Wednesday, 2013-12-11

openstackgerritYuuichi Fujioka proposed a change to openstack/ceilometer: (WIP)Implements monitoring-network
*** herndon has quit IRC00:21
*** sandywalsh_ has quit IRC01:09
*** sandywalsh_ has joined #openstack-ceilometer01:21
*** djbkd has quit IRC01:34
*** herndon has joined #openstack-ceilometer02:00
*** herndon has quit IRC02:20
openstackgerritA change was merged to openstack/ceilometer: Updated from global requirements
openstackgerritA change was merged to openstack/python-ceilometerclient: Updated from global requirements
*** gordc has joined #openstack-ceilometer03:49
*** gordc has quit IRC03:54
openstackgerritSushil Kumar proposed a change to openstack/python-ceilometerclient: Updates tox.ini to use new features
*** SergeyLukjanov has joined #openstack-ceilometer04:27
openstackgerritA change was merged to openstack/ceilometer: add more test cases to improve the test code coverage #5
*** SergeyLukjanov is now known as _SergeyLukjanov05:09
*** _SergeyLukjanov has quit IRC05:10
*** SergeyLukjanov has joined #openstack-ceilometer05:14
openstackgerritJenkins proposed a change to openstack/ceilometer: Imported Translations from Transifex
*** SergeyLukjanov has quit IRC06:21
*** ildikov has joined #openstack-ceilometer06:44
*** ildikov_ has joined #openstack-ceilometer06:46
*** yfujioka has joined #openstack-ceilometer06:59
*** urulama has joined #openstack-ceilometer07:31
*** SergeyLukjanov has joined #openstack-ceilometer07:49
*** ildikov__ has joined #openstack-ceilometer07:59
*** Alienyyg has joined #openstack-ceilometer08:01
AlienyyghiHi guys:  when I use ceilometer alarm-show to list the information about a alarm, I got insuficent data, the alarm and the instance are in the same tenant ,and all be up for a long time, what does the " insuficent data"mean ?08:02
*** ildikov_ has quit IRC08:02
Alienyygand  can anyone tell me how to clear the history of ceilometer ? beacuse I only have on VM now,but when I use ceilometer meter-list , I got many response,It is very hard to figure out the one I need08:03
Alienyygany help?08:06
*** nprivalova has joined #openstack-ceilometer08:52
*** Alexei_987 has joined #openstack-ceilometer09:50
*** ildikov__ has quit IRC10:05
*** ildikov_ has joined #openstack-ceilometer10:06
*** nprivalova has quit IRC10:20
*** SergeyLukjanov is now known as _SergeyLukjanov10:23
openstackgerritThomas Herve proposed a change to openstack/python-ceilometerclient: Improve description of some commands
*** jd__ has quit IRC10:48
*** jd__ has joined #openstack-ceilometer10:48
*** sayali has joined #openstack-ceilometer10:50
*** nprivalova has joined #openstack-ceilometer10:57
*** nprivalova has quit IRC11:09
*** Alexei_987 has quit IRC11:20
openstackgerritMark McLoughlin proposed a change to openstack/ceilometer: doc: fix formatting of alarm action types
openstackgerritMark McLoughlin proposed a change to openstack/ceilometer: doc: remove note about Nova plugin framework
*** nprivalova has joined #openstack-ceilometer12:04
*** Alienyyg has quit IRC12:04
*** SergeyLukjanov has joined #openstack-ceilometer12:10
*** yfujioka has quit IRC12:12
*** ildikov_ has quit IRC12:15
*** nprivalova has quit IRC12:21
*** nprivalova has joined #openstack-ceilometer12:26
*** nprivalova has quit IRC12:42
*** Alexei_987 has joined #openstack-ceilometer12:49
*** nprivalova has joined #openstack-ceilometer12:51
*** urulama has quit IRC12:53
*** nprivalova has quit IRC13:04
*** nprivalova has joined #openstack-ceilometer13:05
*** sandywalsh_ has quit IRC13:07
*** urulama has joined #openstack-ceilometer13:13
*** nprivalova has quit IRC13:17
*** sandywalsh_ has joined #openstack-ceilometer13:20
*** prad has joined #openstack-ceilometer13:23
*** jdob has joined #openstack-ceilometer13:35
sandywalsh_asalkeld, sileht  nijaba any chance of a +1 on ?13:39
Alexei_987jd__: ping13:46
