Friday, 2016-06-24

*** thorst_away has joined #openstack-watcher00:24
*** thorst_away has quit IRC02:16
*** thorst_away has joined #openstack-watcher02:17
*** thorst_away has quit IRC02:25
*** esberglu has quit IRC02:49
*** thorst_away has joined #openstack-watcher03:23
*** thorst_away has quit IRC03:30
openstackgerritTin Lam proposed openstack/watcher: Update docs links to docs.openstack.org  https://review.openstack.org/33372304:00
*** thorst_away has joined #openstack-watcher04:28
*** thorst_away has quit IRC04:35
*** thorst_away has joined #openstack-watcher05:33
*** thorst_away has quit IRC05:40
*** thorst_away has joined #openstack-watcher06:38
*** thorst_away has quit IRC06:45
*** vtech has quit IRC06:45
*** pkoniszewski has quit IRC07:03
*** vtech has joined #openstack-watcher07:08
*** pkoniszewski has joined #openstack-watcher07:12
openstackgerritMerged openstack/python-watcherclient: Updated from global requirements  https://review.openstack.org/33238507:21
openstackgerritMerged openstack/watcher: Updated from global requirements  https://review.openstack.org/33371107:21
*** vincentfrancoise has joined #openstack-watcher07:27
*** alexchadin has joined #openstack-watcher07:37
*** vtech_ has joined #openstack-watcher07:39
*** vtech has quit IRC07:39
*** thorst_away has joined #openstack-watcher07:43
*** thorst_away has quit IRC07:50
openstackgerritEdwin Zhai proposed openstack/watcher: Enable strategy parameters  https://review.openstack.org/29755607:54
vincentfrancoisehi alexchadin08:42
alexchadinvincentfrancoise: hi08:42
vincentfrancoiseI'm looking at the rally plugin but I can't run anything08:43
alexchadinCause we haven't merged rally support for now08:44
alexchadinIf you want, I can give you git patch:)08:45
alexchadinOnce you apply it you will able to run watcher scenarios in rally08:45
vincentfrancoiseI checked out your review to run it08:46
vincentfrancoiseCan you push the missing code for review then ?08:46
alexchadinI need to push that code to rally repo08:47
alexchadinTo do this I need to set rally gate up08:47
vincentfrancoiseI thought it was done like in http://docs.openstack.org/developer/rally/plugins/scenario_plugin.html08:47
vincentfrancoisebut with the code being hosted in the watcher codebase08:48
vincentfrancoiseI am using docker rally to run the tests08:48
alexchadinThey prefer to host code in their repo08:48
*** thorst_away has joined #openstack-watcher08:48
alexchadinto make it "upstream"08:48
vincentfrancoiseOh I see08:48
vincentfrancoisePut me as reviewer when you push for review that changeset then08:49
alexchadinvincentfrancoise: ok :)08:50
vincentfrancoisealso, can you also put the command lines into the README.rst so we can run rally tests locally?08:50
alexchadinwhat sort of commands?08:51
vincentfrancoisea simplified version of what is in http://docs.openstack.org/developer/rally/install.html08:52
vincentfrancoiseand probably a link to this page as well08:52
alexchadinOh08:52
alexchadinOf course08:53
vincentfrancoisebut the the command lines tuned to run the Watcher tests08:53
alexchadinI understand08:53
alexchadinwhile you are here08:53
alexchadinI have some questions:)08:53
vincentfrancoisewhat is it ?08:53
alexchadinregarding cont-opt08:53
vincentfrancoiseah yeah writing the comments on this was not easy so I was probably not explicit/clear enough08:54
alexchadinShould AuditHandler class be metaclass?08:54
vincentfrancoiseI think so08:55
*** thorst_away has quit IRC08:55
alexchadinokay08:55
alexchadinthen08:55
vincentfrancoiseAuditHandler is just to avoid too much code duplication08:55
alexchadinthe execite logic08:55
vincentfrancoiselike BaseAuditHandler would be a Java Interface08:55
alexchadinI see there is 2 way to get it refactored08:56
vincentfrancoisewhilst AuditHandler is the actual abstract class08:56
alexchadinhm08:56
alexchadinThe difference between metaclass and abstract class08:56
alexchadinmetaclass has logic in its methods08:57
alexchadinabstract class is something like interfaces in Python08:57
vincentfrancoiseOh I thought you were referring to the same thing since you make an abstract class via the six.add_metaclass() decorator08:58
