Wednesday, 2016-06-01

*** thorst has joined #openstack-watcher00:07
*** wootehfoot has quit IRC00:22
*** jwcroppe has quit IRC00:41
*** jwcroppe has joined #openstack-watcher00:50
*** aspiers has quit IRC00:54
*** aspiers has joined #openstack-watcher00:56
*** thorst has quit IRC02:20
*** thorst has joined #openstack-watcher02:21
*** thorst has quit IRC02:29
*** thorst has joined #openstack-watcher02:35
*** thorst has quit IRC02:37
*** thorst has joined #openstack-watcher03:38
*** thorst has quit IRC03:46
*** thorst has joined #openstack-watcher04:44
*** thorst has quit IRC04:51
*** ecelik has joined #openstack-watcher05:38
*** thorst has joined #openstack-watcher05:49
*** thorst has quit IRC05:57
*** jwcroppe has quit IRC06:51
*** thorst has joined #openstack-watcher07:14
*** acabot_ has joined #openstack-watcher07:14
*** acabot_ has quit IRC07:15
openstackgerritEdwin Zhai proposed openstack/python-watcherclient: Enable strategy parameter
*** thorst has quit IRC07:22
openstackgerritEdwin Zhai proposed openstack/watcher: Enable strategy parameters
*** danpawlik has joined #openstack-watcher07:26
*** vincentfrancoise has joined #openstack-watcher07:34
*** apoorv has joined #openstack-watcher07:43
*** jwcroppe has joined #openstack-watcher07:56
*** jwcroppe has quit IRC08:04
*** alexchadin has joined #openstack-watcher08:11
*** apoorv has quit IRC08:18
*** thorst has joined #openstack-watcher08:20
*** dtardivel has joined #openstack-watcher08:22
*** thorst has quit IRC08:27
*** jed56 has joined #openstack-watcher08:45
* acabot is preparing the agenda08:46
*** apoorv has joined #openstack-watcher08:58
alexchadinI integrate watcher on testing cluster09:24
alexchadinwhich includes two types of hypervisors: KVM, VMware09:25
*** thorst has joined #openstack-watcher09:25
alexchadinI don't want to take VMware host into account when using strategies09:25
vincentfrancoisewouldn't cause migration issues if you have different kinds of hypervisors?09:25
alexchadinvincentfrancoise: yes, It would09:25
alexchadinI propose to add config option or something else09:26
alexchadinto specify which hosts we get from get_all_hypervisors() function09:26
vincentfrancoiseshouldn't we also specify the hypervisor type in the model ?09:26
alexchadinYes, we should09:27
vincentfrancoiseand then the only thing left to do would be to filter out the VMware hypervisors09:27
alexchadinwill we add it to upstream?09:29
jed56alexchadin: if you don't wants to09:30
jed56take into account some hypervisors you should specify that in the 'audit scope'09:30
vincentfrancoise ?09:30
*** thorst has quit IRC09:32
jed56alexchadin: IMHO, we should only change the action 'migrate'09:32
jed56in most of the cases, the strategy don't care that the hypervisor  is KVM or Vmware09:33
alexchadinjed56: that's it, but it tries to collect ceilometer info from vmware host09:33
alexchadinwhich doesn't have it09:34
jed56This is the responsibility of the 'Action' to manage  heterogeneous metrics09:34
jed56alexchadin: what do you mean ?09:34
alexchadinjed56: when I launch sd-strategy, it gets all hypervisors and its statistics per host09:35
alexchadinwe don't have statistics from vmware host09:35
jed56you specify new plugin in ceilometer :)09:36
jed56by the way, did you have the time to look at monasca or gnocchi ?09:37
alexchadinI didn't:(09:38
alexchadinI will look them this evening, right?09:38
jed56alexchadin: thanks09:42
jed56did you understand by point of view regarding the metrics ?09:42
alexchadinshould we use define-the-audit-scope or change the action 'migrate'?09:44
*** vincentfrancoise has quit IRC10:22
*** alexchad_ has joined #openstack-watcher10:23
*** alexchadin has quit IRC10:25
*** thorst has joined #openstack-watcher10:29
*** thorst has quit IRC10:38
*** alexchad_ has quit IRC10:40
openstackgerritAlexander Chadin proposed openstack/watcher: Add fix for __init__() error
*** alexchadin has joined #openstack-watcher10:40
*** hvprash has joined #openstack-watcher10:42
*** hvprash has quit IRC10:42
*** hvprash has joined #openstack-watcher10:43
*** hvprash has quit IRC10:47
*** alexchadin has quit IRC11:16
jed56alexchadin : in your particular case you should modify migrate11:30
*** thorst has joined #openstack-watcher11:41
*** apoorv has quit IRC11:49
openstackgerritAlexander Chadin proposed openstack/watcher: Add fix for __init__() error
