute one in real deployment? We need something simple and effective at the same time... If we\u0027ll pass list of pollsters via cmd it\u0027ll be kind of ugly to see command like this:\nceilometer-polling --pollsters\u003dip.floating image image.size storage.* energy power switch.* hardware.* network.services.* --logfile /var/log/ceilometer/polling.log - especially if we have simple situation like \u0027poll everything related to the ex-central agent\u0027. Julien, what\u0027s the reason you don\u0027t like to have two namespaces here? If you think it\u0027ll be comfortable to pass list of pollsters, I may just add this functionality here, but I see no harm actually to have two namespaces.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"b753125b60dcea83b9a24d9c2e17ab58dd5917c1","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":50,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"},{"line_number":51,"context_line":"*ceilometer.poll.compute* or both) with possible pollsters to load."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"Alternatives"},{"line_number":54,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_5f75d2f8","line":51,"in_reply_to":"5a890539_bf079680","updated":"2014-11-28 07:55:22.000000000","message":"Just to clarify what I said (I suppose I was vague a little bit): I may keep opportunity to define what to poll via \"polling namespace\" and via \"pollster list\" at the same time in this impl, if it\u0027s ok to keep two namespaces. The only question is what to do if both of these params are set, and which of them has the priority - like should we poll logical OR or AND here?","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"e9fb0a7c5c61c2836be6d69c5dac5d58fa61acda","unresolved":false,"context_lines":[{"line_number":110,"context_line":"----------"},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"* Merge central agent code into the current compute one"},{"line_number":113,"context_line":"  (ceilometer-polling-agent)."},{"line_number":114,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":115,"context_line":"* Provide Devstack with new agents scheme support (for all-in-one and multinode"},{"line_number":116,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_151b4674","line":113,"updated":"2014-11-27 08:38:37.000000000","message":"Just a thought, should we still provide ceilometer-agent-compute(central) which is actually a simple wrapper for backward compatibility?","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"50d420f6e310b69a97bd4e254f420fbb7a7def1e","unresolved":false,"context_lines":[{"line_number":110,"context_line":"----------"},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"* Merge central agent code into the current compute one"},{"line_number":113,"context_line":"  (ceilometer-polling-agent)."},{"line_number":114,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":115,"context_line":"* Provide Devstack with new agents scheme support (for all-in-one and multinode"},{"line_number":116,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_1a68ce20","line":113,"in_reply_to":"5a890539_151b4674","updated":"2014-11-27 10:41:57.000000000","message":"Lianhao, that\u0027s really great idea! This also will prevent me from breaking devstack gate :) Thanks! I\u0027ll fix the spec and POC, please let me now if I need to fix something more here.","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":4491,"name":"Lianhao Lu","email":"llh_misc@outlook.com","username":"lianhao-lu"},"change_message_id":"51f37ccbf44bbd170803d3c36a940bd8f2f5f343","unresolved":false,"context_lines":[{"line_number":110,"context_line":"----------"},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"* Merge central agent code into the current compute one"},{"line_number":113,"context_line":"  (ceilometer-polling-agent)."},{"line_number":114,"context_line":"* Investigate Tempest testing impact and provide new tests if needed."},{"line_number":115,"context_line":"* Provide Devstack with new agents scheme support (for all-in-one and multinode"},{"line_number":116,"context_line":"  installations)."}],"source_content_type":"text/x-rst","patch_set":10,"id":"5a890539_1a452517","line":113,"in_reply_to":"5a890539_1a68ce20","updated":"2014-11-27 15:12:20.000000000","message":"that is all, I think","commit_id":"422c2da87f07f3bc498580842cac2f9ab5e9a1a5"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c303e06e780080c0a82dbbf346bb3ce93ed90ffc","unresolved":false,"context_lines":[{"line_number":45,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":46,"context_line":"case running of polling agent will look like following:"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"    ceilometer-polling --polling--namespace central compute --logfile /var/log/ceilometer/polling.log"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":51,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_4e2a7d2d","line":48,"updated":"2014-11-28 12:24:45.000000000","message":"s/polling--namespace/polling-namespace/\n\nI haven\u0027t checked oslo.config CLI arg support recently, but does it support that form of multiple option-arguments? (i.e. central \u0026 compute are both args to the polling-namespace option).\n\nIf not, then two other alternative approaches would be:\n\n1. use a single option-argument in the form of a comma-separated list:\n\n   ceilometer-polling --polling-namespace central,compute ...