alexchadinSo we are using six.add_metaclass on AuditHandler, right?08:59
vincentfrancoiseyes09:00
alexchadincool09:00
vincentfrancoiseBTW what do you think about doing it this way?09:00
alexchadinI like your idea with inheritance09:01
alexchadinBTW I have questons about execute method09:01
vincentfrancoise^ cool :)09:02
alexchadinIn both ways (oneshot, continuous) we have to update audit state, call planner to schedule, etc09:03
vincentfrancoiseah yes09:03
alexchadinThere could be duplicates in execute method09:03
alexchadin:)09:03
vincentfrancoiseI was thinking about this when writing my comments09:03
vincentfrancoiseI was thinking we could do some non abstract execute() method that call pre_execute() do_execute() and post_execute()09:04
vincentfrancoiseduplicate logic would then go into the pre and post with only the do_execute() differing09:05
alexchadinhm09:05
vincentfrancoisepre_execute() do_execute() and post_execute() would then be abstract in BaseAuditHandler09:05
vincentfrancoiseand pre_execute() and post_execute() could be implemented in AuditHandler09:06
vincentfrancoisewhat do you think?09:06
alexchadinthen do_execute would be empty for oneshot method since there is only one difference between them09:06
alexchadinif statement for continuous09:07
vincentfrancoiseoneshot would only do self.strategy_context.execute_strategy()09:07
vincentfrancoisewhile continuous would also self.strategy_context.execute_strategy() but also the extra bit09:07
alexchadinyes09:10
alexchadinThat's definitely what I wanted09:11
alexchadinYou confirmed my thoughts09:11
*** hvprash has joined #openstack-watcher09:12
*** hvprash has quit IRC09:17
vincentfrancoisecool :)09:18
*** thorst_away has joined #openstack-watcher09:52
*** thorst_away has quit IRC10:00
*** alexchadin has quit IRC10:14
*** alexchadin has joined #openstack-watcher10:14
*** vincentfrancoise has quit IRC10:16
alexchadinjed56: are you here?10:37
*** wootehfoot has joined #openstack-watcher10:51
*** acabot has quit IRC10:54
*** thorst_away has joined #openstack-watcher10:58
*** thorst_away has quit IRC11:05
*** acabot has joined #openstack-watcher11:06
*** thorst_away has joined #openstack-watcher11:07
*** thorst_away is now known as thorst_11:15
*** hvprash has joined #openstack-watcher11:41
*** hvprash has quit IRC11:47
*** hvprash has joined #openstack-watcher11:47
*** hvprash_ has joined #openstack-watcher11:48
*** hvprash has quit IRC11:52
*** hvprash_ has quit IRC12:00
*** hvprash has joined #openstack-watcher12:00
jed56alexchadin: hello12:05
jed56I'm here :)12:05
alexchadinjed56: hi:)12:06
alexchadinI think about layer where we should have our audit handlers12:07
jed56what do you mean ?12:08
alexchadinFor now we have auditendpoint which is RPC enpoint12:08
alexchadinI agree it is wrong way to store default.ContinuousAuditManager in endpoints12:08
alexchadinhttps://review.openstack.org/#/c/321411/6/watcher/decision_engine/manager.py12:08
jed56 yes we discussed that we vincentfrancois12:09
jed56okay12:10
alexchadinMaybe we should have something like AuditManager12:12
alexchadinWhere we can dispatch audit to the right handler12:12
jed56You don't want to put the switch logic in the AuditEndpoint ?12:14
jed56We can imagine to create an AuditManager if we have maybe more than two types of audits12:15
*** dtardivel has joined #openstack-watcher12:15
jed56https://github.com/openstack/watcher/blob/91951f3b012c10c21da2cde37cffbeba0208fc84/watcher/decision_engine/messaging/audit_endpoint.py#L4512:15
jed56It is sad that we can't easily implement inheritance with the openstack objects12:17
jed56We could create a OneshotAudit and ContinuousAudit and use polymorphism to decided what is the handler to use12:17
alexchadinjed56: If we put switch logic in the AuditEndpoint we need to know audit type12:18
*** vincentfrancoise has joined #openstack-watcher12:18
jed56why ?12:19
jed56You query the db to have this information12:19
jed56*can12:19
alexchadinyes12:19
alexchadinhm12:19
jed56There is a reason why you don't want to do that ?12:20