*** jwcroppe has joined #openstack-watcher12:00
*** jwcroppe has quit IRC12:05
acabotagenda for today
*** vincentfrancoise has joined #openstack-watcher12:11
*** hvprash has joined #openstack-watcher12:15
*** hvprash has quit IRC12:15
*** openstackgerrit has quit IRC12:18
*** openstackgerrit has joined #openstack-watcher12:18
danpawlikHello guys, I can tell you, that puppet-watcher module is in review:
*** danpawlik has quit IRC12:40
*** danpawlik has joined #openstack-watcher12:40
vincentfrancoisedanpawlik: awesome!12:41
acabotdanpawlik : looks great ! thanks a lot !12:53
danpawlikacabot: thanks. I must abandon this review and create new, because I should push first commit with files which are generated by cokkiecutter and then I can add my own files12:57
acabotI saw that ;-(12:59
*** hvprash has joined #openstack-watcher13:12
hvprashacabot_ et al. Congrats ! Great Job for having us in big tent13:20
vincentfrancoisehvprash: Hooray! That's quite a big achievement :)13:23
hvprashyes indeed ! Saw only this morning.. catching up from long weekend13:24
vincentfrancoisehvprash: same goes for us! That was quite the news when we arrived at the office :)13:25
hvprashthe team rocks :)13:25
vincentfrancoisetrue thing :)13:26
*** jwcroppe has joined #openstack-watcher13:35
*** seanmurphy has joined #openstack-watcher13:37
acabothvprash : thx !13:40
hvprashthx to you guys :)13:40
acabotI hope it will help you to get more resources from Walmart on the project ;-)13:41
*** brunograz has joined #openstack-watcher13:42
hvprashacabot, i am trying hard to get resources for this project. Hoping to have some good news in the coming weeks13:43
*** ecelik has quit IRC13:48
*** alexchadin has joined #openstack-watcher13:52
*** esberglu has joined #openstack-watcher13:57
danpawlikacabot: I have create new pull request: :) Now is only generated files using puppet-openstack bootstrap. Problem is, that this tool was not recently updated and I must "fix" initial commit  xD13:58
acabotI updated the agenda ;-)13:59
*** ostroverkhov has joined #openstack-watcher14:05
*** zarkone has joined #openstack-watcher14:42
zarkonehello all! Struggling with a watcher-decision-engine: for some reason, after I do `watcher audit-create`, decision-engine console remains silent even in debug.. what I've missed?14:43
vincentfrancoisezarkone: the watcher CLI just got overhauled last week14:44
vincentfrancoiseso the command changed a bit14:44
vincentfrancoiseyou can refer to to see the new commands14:45
vincentfrancoiseif you just type in "watcher"14:45
vincentfrancoiseyou should also enter the watcher prompt with command auto-completion ;)14:45
zarkonevincentfrancoise: thanks, but I have to use the 0.22 and corresponding watcher-client right now.. So, on my test env I see how decision engine create action plan, but on production env -- not.. Seems like something is different in config, but I can't catch what exactly14:46
zarkoneI can't wait till I could upgrade to >=0.26, by the way =)14:46
zarkonebut because of Liberty deps I can't do it right now14:47
vincentfrancoiseoh I see14:47
*** lrensing has joined #openstack-watcher14:47
vincentfrancoisecan you make a pastebin out of the exact command you are typing then? Maybe I can help14:48
*** seanmurphy has quit IRC14:59
tpeoplesjed56: what is this proof of concept you're talking about?15:01
*** brunograz has quit IRC15:01
*** alexchadin has quit IRC15:01
gzhai2vincentfrancoise: any question for strategy parameter check?15:01
jed56tpeoples: basically, the POC retrieve the request_spec of an instance.15:01
jed56Then, we can specify in the requestSpec a "requested_destination" ( which have been computed by watcher.15:02
jed56Finally, we can call select_destinations(), if the requested_destination satify the scheduler rules, the requested_host is returned otherwise the exception NoValidHost is raise.15:02
jed56There is an issue with this solution15:02
vincentfrancoisegzhai2: basically, we currently use voluptuous to validate parameters15:02
jed56 We need to apply directly the migration, otherwise Nova is not aware of this change.15:02
gzhai2vincentfrancoise: same in my code.15:02
jed56If we call again nova for validate the next action the previous one is not taking into account.15:02
tpeoplesjed56: hmmm. have you discussed that with bauzas?15:03
vincentfrancoisegzhai2: but in your case, we need to serialize the voluptuous Schema to save in the DB so it can be used by the Watcher API when the operator submits its inputs15:03