*** prad has quit IRC14:06
*** thomasem has joined #openstack-ceilometer14:11
*** prad has joined #openstack-ceilometer14:12
*** gordc has joined #openstack-ceilometer14:24
*** terriyu has joined #openstack-ceilometer14:40
*** eglynn has joined #openstack-ceilometer14:45
*** jdob has quit IRC15:00
*** SergeyLukjanov_ has joined #openstack-ceilometer15:03
*** sayali_ has joined #openstack-ceilometer15:05
*** jdob has joined #openstack-ceilometer15:06
*** SergeyLukjanov has quit IRC15:06
*** sayali has quit IRC15:09
*** sayali_ has quit IRC15:09
*** SergeyLukjanov has joined #openstack-ceilometer15:10
*** SergeyLukjanov_ has quit IRC15:12
*** litong has joined #openstack-ceilometer15:20
*** sayali_ has joined #openstack-ceilometer15:21
Alexei_987jd__: ping15:28
jd__Alexei_987: pong15:31
Alexei_987jd__: Hi do you have some time to chat?15:31
Alexei_987jd__: I'm working on the models refactoring that we discussed recently -
jd__Alexei_987: yes15:32
Alexei_987jd__:  Doug blocked this patch but I cannot reach him to ask why :)15:32
Alexei_987jd__: do you have any idea about what validation framework he was talking about?15:33
Alexei_987jd__: cause validation is not the only purpose of this patch and I need it for other stuff as well15:33
jd__oh I definitely know what he is talking about15:34
Alexei_987jd__ please share your knowledge with me :)15:34
jd__you wrote something that already exists in at least 3 places: nova objects, jsonschema and WSME15:34
Alexei_987I'm afraid you don't get the main idea of the patch15:34
jd__Doug and I keep complaining that we have too many of these validation things in OpenStack15:34
Alexei_987jd__ nova-objects - we don't have them in ceilometer (and I hope we'll never have)15:35
jd__Alexei_987: well having "'name': str" covers that from what I understand15:35
Alexei_987yes but it's not the only purpose15:35
Alexei_9871) I need to have static field definition in the class15:35
jd__I agree it's not the only purpose15:35
Alexei_987validation is only a side effect15:36
Alexei_9872) My IMHO models is where we keep the data15:36
Alexei_987validation should be near the data15:36
Alexei_987so it's the correct place to put it15:36
Alexei_987not WSME15:36
jd__I think we're not against the idea of the patch, just on the implementation of the data validation that we don't want15:36
jd__i.e. this patch should only be about writing data schema and that's it, leveraging something else for the validation15:37
jd__I talk about WSME because WSME does data validation for example15:37
Alexei_987why leverage it to something else ?15:37
Alexei_987when object is created it should be valid15:37
Alexei_987so the right place is __init__15:37
jd__I think we're not talking about the same things15:37
jd__take a step back15:37
Alexei_987so what is WSME validation?15:38
Alexei_987and jsonschema validation :)15:38
Alexei_987what's the purpose of this stuff?15:38
jd__if you take a look at ceilometer.api.controllers.v2, you'll see there is a bunch of class15:38
Alexei_987exactly :)15:38
jd__like Sample, Meter, etc… with fields15:38
Alexei_987and I won't to move it out of there15:39
Alexei_987cause it's crappy15:39
jd__that's a schema, describing a data structure and the type of the fields15:39
Alexei_987but it's not supposed to be in there15:39
jd__that's used by WSME to validate the data that it receives15:39
jd__that's not my point Alexei_98715:39
Alexei_987ok I get it15:39
jd__my point is that WSME is doing "oh here's a schema and data, let's validate"15:39
jd__nova objects is doing the same thing, in a different manner, in a different place15:40
Alexei_987both are doing it wrong :)15:40
jd__jsonschema is doing the same thing in a different context15:40
Alexei_987data is described in storage/models15:40
jd__and now you are writing something doing it, in a different place, with a different mechanism15:40
Alexei_987yes :)15:40
Alexei_987cause I want to remove both WSME validation + storage validation15:40
jd__I'm not telling that validating data is wrong15:40
jd__your idea of validating in this part of the code is likely a good idea15:41
jd__we don't want you to write a validation mechanism15:41
Alexei_987I could reuse existing one :)15:41
