Tuesday, 2016-12-27

openstackgerritHidekazu Nakamura proposed openstack/watcher: Add additional depencencies of CentOS 7  https://review.openstack.org/41503800:36
*** stack_ has joined #openstack-watcher02:10
*** stack_ has left #openstack-watcher02:11
*** licanwei has joined #openstack-watcher02:12
openstackgerritHidekazu Nakamura proposed openstack/watcher: Update configuration document  https://review.openstack.org/41508804:02
*** hidekazu has joined #openstack-watcher04:09
*** atuly has quit IRC04:22
openstackgerritHidekazu Nakamura proposed openstack/watcher: Fix TypeError if no input_parameters added  https://review.openstack.org/41509604:52
*** vtech has joined #openstack-watcher05:37
*** vtech has quit IRC05:46
*** atuly has joined #openstack-watcher06:27
*** licanwei has left #openstack-watcher06:43
openstackgerritAlexander Chadin proposed openstack/watcher: New default planner  https://review.openstack.org/40699108:33
*** alexchadin has joined #openstack-watcher08:34
*** vincentfrancoise has joined #openstack-watcher08:39
openstackgerritMerged openstack/watcher: Fix reference http  https://review.openstack.org/41486509:26
openstackgerritMerged openstack/watcher: Add additional depencencies of CentOS 7  https://review.openstack.org/41503809:27
vincentfrancoisealexchadin: hi09:31
alexchadinvincentfrancoise: morning09:31
vincentfrancoisealexchadin: I was quickly going through your latest PS for the planner09:32
vincentfrancoiseand I was wondering why you added a new resize action09:32
vincentfrancoiseit seems to me it is not really related to this BP09:33
alexchadinvincentfrancoise: Resize may be need in some new strategies09:33
alexchadinvincentfrancoise: Yes, it isn't related to the BP directly09:34
vincentfrancoisecan you instead push this in a separate PS the meake the new planner code easier to read?09:34
alexchadinvincentfrancoise: but it shows how `parents` field is filled09:34
alexchadinvincentfrancoise: It will take some more time to separate it, but I think yes, I can do it09:35
alexchadinif it is necessary09:35
vincentfrancoiseoh let me see the parents09:35
alexchadinDavid said that my previous patch made actions centralized, not pluggable09:36
vincentfrancoiseit's just that I wasn't understanding the reason why you pushed it along with the planner09:36
alexchadinthat's why I made new `validate_parents` method and actions is now pluggable again09:37
alexchadinvincentfrancoise: you mean resize?09:37
vincentfrancoisealexchadin: yeah but now that you explain it I get why09:38
alexchadinvincentfrancoise: I didn't test applying of resize action, but planner build actions well09:40
vincentfrancoisethis PS is hard to review because I didn't follow all the talks that happened when I was on holidays :(09:40
alexchadinvincentfrancoise: the main goal is to make applying of actions in parallel09:41
alexchadinvincentfrancoise: we have separate this goal to two BP: modifying of planner and changes to applier09:42
vincentfrancoisealexchadin: yeah I remember that09:42
alexchadinvincentfrancoise: I've uploaded first BP that adds `parents` field to action table that will have list of action uuids09:43
alexchadinvincentfrancoise: but, some actions shouldn't be linked09:43
alexchadinvincentfrancoise: for example, migration, nop, sleep, change_nova_node_status - all of them can be done without any dependencies09:44
alexchadinvincentfrancoise: but there are resize and turn_node_to_acpi_s3 (it isn't defined as action type yet) actions that should have parents09:45
alexchadinresize of specified action must go after migration of this action09:45
alexchadinand turn_node_to_acpi_s3 must go only when all migration and resize actions are done to specified host09:46
vincentfrancoisealexchadin: so validate_parents() will be the acting according the parents types?09:47
alexchadinvincentfrancoise: yeap09:47
alexchadinvincentfrancoise: I've added dummy_with_resize strategy, take a look at it09:47
vincentfrancoisealexchadin: does it have a tempest test to run the resize?09:50
alexchadinvincentfrancoise: not yet09:50
vincentfrancoisealexchadin: ok09:50
*** alexchadin has quit IRC10:16
*** alexchadin has joined #openstack-watcher10:16
alexchadinvincentfrancoise: does review.openstack.org work good for you?10:22
openstackgerritAlexander Chadin proposed openstack/watcher: New default planner  https://review.openstack.org/40699110:22
