)]}'
{"specs/liberty/pollsters-no-transform.rst":[{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"b60fb3f0a13aefbf3447667497ec1f5db30d6e17","unresolved":false,"context_lines":[{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"The primary alternative is to leave things as they are. Obviously if we do this"},{"line_number":48,"context_line":"we lose the benefits described abvoe."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"Data model impact"},{"line_number":51,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_11074882","line":48,"updated":"2015-06-22 15:55:15.000000000","message":"s/abvoe/above/","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"b60fb3f0a13aefbf3447667497ec1f5db30d6e17","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_3177ec03","line":70,"updated":"2015-06-22 15:55:15.000000000","message":"this will render the interval attr in pipeline moot no? we need a way to define interval on polling agents wihtout pipeline.","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"c5804f13966d124626a4306c5c87cf6c86dfbb5e","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_b29e9752","line":70,"in_reply_to":"fa32b979_1ce51fa7","updated":"2015-06-22 18:03:42.000000000","message":"but we will not need the pipeline anymore right? in the polling agents that is. i guess one question is are we actually building the meters in the polling agent or are we letting the meters be built in notification agent and just grabbing the appropriate data in polling agent.","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"b59b33ee1b4d81e4a1244ffe660ce7be184fc1a0","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_1ce51fa7","line":70,"in_reply_to":"fa32b979_3177ec03","updated":"2015-06-22 16:12:27.000000000","message":"So the idea is that initially the first half (sources) of pipeline.yaml is still used, but not the second half (sinks).\n\nIf that makes sense I can clarify the spec.","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"3ed75069cc1ed6ea24165c0981b340004935406c","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_66aa3173","line":70,"in_reply_to":"fa32b979_4312a658","updated":"2015-06-24 13:31:49.000000000","message":"yep. i guess the main question was our we still loading the pipeline.yaml into polling agents or are we creating a \u0027source_only\u0027 pipeline.yaml? because it might be confusing for deployers to load a \u0027full\u0027 pipeline.yaml\u0027 into polling agents when only half is ever used\n\nthis probably isn\u0027t a huge concern as it appears most people use the same pipeline.yaml across notification and polling agents","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":13560,"name":"Rohit Jaiswal","email":"rohit.jaiswal@gmail.com","username":"rjaiswal"},"change_message_id":"8e12879905204c1f5538c7971635156fb32b6e9f","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_a5354b45","line":70,"in_reply_to":"fa32b979_66aa3173","updated":"2015-06-25 17:31:05.000000000","message":"Currently, the decoupled format of the pipeline is validated to ensure a valid sink linked to a source and uniqueness of each pipe (source:sink pair).\n\nHow will having a source_only pipeline for polling agents affect this validation for the pipeline as a whole?","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"104d0b4eaedc123524d0893f0497d02e705eee8a","unresolved":false,"context_lines":[{"line_number":67,"context_line":"Pipeline impact"},{"line_number":68,"context_line":"---------------"},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Pipeline behavior remains the same but will be running in a different agent."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Other end user impact"},{"line_number":73,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa32b979_4312a658","line":70,"in_reply_to":"fa32b979_b29e9752","updated":"2015-06-24 08:16:47.000000000","message":"That word \"pipeline\", it\u0027s horrible. Means too many things at once in the ceilo universe.\n\nSame with \"sample\" and \"meter\".\n\nThe polling agents will still yield Samples so that there is no need to change the pollster plugin implementations (an important consideration I think). So I guess the answer is, yes, they build meters. These will be sent to notification agent, hopefully in _exactly_ the same way that the new pipeline-less API server is going to do.\n\nThis getting any closer to making sense?","commit_id":"7db4b54b5c460e2b16f01c3b5eebf0aa611d992f"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"07ebbb5cd5c4941661f80a8ff65b606eecd52a5f","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_6c55b2f8","line":43,"updated":"2015-06-26 21:17:04.000000000","message":"i remember why i asked if the polling agent would build sample or not. the reason was that if we let the sample be built on the notification agent end, we can also build \u0027healthcheck\u0027 events from them.\n\nright now it seems with this solution we still need to rely on non-sense volume\u003d1 samples for healthcheck\n\ni was thinking this could be more like ceilometermiddleware. where it builds a single notification payload which pretty much contains resource data and also metric data[1]. ignore the fact it\u0027s CADF but the basic premise is from that payload we derive events and samples on notification agent\n\n[1] https://github.com/openstack/ceilometer/blob/master/ceilometer/tests/objectstore/test_notifications.py#L44-L88","commit_id":"3dcdfc425b36816382df125f0e73976b049d89b0"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"50d9c4b05625ee5012c5a33839f8540c8c0d0948","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"ba3cc151_471d4e72","line":43,"in_reply_to":"ba3cc151_0103b674","updated":"2015-06-29 13:50:01.000000000","message":"Is it fair to say \"Yo, man, one step at a time!