jd__we want you to write the schema in, and use "something" to validate the data against the schema15:41
jd__the question is what to reuse :-)15:41
Alexei_987I don't care which one :)15:42
Alexei_987but it should use the lambdas approach15:42
Alexei_987it will allow to add business logic validation later15:42
Alexei_987e.g. that referenced object exists15:43
Alexei_987and so on15:43
Alexei_987cause we cannot rely on storage to maintain references15:43
jd__yeah, so what we need to do first is to find the right tool to validate and use it everywhere15:43
jd__that's on my TODO list15:43
Alexei_987but lambdas patches don't block this search15:43
Alexei_987so I think that we can add this patch with simple type validation at first15:44
jd__so I definitely suggest you keep this patch under your elbow for now, but it's unlikely it's going to be doable until we clear that out15:44
Alexei_987and move proper validation there later15:44
jd__Alexei_987: not sure you can convince us :)15:44
Alexei_987well.. I could work on the whole patch series15:44
Alexei_987and it should result in -1000 lines of ceilometer code15:44
Alexei_987maybe this should convince you :)15:45
openstackgerritlitong01 proposed a change to openstack/ceilometer: test code should be excluded from test coverage summary
Alexei_987cause what I like the most is deleting code :)15:45
Alexei_987jd__: the problem I see is that we have some functionality duplicated in ceilometer - query creation in controllers level + query transformation in storage level15:46
Alexei_987this stuff should move to common query code in models15:47
Alexei_987cause it's objects are used on both levels15:47
Alexei_987jd__ what do you think ^15:47
Alexei_987and to simplify query code I need to have 1) static field description in models 2) simple validation that makes sure that object is valid if we created it15:48
*** dhellmann has joined #openstack-ceilometer15:51
dhellmannAlexei_987: my main issue with the patch is that it reproduces some features of things like WSME,, and even some parts of the nova object code that another team is trying to move into oslo15:52
Alexei_987ok 1) WSME - is not the place to validate objects cause they are defined in storage/models and validation should be during this object creation15:53
Alexei_987this will allow to use it both in WSME layer15:53
Alexei_987and storage layer15:53
Alexei_9872) nova object is a complicated feature and mostly useless for ceilometer. I hope that we'll never have them here15:54
dhellmannAlexei_987: well, the API layer is doing different types of validation than this might, but the bigger issue is adding a 4th way to declare  classes and validators15:54
dhellmannif one of the existing frameworks doesn't do what we need, let's update it instead of making a new one15:54
*** urulama has quit IRC15:54
*** viktors has joined #openstack-ceilometer15:55
Alexei_987dhellmann: I don't care much about what framework we use and I can reuse existing one15:55
Alexei_987but I care about where and when validation is happening15:55
Alexei_987and it SHOULD happen only inside model __init__15:55
jd__a minute guy and I'm with you15:55
Alexei_987nowhere else15:55
Alexei_987if we'll replace our models with nova objects I will -1 this patch but it's ok afterall15:56
dhellmannAlexei_987: different layers may want or need to do different validation15:56
Alexei_987no if they operate with the same object15:56
Alexei_987we only have 1 object definition15:56
Alexei_987and it's used in several layers15:57
Alexei_987object cannot be valid for 1 layer and invalid for the other15:57
Alexei_987that's just doesn't make sense15:57
dhellmannAlexei_987: you're assuming the API data structure and the storage data structures match exactly, but they don't always15:58
Alexei_987yes.. but they use 1 object to communicate :)15:59
Alexei_987and this object should validate itself when created15:59
Alexei_987if it's created with wrong data on WSME level it doesn't matter what data structure is used in WSME16:00
