Thursday, 2016-12-15

*** lamt has quit IRC00:35
*** dtardivel has quit IRC01:04
*** harlowja has quit IRC03:03
*** atuly has joined #openstack-watcher03:05
*** atuly has quit IRC03:33
*** atuly has joined #openstack-watcher03:33
*** thorst_ has joined #openstack-watcher03:43
openstackgerritOpenStack Proposal Bot proposed openstack/watcher: Updated from global requirements
*** atuly has quit IRC04:09
*** thorst_ has quit IRC04:31
*** thorst_ has joined #openstack-watcher04:32
*** thorst_ has quit IRC04:40
*** thorst_ has joined #openstack-watcher05:10
*** thorst_ has quit IRC05:18
*** noama has quit IRC05:28
*** noama has joined #openstack-watcher05:28
*** thorst_ has joined #openstack-watcher06:06
*** thorst_ has quit IRC06:15
*** thorst_ has joined #openstack-watcher07:01
*** thorst_ has quit IRC07:09
openstackgerritlicanwei proposed openstack/python-watcherclient: too many digits after the decimal point
*** thorst_ has joined #openstack-watcher07:41
*** thorst_ has quit IRC07:49
*** atuly has joined #openstack-watcher08:10
*** thorst_ has joined #openstack-watcher08:35
*** alexchadin has joined #openstack-watcher08:37
openstackgerritAntoine Cabot proposed openstack/python-watcherclient: Remove readme reference because of sphinx error
*** thorst_ has quit IRC08:43
openstackgerritDavid TARDIVEL proposed openstack/watcher-dashboard: Moved README.rst content to installation.rst doc
openstackgerritMerged openstack/python-watcherclient: Remove readme reference because of sphinx error
openstackgerritDavid TARDIVEL proposed openstack/python-watcherclient: Support parents field along with planner changes
openstackgerritDavid TARDIVEL proposed openstack/python-watcherclient: too many digits after the decimal point
openstackgerritMerged openstack/watcher-dashboard: Moved README.rst content to installation.rst doc
openstackgerritAlexander Chadin proposed openstack/watcher-specs: Add improvements to the planner and workflow mechanisms
*** thorst_ has joined #openstack-watcher09:30
openstackgerritDavid TARDIVEL proposed openstack/watcher-dashboard: Show team and repo badges on README
*** thorst_ has quit IRC09:38
openstackgerritMerged openstack/watcher-dashboard: Show team and repo badges on README
*** exploreshaifali has joined #openstack-watcher10:09
*** openstackgerrit has quit IRC10:18
*** alexchad_ has joined #openstack-watcher10:21
*** alexchadin has quit IRC10:21
*** thorst_ has joined #openstack-watcher10:25
*** thorst_ has quit IRC10:32
*** hidekazu has quit IRC10:36
*** exploreshaifali has quit IRC10:38
alexchad_acabot: could you please invite David here?10:44
*** openstackgerrit has joined #openstack-watcher10:45
openstackgerritAlexander Chadin proposed openstack/python-watcherclient: Add functional tests to watcherclient
acabotalexchad_ : done10:46
*** dtardivel has joined #openstack-watcher10:46
*** alexchad_ is now known as alexchadin10:46
alexchadinacabot: thank you!10:47
alexchadindtardivel: hi!10:47
dtardivelalexchadin: hi10:47
alexchadindtardivel: I will fix tempest tests now, but about this note "Moreover, parents field is not set in db"10:47
alexchadindtardivel: they should be linked10:48
dtardivelok, I will fix dashboard issue if you want10:48
alexchadindtardivel: migrate and change_nova_service_state are independent10:48
alexchadinto each other10:49
dtardivelI ran a basic-consolidation on my devstack with 2 instances deployed. The computed action are corrected, by not linked10:49
openstackgerritAntoine Cabot proposed openstack/watcher-specs: Remove readme reference to fix Sphinx
alexchadindtardivel: it would be great since I'm not good with Django:)10:49
dtardivelalexchadin: parents is set to [] on both actions10:49
alexchadindtardivel: yes, it is normal use case10:49
alexchadindtardivel: these two actions will be run in parallel10:50
dtardivelwhaooou, I thought this feature will be implemented later10:51
alexchadindtardivel: They should be, but I didn't make any changes to applier yet10:52