\"?\n\nAn in let\u0027s incrementally improve towards nirvana.","commit_id":"3dcdfc425b36816382df125f0e73976b049d89b0"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"caf1b9193801bc74fcf0f0fa727b8ad92cc5c1c3","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"ba3cc151_0103b674","line":43,"in_reply_to":"fa32b979_09116d04","updated":"2015-06-29 13:06:06.000000000","message":"yeah, i understand what is being proposed, i\u0027m trying to figure out a way how to drop the healthcheck (volume\u003d1) samples and get them as events since this seems like a good step to implement it.\n\nalso, i guess the idea was that if we build the sample on pollster end, the payload now consists of x samples for y resources. since all samples have metadata, you now of x duplicate metadata entries. if we wanted to minimise the payload (not that i think it\u0027s large), it\u0027d be a reason to not build full sample at the pollster.","commit_id":"3dcdfc425b36816382df125f0e73976b049d89b0"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"5b733fc7f1711e873a514a13115ceeb9418c8df1","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_09116d04","line":43,"in_reply_to":"fa32b979_6c55b2f8","updated":"2015-06-27 07:51:10.000000000","message":"I think we can pretty much achieve this anyway (or even have to if we want to maintain some measure of genericness without blowing up existing plugins):\n\n* Existing pollsters yield samples (eg: https://github.com/openstack/ceilometer/blob/master/ceilometer/network/floatingip.py#L60). We don\u0027t want to change that.\n* Notifications are the common method for sending stuff on the bus.\n\nSo I think pretty much what the pollsters are going to end up doing is serializing a sample into a notification and then the notification agent will take it and decide what to do with it, yeah?","commit_id":"3dcdfc425b36816382df125f0e73976b049d89b0"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"fd00dff97421bfcbcfc583609898441ebc7a3134","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba3cc151_4b2edb8c","line":43,"updated":"2015-07-03 12:39:38.000000000","message":"tbh, i think we should hash out a rough schema for payload. just so we can deal with versioning and having a liberty \u0027polling lib\u0027 work with m* \u0027listener lib\u0027 (this seems to be very real since i\u0027m ask this more and more)\n\nie.\npayload \u003d {\n\u0027resource\u0027 : {id:1, proj: proj, attr:1, attr2:2},\n\u0027samples\u0027: [{name:blah,type}]}\n\nor something similar to ceilometermiddleware.","commit_id":"a649270528a3b86a0f3aad39076930e5c61e63b1"},{"author":{"_account_id":6537,"name":"gordon chung","email":"gord@live.ca","username":"chungg"},"change_message_id":"f3463245d2c1605f6be76898c38f58cf19808c61","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba3cc151_84800d48","line":43,"in_reply_to":"ba3cc151_2ba91724","updated":"2015-07-06 19:55:35.000000000","message":"the reasoning for schema is that because of the split in project, we now have interoperability between different versions to think about... and we in ceilometer have experienced all to often where a change in another service will destroy our listeners.\n\nschema are a pain when dealing with other projects but i\u0027m sorta hoping it\u0027s a lot easier if it\u0027s internal to us.","commit_id":"a649270528a3b86a0f3aad39076930e5c61e63b1"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"c828a73627c4f149bc8d18c4bf201a75fb02578f","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"We also get an opportunity to review and audit the polling code, which is a"},{"line_number":42,"context_line":"useful thing to do on a regular basis."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Alternatives"},{"line_number":45,"context_line":"------------"},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba3cc151_2ba91724","line":43,"in_reply_to":"ba3cc151_4b2edb8c","updated":"2015-07-03 12:54:54.000000000","message":"The mental model I had in my head was that the pollsters were going to be:\n\na) as dumb as possible\nb) require no change to existing pollster plugins (which are generators that yield a sample)\n\nYour proposed payload implies that we\u0027ll look for opportunities to batch up samples which are of the same resource. This will require some management of state and a bit of brains. Both of these are bad news if it is at all possible to avoid them.\n\nIt will also require decomposing the Sample into a new form.\n\nWhat I was thinking is that we fire samples onto the bus as soon as we get them and let whatever the listener is figure it out. I feel this allows the most flexibility in the listening audience and also makes the individual I/O events unitary and self-contained which is, in general, a good idea in a distributed system.\n\nFire and forget and all that jazz.\n\n(Note that I\u0027m not trying to say I\u0027m against having a rough schema, rather that I\u0027m commenting on the granularity.)","commit_id":"a649270528a3b86a0f3aad39076930e5c61e63b1"}]}