dhellmannthe API layer also needs to enforce permission rules that do not apply in the storage layer16:00
Alexei_987permissions ok to be in WSME :)16:00
dhellmannok, I wasn't clear: the data models can be different without one of them being wrong. We may store things differently than we present them to the user.16:00
Alexei_987yes but why we need to validate such conversion?16:01
Alexei_987we have 2 usecases:16:01
Alexei_9871) data from from the user to the storage16:01
Alexei_9872) data comes from storage to user16:01
dhellmannand data from events to the storage16:01
Alexei_987this can be considered 1)16:01
dhellmannno, it can't16:01
dhellmannthat's what I mean by permissions16:01
dhellmannthe user can post data to the API that we might reject16:01
Alexei_987why not? event is the same user16:02
dhellmannthe notification stream is already protected, so we accept it as a source of truth16:02
Alexei_987ok permissions is the other topic16:02
Alexei_987but in both cases + events we use storage/models to transfer data16:02
Alexei_987from 1 layer to the other16:03
dhellmannyes, that is correct16:03
Alexei_987and we should check that it's valid when we create a data object16:03
Alexei_987there is no point in any other validation16:03
Alexei_987it will fail anyway16:03
jd__dhellmann: any reason you came up with colander?16:03
dhellmannjd__: it was discussed as an alternative to having the nova team create their own serialization framework16:04
Alexei_987lets say that we have 3 validation layers: api, model, storage16:04
Alexei_987but we use & to pass data16:04
Alexei_987so true & false & true will fail16:04
dhellmannAlexei_987: if you can show that we are not doing any validation in the WSME models that shouldn't be moved to the storage models, then that's fine. I am objecting primarily to making up yet another way of declaring those validators. Please look at some existing libraries and pick one.16:05
*** SergeyLukjanov has quit IRC16:05
Alexei_987well the reason I'm using lambdas is because it will allow to add business logic validation later16:05
Alexei_987and maintain reference integrity for storage layer16:05
dhellmannAlexei_987: I don't care about the lambdas. I care about using the __slots__ data structure to define them.16:06
dhellmannjd__: I don't have any particular fondness for colander, except that Christophe pointed it out when I asked about splitting similar functionality out of WSME16:08
jd__dhellmann: cool, I'm building a list of candidates, I'm going to study all of this in the near future16:08
jd__you know the future where I have time to do it16:09
Alexei_987dhellmann: __slots__ is cool16:09
Alexei_987dhellmann: it allows to define all the fields we can have16:09
Alexei_987+ it makes object smaller16:09
dhellmannAlexei_987: if there was no prior art to do this, then I would agree. Don't reinvent this wheel.16:09
Alexei_987yes but currently we don't have a list of object fields in the model16:10
dhellmannAlexei_987: and using slots to reduce memory is fine, it's the combination of that with making up a new way to declare fields and validators that I don't like16:10
Alexei_987again - we don't have any fields declaration in model16:10
Alexei_987and I NEED it16:10
Alexei_987__slots__ is the right place to do it16:10
dhellmannAlexei_987: have you even looked at the other libraries I mentioned?16:11
Alexei_987yes :) i'm familiar with nova objects and WSME validators16:11
Alexei_987they are doing the same thing16:11
dhellmannyes, that's my point16:11
Alexei_987nova object just uses another variable for this16:11
Alexei_987and again my reasoning is the same 1) we don't have nova objects16:12
Alexei_9872) WSME cannot describe model fields16:12
Alexei_987cause it's on the other layer16:12
dhellmannI agree that using WSME directly is not the right approach.16:12
dhellmannI'm currently discussing whether nova objects will go into oslo, too, in some form.16:13
Alexei_987ok let's discuss nova objects16:13
Alexei_987their main idea is rpc remoting16:13
Alexei_987we don't have rpc remoting16:13
Alexei_987+ we don't have versioning16:13
