Monday, 2017-03-06

*** zhurong has joined #openstack-watcher01:13
*** zhurong has quit IRC02:15
*** zhurong has joined #openstack-watcher02:19
*** necuser__ has quit IRC02:38
*** necuser__ has joined #openstack-watcher02:50
*** thorst has quit IRC03:08
openstackgerritMerged openstack/watcher-dashboard master: [Fix gate]Update test requirement
openstackgerritMerged openstack/watcher master: [Fix gate]Update test requirement
*** luyao has quit IRC04:16
*** thorst has joined #openstack-watcher04:28
*** thorst has quit IRC04:28
*** karthikpr has quit IRC05:04
*** thorst has joined #openstack-watcher05:29
*** thorst has quit IRC05:34
*** Yumeng has joined #openstack-watcher05:36
*** alexchadin has joined #openstack-watcher06:26
*** thorst has joined #openstack-watcher06:30
*** thorst has quit IRC06:34
*** aovchinnikov has joined #openstack-watcher06:55
aovchinnikovHello everyone07:02
aovchinnikovI have a question regarding watcher's behaviour. How does it handle nova scheduling rules when deciding where to migrate a VM during cloud optimization?07:03
aovchinnikovFor example, consider two VMs booted with SameHostFilter. Is it guarantied that such rule won't be optimized?07:03
alexchadinhi aovchinnikov07:24
alexchadinaovchinnikov: Watcher doesn't respect nova scheduling filters for now. There is no API for Nova Scheduler so we rely only on Watcher Audit scope currently07:26
aovchinnikovalexchadin ok, so what are the plans to mitigate this in the future?07:29
aovchinnikovbecause there are many real-world scenarios where both optimization and respect for nova rules are highly desirable07:29
alexchadinaovchinnikov, I agree, but Nova team doesn't know what to do with requests from Watcher, Congress. There are plans (based on latest info from PTG) to get rid of Scheduler and make scheduling process decentralized (maybe, using Placement API), but it is only plans for future.07:32
aovchinnikovwell, that sounds like a very distant future to me07:33
aovchinnikovdoes nova team have any estimates on when something will be done to address this issues?07:34
aovchinnikovbecause till there is no way to tell which hosts are forbidden for a particular instance watcher will be somewhat limited in its usability07:36
alexchadinaovchinnikov maybe Pike, maybe Queens :) This problem didn't appear yesterday, we have a long discussion with them07:36
alexchadinaovchinnikov you can use Audit Scope, but it is more applicable for small clouds07:36
aovchinnikovhave you considered any alternative routes to follow while there is no Queens?07:37
aovchinnikovlike peeping into nova's DB or something?07:38
alexchadinaovchinnikov: there is always a downstream way to modify it by yourself and share it somewhere. We are not nova-related project so we have to use common ways to communicate with Nova services07:38
alexchadinaovchinnikov, I had worked on Nova LoadBalancer (experimental project of Servionica) before I came to Watcher and we respected filters by modifying source code of Nova07:39
aovchinnikov>Audit Scope... well, there is a good chance that I'll have to run it on a medium-sized cloud with somewhat intricate rules applied to VMs, so I was hoping for some sort of a direction which is somewhat aligned with community's one07:40
aovchinnikovyes, downstream changes was one of options we've considered, but that is not a very good solution for long-term07:41
alexchadinaovchinnikov: Here is example of scope
aovchinnikovalexchadin: have you considered reusing parts of nova scheduler inside watcher? that has a potential of solving this problem07:44
aovchinnikovwhile bringing in a plethora of others, but anyway07:44
aovchinnikovalexchadin: thank you for the link, I'll give scopes a try07:45
alexchadinaovchinnikov: we don't want to multiply entities. We would need to copy almost all scheduler code base and then we need to be synced with nova filters07:46
aovchinnikovyes, those are some of aforementioned other issues07:46
aovchinnikovalexchadin: thank you for your help!07:47
alexchadinaovchinnikov: there would be a lot of new changes only for a while and cutting it in future... It doesn't make me sense07:47
alexchadinaovchinnikov you are welcome!07:47
*** alexchadin has quit IRC07:59
*** alexchadin has joined #openstack-watcher08:00
openstackgerritSanthosh Fernandes proposed openstack/watcher master: Add gnocchi support in basic_consolidation strategy
*** sanfern has joined #openstack-watcher08:31
*** thorst has joined #openstack-watcher08:32
*** thorst has quit IRC08:36
*** vmahe has joined #openstack-watcher08:37
*** vincentfrancoise has joined #openstack-watcher08:40
*** thorst has joined #openstack-watcher09:33
*** thorst has quit IRC09:37
*** pksingh has joined #openstack-watcher09:53
*** alexchadin has quit IRC10:11
*** alexchadin has joined #openstack-watcher10:30
pksinghvincentfrancoise: Hello10:33
*** openstackgerrit has quit IRC10:33
vincentfrancoisepksingh: hi10:33