alexchadindtardivel: Applier will run these actions in sequence for now10:52
alexchadindtardivel: do you remember, that we divided BP by two?10:53
dtardivelyes I remenber10:53
dtardivelone thing, when I run the action plan, I see actions are executed in chained way, but in opposite way10:57
dtardivelmigration are triggered before compute state updates. And before your patchset, it was the opposite10:58
atulyhi alexchadin10:59
alexchadinhi atuly11:00
alexchadindtardivel: I will be back in 5 minutes11:00
atulyI was reviewing11:00
atulyone basic question : weights_dict = {11:00
atuly        'turn_host_to_acpi_s3_state': 0,11:00
atuly        'resize': 1,11:00
atuly        'migrate': 2,11:00
atuly        'sleep': 3,11:00
atuly        'change_nova_service_state': 4,11:00
atuly        'nop': 5,11:00
atulycan we some meaning name for nop11:00
atulyi guess it means no_operational11:01
atulycorrect me if wrong ?11:01
alexchadinatuly: I think, you need to ask about it vincentfrancoise or jed56, cause when I came to watcher team it was already named that way:) I think you are correct11:04
atulyanyways if it not operator dependent and only specific within the file we may continue to use it as it is.11:07
alexchadindtardivel: ping11:08
*** thorst_ has joined #openstack-watcher11:19
alexchadindtardivel: now actions are adding to database in weight order11:23
*** thorst_ has quit IRC11:27
openstackgerritKevin Mullery proposed openstack/watcher: Graph cluster model implementation
*** chrisspencer has quit IRC12:06
*** chrisspencer has joined #openstack-watcher12:08
*** thorst_ has joined #openstack-watcher12:14
*** thorst_ has quit IRC12:21
openstackgerritAlexander Chadin proposed openstack/watcher: New default planner
*** alexchadin has quit IRC12:56
*** thorst_ has joined #openstack-watcher13:19
dtardivelatuly: nop means 'no operation'. it is just used for the dummy strategy for training13:47
*** lamt has joined #openstack-watcher14:10
openstackgerritDavid TARDIVEL proposed openstack/python-watcherclient: Fix TOKENID format which should without dashed
openstackgerritDavid TARDIVEL proposed openstack/watcher-dashboard: 'next_uuid' Action field is replaced by 'parents'
*** alexchadin has joined #openstack-watcher15:11
alexchadindtardivel: ping15:14
dtardivelalexchadin: yep15:14
alexchadindtardivel: I've just seen your response, thank you15:14
dtardivelis it clear enough ?15:14
alexchadinI agree that it is static implementation. But we need to control order of every action type and its pre-requisites/post-reqs. for example, change_state type is not related to migrate type. They can be done in parallel15:16
alexchadinMaybe, we can implement action types as plugins15:17
dtardivelyes I understand, but i would prefer (and vincentfrancoise as well), we treat with // in another spec/changeset.15:18
alexchadinBTW, turn_host_to_acpi_s3_state requires all resize/migrate actions related to the same host to be done before launching15:18
dtardivelThe general question: how can we set affinity/anti-affinyty rules between actions ?15:19
dtardivelI kwow beginning of // constraints has been introduced in your spec, but I think we should address it in  a generic way, including how to handle new action plugin dynamically .... and I have not the response today :)15:21
alexchadinwhy not to support it in static way now? we can create new BP if someone wants to make it pluggable15:23
alexchadinif we make it static now, we can implement parallel executing faster15:24
dtardivelMy concern in this case, is we will broke the possibility to extend Actions by plugin.15:26
dtardivelwe ask to acabot ....15:26
alexchadinactions are something that appear rarely, watcher team may discuss new action impact15:27
alexchadinI totally agree that we need pluggable actions15:27
alexchadinbut we can introduce it later, once we have some thoughts about it15:27
dtardivelaction are already pluggable today :)15:27