\n\n2. use multiple occurrances of the option/option-arg pairs on the command line:\n\n   ceilometer-polling --polling-namespace central --polling-namespace compute ...","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"5689ff7e262fe754819ba7b65b89b4dc20201108","unresolved":false,"context_lines":[{"line_number":45,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":46,"context_line":"case running of polling agent will look like following:"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"    ceilometer-polling --polling--namespace central compute --logfile /var/log/ceilometer/polling.log"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":51,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_96f704d4","line":48,"in_reply_to":"5a890539_4e2a7d2d","updated":"2014-12-01 09:25:33.000000000","message":"Eoghan, actually my current POC has the format I\u0027ve used in this spec - I\u0027ve used nargs + choices opportunities for oslo.config - take a look on the https://review.openstack.org/#/c/124719/11/ceilometer/cmd/polling_agent.py,cm - I\u0027ve checked it, and this seems to work ok. I msy fix this double dash-dash if you\u0027d like to","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"6fdca5f50ab99a3dd763c113ec058c00b8e0a6a4","unresolved":false,"context_lines":[{"line_number":45,"context_line":"(with adding here *ipmi* when this agent will be added here as well). In this"},{"line_number":46,"context_line":"case running of polling agent will look like following:"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"    ceilometer-polling --polling--namespace central compute --logfile /var/log/ceilometer/polling.log"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"This parameter will be passed to the agent manager class directly, and will"},{"line_number":51,"context_line":"define what setup.cfg namespace (*ceilometer.poll.central*,"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_b657a026","line":48,"in_reply_to":"5a890539_96f704d4","updated":"2014-12-01 10:05:12.000000000","message":"A-ha, cool, that multi-choice option looks a good approach for this.","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c303e06e780080c0a82dbbf346bb3ce93ed90ffc","unresolved":false,"context_lines":[{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Also this change will propose new way of how do pollsters to be used are"},{"line_number":56,"context_line":"chosen for polling agent instance - via CLI script parameter"},{"line_number":57,"context_line":"*--pollster-list\u003d\u003clist of pollster entrypoints or wildcards\u003e*. In this case"},{"line_number":58,"context_line":"polling agent running command will look like this:"},{"line_number":59,"context_line":""},{"line_number":60,"context_line":"    ceilometer-polling --pollster-list image image.size storage.* --logfile /var/log/ceilometer/polling.log"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_8e77053b","line":57,"updated":"2014-11-28 12:24:45.000000000","message":"I\u0027d recommend a common root for --pollster-list and --polling-namespace, i.e. either \"polling\" or \"pollster\", but keep it consistent.\n\nThe reason is that a common root would allow the CLI args to be defined in the same group (so for example we could have \"namespace\" and \"list\" as options in the \"pollster\" group). The group-name is then prepended to the CLI option.","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"5689ff7e262fe754819ba7b65b89b4dc20201108","unresolved":false,"context_lines":[{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Also this change will propose new way of how do pollsters to be used are"},{"line_number":56,"context_line":"chosen for polling agent instance - via CLI script parameter"},{"line_number":57,"context_line":"*--pollster-list\u003d\u003clist of pollster entrypoints or wildcards\u003e*. In this case"},{"line_number":58,"context_line":"polling agent running command will look like this:"},{"line_number":59,"context_line":""},{"line_number":60,"context_line":"    ceilometer-polling --pollster-list image image.size storage.* --logfile /var/log/ceilometer/polling.log"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_d6efcce7","line":57,"in_reply_to":"5a890539_8e77053b","updated":"2014-12-01 09:25:33.000000000","message":"I think that in this case \u0027pollster\u0027  is preferable, but this might lead to kind of meaning inconsistency - because it\u0027ll be surely pollster list (with or without wildcards, but still), but as Julien mentioned in last reviews, pollster-namespace (that I used before) is not fully true. Although I may return pollster-namespace back :)","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"8d39d8170649387350add0a86560cae677a2097c","unresolved":false,"context_lines":[{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Also this change will propose new way of how do pollsters to be used are"},{"line_number":56,"context_line":"chosen for polling agent instance - via CLI script parameter"},{"line_number":57,"context_line":"*--pollster-list\u003d\u003clist of pollster entrypoints or wildcards\u003e*. In this case"},{"line_number":58,"context_line":"polling agent running command will look like this:"},{"line_number":59,"context_line":""},{"line_number":60,"context_line":"    ceilometer-polling --pollster-list image image.size storage.* --logfile /var/log/ceilometer/polling.log"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_bcbdbf05","line":57,"in_reply_to":"5a890539_9994958d","updated":"2014-12-01 10:43:28.000000000","message":"Also that\u0027s not fully true, because polling !