*** karthikpr has joined #openstack-watcher10:33
pksinghvincentfrancoise: i was looking at,10:33
*** thorst has joined #openstack-watcher10:34
pksinghvincentfrancoise: can we create some conf file and move this to that file?10:34
vincentfrancoisepksingh: no we cannot because these configs are dynamically loaded10:34
vincentfrancoisepksingh: all plugins have this dynamic config system10:35
pksinghvincentfrancoise: so if i need to use datasource at different points in different strategies then do i need to add them to all places?10:35
vincentfrancoisepksingh: as of now, yes10:36
vincentfrancoisepksingh: this is why the abstraction layer for datasources was discussed at the PTG10:36
pksinghvincentfrancoise: can't i define in oslo_config at common place under conf directory and access it here?10:37
vincentfrancoisepksingh: because right now both the configuration and the implementation are defined within each and every plugin10:37
*** karthikpr has quit IRC10:38
*** thorst has quit IRC10:38
vincentfrancoisepksingh: the thing you might do is to use string interpolation10:39
vincentfrancoisepksingh: you define a datasource parameter as a static parameter and use string interpolation as default value in the strategies10:39
pksinghvincentfrancoise: and they are loaded by in conf directory?10:39
vincentfrancoisepksingh: the dynamic config options you mean? yes10:41
pksinghvincentfrancoise: so if i add more config options in future then adding them to each plugin/strategy would cause a lot of duplicate work10:41
pksinghvincentfrancoise: yes i mean dynamic config options,10:42
vincentfrancoisepksingh: it's not duplicate work because plugins are not just strategy plugins so it will only feel duplicated for the "strategies" subset10:43
pksinghvincentfrancoise: yes i mean if that config is applicable only for strategies, then i have to add that to each strategy in get_config_opts?10:44
vincentfrancoisepksingh: and also, how would you make the difference between a parameter that is specific to a strategy and one that is actually shared with other?10:45
vincentfrancoisepksingh: ah10:45
vincentfrancoisepshedimb: codewise you don't actually need to duplicate the code10:45
vincentfrancoisejust add the common parameter to
pksinghvincentfrancoise: that's what i wanted, thanks10:47
*** alexchadin has joined #openstack-watcher10:48
vincentfrancoisepksingh: then make sure you do a get_config_opts(cls): return super(cls, MyStrategy) + [other params]10:48
vincentfrancoisein **ALL** the subclasses10:48
vincentfrancoiseotherwise, some parameters will be ignored10:48
pksinghvincentfrancoise: yes you are right10:49
*** openstackgerrit has joined #openstack-watcher10:52
openstackgerritBéla Vancsics proposed openstack/watcher master: Reduced the code complexity
*** pksingh has quit IRC10:55
*** sanfern has quit IRC11:19
*** alexchadin has quit IRC11:28
*** alexchadin has joined #openstack-watcher11:44
*** zhurong has quit IRC11:58
*** alexchadin_ has joined #openstack-watcher12:07
*** alexchadin_ has quit IRC12:07
*** sanfern has joined #openstack-watcher12:13
*** karthikpr has joined #openstack-watcher12:16
*** alexchadin has quit IRC12:17
*** karthikpr has quit IRC12:20
*** alexchadin has joined #openstack-watcher12:21
*** thorst has joined #openstack-watcher12:44
*** aovchinnikov has quit IRC13:04
edleafeaovchinnikov: alexchadin: Small steps:
alexchadinthank you, edleafe, I will take a look13:45
*** pksingh has joined #openstack-watcher14:20
*** pksingh has quit IRC14:41
*** alexchadin has quit IRC15:01
pshedimbvincentfrancoise, Yeah. Not codewise. But I meant metric collection in the beginning.15:06
*** karthikpr has joined #openstack-watcher15:12
*** karthikpr has quit IRC15:16
*** alexchadin has joined #openstack-watcher15:16
*** karthikpr has joined #openstack-watcher15:16
*** karthikpr has quit IRC15:18
*** alexchadin has quit IRC15:34
*** danpawli1 has quit IRC16:13
*** vmahe has left #openstack-watcher16:23
*** karthikpr has joined #openstack-watcher17:18
*** vincentfrancoise has quit IRC17:21
*** karthikpr has quit IRC17:44
*** Guest89812 has joined #openstack-watcher18:19
*** wootehfoot has joined #openstack-watcher18:19
*** Guest89812 has quit IRC18:24
*** karthikpr has joined #openstack-watcher18:42
*** sanfern has quit IRC18:46
*** karthikpr has quit IRC18:46
*** karthikpr has joined #openstack-watcher19:04
*** wootehfoot has quit IRC20:59
*** karthikpr has quit IRC21:49
*** karthikpr has joined #openstack-watcher21:55
*** wootehfoot has joined #openstack-watcher21:58
*** thorst has quit IRC22:04
*** thorst has joined #openstack-watcher22:05
*** karthikpr has quit IRC22:08
*** thorst has quit IRC22:09
*** sanfern has joined #openstack-watcher22:34
openstackgerritChris Spencer proposed openstack/watcher-specs master: Define grammar for workload characterization
*** wootehfoot has quit IRC23:14

Generated by 2.14.0 by Marius Gedminas - find it at!