vincentfrancoisealexchadin: yeap10:23
vincentfrancoisealexchadin: there is something I want to double check before telling you10:23
alexchadinvincentfrancoise: yes?10:23
vincentfrancoisealexchadin: about using the action factory within the planner10:23
vincentfrancoisealexchadin: if you use the action factory to load the action plugins within the planner10:24
vincentfrancoisealexchadin: that means you need to install an action both where you have the decision engine and where the applier are running10:24
alexchadinbut we usually install all the code despite of services we plan to use10:25
vincentfrancoisealexchadin: so the separation of concern is not full here10:25
alexchadinwe have full nova code set on compute node, along with conductor, api, etc10:26
vincentfrancoisealexchadin: in practice yes but that's the idea behind using action dicts within the strategies in the first place10:27
vincentfrancoisealexchadin: like https://github.com/openstack/watcher/blob/master/watcher/decision_engine/strategy/strategies/basic_consolidation.py#L295-L297 does not reference the action object on purpose10:28
alexchadinI see10:28
vincentfrancoisealexchadin: the idea is that api, decision_engine and applier should be considered as separate projects10:29
vincentfrancoiseeven if under the same repo10:29
alexchadinwe can move actions to the common code set:)10:29
alexchadinlike it is done with objects10:29
vincentfrancoisealexchadin: action is already in 'objects'10:30
alexchadinI mean, to move watcher/applier/actions folder somewhere10:31
vincentfrancoisealexchadin: that's why I was telling you that I wanted to double check10:33
vincentfrancoisealexchadin: it's because I don't have a good solution to propose you yet :p10:33
alexchadinmerry christmas:)10:34
vincentfrancoisealexchadin: merry christmas to you too :)10:34
alexchadinvincentfrancoise: we can move watcher/watcher/applier/actions to the watcher/watcher/, on the same level where are notifications, objects, tests..10:46
vincentfrancoisealexchadin: we could but I do not like the idea of making actions a first-order package10:47
alexchadinvincentfrancoise: me too10:47
vincentfrancoisealexchadin: as Time Peters says: Special cases aren't special enough to break the rules. :p10:48
alexchadinsoooo we won't make any changes:D10:48
vincentfrancoisealexchadin: that's why I'm thinking on how to make this all work10:49
vincentfrancoisewill come back to you if I find anything10:49
alexchadinI will be back in one hour estimate10:49
*** alexchadin has quit IRC10:49
vincentfrancoiseI'll be on lunch break soon so if I come back to you that would be in ~3h I guess10:50
*** alexchadin has joined #openstack-watcher13:11
*** alexchadin1 has joined #openstack-watcher13:14
*** alexchadin has quit IRC13:18
*** alexchadin1 has quit IRC13:18
*** alexchadin has joined #openstack-watcher13:19
vincentfrancoisealexchadin: ping13:21
alexchadinvincentfrancoise: pong13:26
vincentfrancoisealexchadin: where is turn_host_to_acpi_s3_state turn_host_to_acpi_s3_state?13:27
vincentfrancoisealexchadin: I can't see any new action with this name13:27
alexchadinvincentfrancoise: so, there is no action yet13:28
alexchadinvincentfrancoise: cause there is no strategy that would use it soon13:29
alexchadinvincentfrancoise: I've leaved it just to show how parents field should work13:29
vincentfrancoisealexchadin: IMHO we should have a strategy that takes as parameter an explicit action to execute13:30
vincentfrancoisealexchadin: that way we could easily test all the actiona13:30
vincentfrancoiseinstead of doing a new dummy_with* everytime13:31
alexchadinvincentfrancoise: maybe13:31
vincentfrancoisebut that's just a nice to have13:32
alexchadinvincentfrancoise: something like "fullset" strategy:)13:32
*** alexchadin has quit IRC14:04
*** lamt has joined #openstack-watcher14:23
*** openstackgerrit_ has joined #openstack-watcher15:55
*** openstackgerrit_ has quit IRC15:57
openstackgerritMerged openstack/watcher: Fix TypeError if no input_parameters added  https://review.openstack.org/41509616:21
*** vincentfrancoise has quit IRC17:57
*** vtech has joined #openstack-watcher22:57
*** vtech has quit IRC23:21
*** hidekazu has quit IRC23:34
*** lamt has quit IRC23:36

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