\u003d pollster in fact. It might mess the polling itself as a process with different global separation compute-central we have with pollsters as basically classes to be used. And the moment you mentioned, yes.","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"6fdca5f50ab99a3dd763c113ec058c00b8e0a6a4","unresolved":false,"context_lines":[{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Also this change will propose new way of how do pollsters to be used are"},{"line_number":56,"context_line":"chosen for polling agent instance - via CLI script parameter"},{"line_number":57,"context_line":"*--pollster-list\u003d\u003clist of pollster entrypoints or wildcards\u003e*. In this case"},{"line_number":58,"context_line":"polling agent running command will look like this:"},{"line_number":59,"context_line":""},{"line_number":60,"context_line":"    ceilometer-polling --pollster-list image image.size storage.* --logfile /var/log/ceilometer/polling.log"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_9994958d","line":57,"in_reply_to":"5a890539_d6efcce7","updated":"2014-12-01 10:05:12.000000000","message":"TBH it was only a minor point, so don\u0027t sweat it.\n\nThe reason pollster-namespace is not fully true is because this value is also used to construct the group prefix for partitioning?\n\nFair enough, makes sense.","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c303e06e780080c0a82dbbf346bb3ce93ed90ffc","unresolved":false,"context_lines":[{"line_number":60,"context_line":"    ceilometer-polling --pollster-list image image.size storage.* --logfile /var/log/ceilometer/polling.log"},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"If both of these parameters will be set, we need to use them both (possibly using"},{"line_number":63,"context_line":"logical AND operaiton to find out final list of pollsters to be used)."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"Alternatives"},{"line_number":66,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_ae18411f","line":63,"updated":"2014-11-28 12:24:45.000000000","message":"s/operaiton/operation/","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"c303e06e780080c0a82dbbf346bb3ce93ed90ffc","unresolved":false,"context_lines":[{"line_number":101,"context_line":"---------------------"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"* It\u0027ll be less services Ceilometer will run on node (if it\u0027ll be all-in-one"},{"line_number":104,"context_line":"  installation) and we\u0027ll have more unified way of Ceilometer installation."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"Developer impact"},{"line_number":107,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_13c0a2df","line":104,"updated":"2014-11-28 12:24:45.000000000","message":"There are also downstream impacts on packagers and deployment tools.\n\nTypically there are *separate* .rpm/.deb currently for the central and compute agents, containing different .service/.upstart definitions for launching the corresponding service.\n\nOne simple approach would be for these existing packages to be modified to use the appropriate namespace for the individual agent, e.g.:\n\n s/ExecStart\u003d...ceilometer-agent-compute/\n   ExecStart\u003d...ceilometer-agent-polling --pollster-namespace\u003dcompute/\n\nand also add a net-new package for the combined package. This new package could be selected by meta-installer tools running in allinone/simple mode.\n\nWe should also call out the impact on deployment tools such as the puppet-ceilometer module. Note for example the difference between:\n\n  https://github.com/stackforge/puppet-ceilometer/blob/master/manifests/agent/central.pp\n\nand:\n\n  https://github.com/stackforge/puppet-ceilometer/blob/master/manifests/agent/compute.pp\n\nThere\u0027s significantly more complexity in the compute case, as there\u0027s an ordering dependency on the nova compute service.\n\nAt the least we\u0027d need a new manifest for the combined agent that duplicates/extends (?) this logic.\n\nNote that I\u0027m not implying that all the above work needs to be undertaken by the ceilometer upstream team before we can declare the BP implemented. Instead these changes will be made by the folks already involved in packaging/puppetizing ceilometer (eglynn,pbrady,cdent,EmilienM,zul etc). However I think it would be useful to call out the impacts in this spec.","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":3012,"name":"Dina Belova","email":"dbelova@mirantis.com","username":"dbelova"},"change_message_id":"5689ff7e262fe754819ba7b65b89b4dc20201108","unresolved":false,"context_lines":[{"line_number":101,"context_line":"---------------------"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"* It\u0027ll be less services Ceilometer will run on node (if it\u0027ll be all-in-one"},{"line_number":104,"context_line":"  installation) and we\u0027ll have more unified way of Ceilometer installation."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"Developer impact"},{"line_number":107,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_f6c1284f","line":104,"in_reply_to":"5a890539_13c0a2df","updated":"2014-12-01 09:25:33.000000000","message":"ok, will add some lines here, thanks!","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"6fdca5f50ab99a3dd763c113ec058c00b8e0a6a4","unresolved":false,"context_lines":[{"line_number":101,"context_line":"---------------------"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"* It\u0027ll be less services Ceilometer will run on node (if it\u0027ll be all-in-one"},{"line_number":104,"context_line":"  installation) and we\u0027ll have more unified way of Ceilometer installation."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"Developer impact"},{"line_number":107,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"5a890539_d99d3d68","line":104,"in_reply_to":"5a890539_f6c1284f","updated":"2014-12-01 10:05:12.000000000","message":"Thanks!","commit_id":"d1d11e79f455a097433d513a08b496dc349bdc2e"}]}