alexchadinThere isn't, but I always tried to reduce number of queries to the DB12:21
alexchadinTo make code a little faster12:21
jed56alexchadin: yes this is good idea to reduce the number of queries12:21
jed56let's me take a look to the code12:22
jed56IMHO, we can move the query in order to do it only once12:22
alexchadinWe can send audit object12:22
alexchadinto the execute function12:22
jed56so, the query is made by the DefaultStrategyContext class12:23
*** hvprash has quit IRC12:23
*** hvprash has joined #openstack-watcher12:23
jed56We can move this logic to the AuditManager12:24
jed56and like you said pass the object and not anymore the audit_uuid12:24
alexchadinjed56: yeap:)12:24
jed56I don't think this will cause problems12:25
alexchadinthank you for your help12:25
jed56welcome !12:25
vincentfrancoisealexchadin: I juste tried to run your rally tests again but I get errors12:35
*** hvprash has quit IRC12:35
alexchadinwhich types of errors?12:35
vincentfrancoisehttp://pastebin.com/utpHRgd612:35
alexchadinthank you for bug12:37
alexchadins/audit_templates/audit_template12:37
alexchadinI forgot to change it in basic.py12:37
alexchadinvincentfrancoise: have you changed it?12:39
*** alexchadin has quit IRC12:42
*** hvprash has joined #openstack-watcher12:57
vincentfrancoisealexchadin: http://pastebin.com/aGAaSCqx13:00
*** edleafe is now known as figleaf13:01
*** hvprash has quit IRC13:02
*** esberglu has joined #openstack-watcher13:19
*** hvprash has joined #openstack-watcher13:20
*** vtech_ has quit IRC13:23
*** hvprash_ has joined #openstack-watcher13:24
*** esberglu has quit IRC13:24
*** hvprash has quit IRC13:24
*** hvprash has joined #openstack-watcher13:27
*** dronshaw has joined #openstack-watcher13:30
*** hvprash_ has quit IRC13:30
figleafjed56: I've added your spec to the agenda for next Monday's scheduler meeting13:34
figleafjed56: https://wiki.openstack.org/wiki/Meetings/NovaScheduler#Agenda_for_next_meeting13:35
figleafjed56: I want to get a clear answer from the scheduler subteam on whether your approach is something that we should encourage or discourage13:35
figleafjed56: It's at 1400 UTC in #openstack-meeting-alt13:36
jed56figleaf: thanks you very much13:36
jed56we will be there with acabot13:36
figleafjed56: great!13:36
jed56thank you very much for that13:37
*** esberglu has joined #openstack-watcher13:39
jed56jinquan: hi13:39
jed56jinquan: do you feel confident to develop the configurable-weights-default-planner blueprint without specification ?13:40
jed56jinquan: IMHO, we didn't need a spec but if you need one I can write one.13:41
jed56jinquan: if you have any questions don't hesitate13:41
*** vtech has joined #openstack-watcher13:46
*** vtech has quit IRC13:55
*** ecelik has quit IRC13:57
*** jwcroppe has joined #openstack-watcher14:05
*** vtech has joined #openstack-watcher14:12
*** brunograz has quit IRC15:30
*** vtech has quit IRC15:33
*** vtech has joined #openstack-watcher15:54
*** wootehfoot has quit IRC15:57
*** vtech has quit IRC16:17
openstackgerritVincent Fran├žoise proposed openstack/watcher: Loadable Cluster Data Model Collectors  https://review.openstack.org/33398716:27
*** vincentfrancoise has quit IRC16:30
*** acabot has quit IRC16:50
*** acabot has joined #openstack-watcher17:01
*** jed56 has quit IRC17:15
*** dtardivel has quit IRC17:17
*** vtech has joined #openstack-watcher17:41
*** openstackgerrit has quit IRC18:03
*** openstackgerrit has joined #openstack-watcher18:03
*** vtech has quit IRC18:22
*** openstack has joined #openstack-watcher18:41
*** dronshaw has quit IRC19:21
*** esberglu has quit IRC19:26
*** esberglu has joined #openstack-watcher19:34
*** esberglu has quit IRC21:07
*** esberglu has joined #openstack-watcher21:14
*** esberglu has quit IRC21:14
*** thorst_ has quit IRC21:49
*** thorst has joined #openstack-watcher21:50
*** figleaf is now known as edleafe21:52
*** thorst has quit IRC21:58
*** harlowja has quit IRC22:17
*** harlowja has joined #openstack-watcher22:30
*** thorst has joined #openstack-watcher22:57
*** thorst has quit IRC23:04

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!