Alexei_987and their implementation is already full of hacks to support this 2 features16:13
dhellmannI have issues with their implementation, too, because they did the same thing -- they built a framework from scratch that we would have to maintain, instead of using an existing library16:14
Alexei_987we don't need this 2 features - we don't need nova objects16:14
dhellmannAlexei_987: you're missing my point -- these other implementations do more than what yours does, but they also do the things yours does. Why have multiple implementations of the same set of features?16:14
Alexei_987well feature set is a little different16:14
dhellmannI don't have time right now to carry on this conversation live, I'm afraid.16:14
*** prad_ has joined #openstack-ceilometer16:31
*** prad has quit IRC16:31
*** prad_ is now known as prad16:31
openstackgerritlitong01 proposed a change to openstack/ceilometer: test code should be excluded from test coverage summary
*** herndon has joined #openstack-ceilometer17:02
openstackgerritJames E. Blair proposed a change to openstack/python-ceilometerclient: Have tox install via develop
*** herndon has quit IRC17:04
*** nprivalova has joined #openstack-ceilometer17:12
*** thomasem has quit IRC17:20
*** SergeyLukjanov has joined #openstack-ceilometer17:20
*** nprivalova has quit IRC17:36
*** thomasem has joined #openstack-ceilometer17:43
*** thomasem has quit IRC17:44
*** thomasem has joined #openstack-ceilometer17:44
*** Alexei_987 has quit IRC18:07
*** sayali_ has quit IRC18:23
openstackgerritA change was merged to openstack/ceilometer: Add configuration-driven conversion to Events
*** thomasem has quit IRC18:54
*** gordc has quit IRC18:55
*** gordc has joined #openstack-ceilometer18:56
*** prad has quit IRC18:59
*** insanidade has joined #openstack-ceilometer19:01
*** jdob has quit IRC19:03
*** jdob has joined #openstack-ceilometer19:03
*** prad has joined #openstack-ceilometer19:06
*** openstackgerrit has quit IRC19:34
*** openstackgerrit has joined #openstack-ceilometer19:34
*** prad has quit IRC19:40
*** herndon has joined #openstack-ceilometer19:41
*** prad has joined #openstack-ceilometer19:43
insanidadeanyone awake ?19:47
herndonany cores around? jd__ eglynn llu dhellmann? This patch has been sitting for a while, would like to get some feedback on it so we can keep working on the subsequent patches:
eglynnherndon: I'm around just about, but have to head off very shortly ... I'll try to get to it this review later evening, otherwise first thing tmrw19:51
*** eglynn is now known as eglynn-afk19:52
*** thomasem has joined #openstack-ceilometer19:58
*** SergeyLukjanov has quit IRC20:02
*** herndon has quit IRC20:07
*** insanidade has quit IRC20:28
*** sandywalsh_ has quit IRC20:42
*** gordc has quit IRC20:53
*** sandywalsh_ has joined #openstack-ceilometer20:54
*** DanD has joined #openstack-ceilometer21:10
*** gordc has joined #openstack-ceilometer21:13
*** thomasem has quit IRC21:25
*** thomasem has joined #openstack-ceilometer21:30
*** thomasem has quit IRC21:31
*** eglynn-afk has quit IRC21:42
openstackgerritJulien Danjou proposed a change to openstack/ceilometer: config: specify a template for mktemp
openstackgerritJulien Danjou proposed a change to openstack/ceilometer: service: fix service alive checking
*** jdob has quit IRC22:03
*** eglynn-afk has joined #openstack-ceilometer22:14
*** herndon has joined #openstack-ceilometer22:28
openstackgerritJohn Herndon proposed a change to openstack/ceilometer: Event Storage Layer
*** eglynn-afk has quit IRC22:47
*** litong has quit IRC22:51
openstackgerritDavid Peraza proposed a change to openstack/ceilometer: Oslo sync to recover from db2 server disconnects
*** herndon has quit IRC23:21
*** gordc has quit IRC23:29
*** terriyu has quit IRC23:32
*** openstackgerrit has quit IRC23:35
*** openstackgerrit has joined #openstack-ceilometer23:36
*** openstackgerrit has quit IRC23:56
*** openstackgerrit has joined #openstack-ceilometer23:56

Generated by 2.14.0 by Marius Gedminas - find it at!