jed56tpeoples: yes, I spend 25 minutes with him today15:03
*** seanmurphy has joined #openstack-watcher15:03
jed56there are many bugs in nova :)15:03
vincentfrancoisegzhai2: so the solution would be for us to use jsonschema to perform the validation so we can pass the schema from the decision engine down to the API15:04
jed56for example : we an affinity rules you can't migrate any VMs15:04
gzhai2vincentfrancoise: Yes. As end user need read schema first before submitting parameter, right?15:04
jed56IMHo, one solution can be to have a specific session with the nova scheduler with a kind of  virtual cluster state which take into account the changes.15:04
vincentfrancoisegzhai2: BUT that means that a plugin developer will have to use voluptuous for strategy parameters but voluptuous for actions15:05
vincentfrancoisegzhai2: so 2 libs for the same thing is not good15:05
jed56tpeoples with powerSystem you are only performing one action  ?15:05
jed56if you use select_destination()15:05
jed56right ?15:06
vincentfrancoisegzhai2: the idea would be not just to read but to make the API validate the parameters based on the schema15:06
tpeoplesjed56: we only use live migrations currently15:06
tpeoplesbut... yes15:06
tpeoplesjust one at a time15:06
gzhai2vincentfrancoise: both are voluptuous, why 2 libs?15:06
jed56tpeoples: okay so in your case its working15:07
gzhai2vincentfrancoise: What's the benefit of API validation?15:07
vincentfrancoisegzhai2: no jsonschema is different from voluptuous15:07
vincentfrancoisegzhai2: things should be easy for the operator15:07
gzhai2vincentfrancoise: jsonschema for actions, right?15:07
vincentfrancoisegzhai2: no for strategy parameters ;)15:08
gzhai2vincentfrancoise: I can propagate the exception, so API can return error.15:08
vincentfrancoisegzhai2: so if he's typing an invalid parameter for his strategy15:08
gzhai2is it okay?15:08
vincentfrancoisegzhai2: the error would only be caught be seen upon triggering an audit15:08
vincentfrancoisegzhai2: I'll submit my reviews tomorrow so it will be clearer for you to understand my point15:09
tpeoplesjed56: Do you think at some point you could try to summarize what you find on a ML thread so everyone is on the same page?15:09
gzhai2yes. as "-p" is appended to 'audit create'15:09
gzhai2" BUT that means that a plugin developer will have to use voluptuous for strategy parameters  but voluptuous for actions. so 2 libs for the same thing is not good"15:10
jed56tpeoples: yes, of course. I will first  delve more in nova15:10
gzhai2not fully understand it. both are voluptuous, but 2 libs.15:11
vincentfrancoisegzhai2: jsonschema is not the same as voluptuous ( vs
vincentfrancoisethe structure of the schema dict is formatted differently15:13
*** ostroverkhov has quit IRC15:14
gzhai2vincentfrancoise: pls. upload your review. So I can check. Thank you.15:16
*** hvprash_ has joined #openstack-watcher15:16
gzhai2vincentfrancoise: I agree that API error is good for invalid parameter, but propergate exception can achieve it.15:17
gzhai2vincentfrancoise: BTW, do you agree that the only way to specify strategy paramter is appending "watche audit create"?15:19
*** hvprash has quit IRC15:20
*** danpawlik has quit IRC15:22
vincentfrancoisegzhai2: sorry I was afk15:46
vincentfrancoisegzhai2: you could potentially specify audit parameters in the audit template if and only if you have specified the strategy within it.15:47
vincentfrancoisegzhai2: For now, I would argue we should keep it simple and only expose the parameters at the audit level as you said15:48
*** ostroverkhov has joined #openstack-watcher15:52
*** seanmurphy has quit IRC16:09
*** ostroverkhov has quit IRC16:24
*** vincentfrancoise has quit IRC16:54
*** zarkone has quit IRC17:21
*** wootehfoot has joined #openstack-watcher17:48
*** esberglu has quit IRC18:27
*** dtardivel has quit IRC19:10
*** wootehfoot has quit IRC19:39
*** hvprash_ has quit IRC20:50
*** lrensing has quit IRC21:09
*** ostroverkhov has joined #openstack-watcher21:47
*** thorst has quit IRC22:23
*** thorst has joined #openstack-watcher22:30
*** thorst has quit IRC22:34
*** thorst has joined #openstack-watcher22:50
*** ostroverkhov has quit IRC22:52
*** thorst has quit IRC22:54
*** jwcroppe has quit IRC22:56

Generated by 2.14.0 by Marius Gedminas - find it at!