dtardivelalexchadin: I have again an issue with your implementation, if I run a basic consolidation with or without your patch. Applier does not run action in the same way15:29
dtardivelit's a opposite way with your patch, and it is not normal15:30
*** alexchadin has quit IRC15:30
*** alexchadin has joined #openstack-watcher15:30
alexchadinsorry for disconnecting15:30
alexchadincould you give me some logs?15:31
dtardivelfrom what ? the DE ?15:31
alexchadinyou say that actions run in other order?15:33
alexchadinin comparison with old version15:34
dtardivelas explain in my comment, on devstack, a basic consolidation (b4 your patch) build an action plan with 1.  change State, and 2 Migrate. And the applier runs these actions in this order.15:34
dtardivelWith your patch, the built action plan is composed by a change_state and a migrations actions, without parents. But Applier runs, in first, the migrate action and after the change_state action15:35
alexchadinI see the reason15:37
dtardiveldid you remember we found a bug in applier side, where this one didn't read 'next' field to build the taskflow workflow ?15:38
alexchadinApplier gets list of actions15:38
dtardivelapplier take the actions in the order theses ones have been save in db15:38
dtardiveland with your patch, migrate is the first action saved in db :)15:38
alexchadinand it is strange15:39
alexchadinsince we have this one: scheduled = sorted(to_schedule, key=lambda x: (x[0]), reverse=True)15:39
dtardivelyes but your update the weights as weel no ?15:39
alexchadinso, probably change_nova_service_state should be the first one15:39
dtardivelas well15:39
alexchadinlet me write new unit test:)15:42
dtardivelas the graph is built by 'parents' link, we should find the first entry action of the graph (the one(s) that has(ve) no parent(s)). is it not the algo of applier ?15:43
dtardivelvincentfrancoise will come back next monday. Will you at office next week as well ?15:44
alexchadindtardivel: yes, I will. take a look at this pseudo:
alexchadindtardivel: It is general algorithm of executing actions in applier15:45
dtardivelok, so in my use case, I have no parent for both actions15:47
dtardivelso no link in mainflow15:48
alexchadinso, they will be executed at once15:49
dtardivelalexchadin: I can not in fact test // action on devstack. I have only 2 actions in my action plans15:49
dtardivelI will try to test in within our OS infra in our lab.15:49
dtardivelI'm reviewing your last patch on auto trigger. It's look ok for me, I need to test it quickly on my devstack15:51
alexchadinIf I'm not online, write a message on gerrit15:52
alexchadinthank you so much:)15:52
dtardivelyou're welcome15:52
*** thorst_ has quit IRC15:58
*** thorst_ has joined #openstack-watcher16:02
openstackgerritMerged openstack/watcher-specs: Remove readme reference to fix Sphinx
openstackgerritDavid TARDIVEL proposed openstack/python-watcherclient: Add auto_trigger support
alexchadindtardivel: I tried to add change_state and migrate actions and I get this one:
dtardivelis it in DE or Applier level ?16:13
*** alexchadin has quit IRC16:13
*** alexchadin has joined #openstack-watcher16:13
*** michaelgugino has joined #openstack-watcher16:13
*** alexchad_ has joined #openstack-watcher16:14
alexchad_It is in DE level16:14
*** alexchadin has quit IRC16:14
*** jwcroppe has joined #openstack-watcher16:18
*** thorst_ has quit IRC16:23
openstackgerritAntoine Cabot proposed openstack/watcher-specs: Spec for Noisy Neighbor Startegy renamed cache occupancy strategy to noisy neighbor
openstackgerritAntoine Cabot proposed openstack/watcher-specs: Add improvements to the planner and workflow mechanisms
openstackgerritAntoine Cabot proposed openstack/watcher-specs: Support Description For Dynamic Action
*** thorst_ has joined #openstack-watcher16:33
*** alexchad_ has quit IRC17:29
*** thorst_ has quit IRC17:32
*** alexchadin has joined #openstack-watcher17:50
*** acabot has quit IRC18:11
*** harlowja has joined #openstack-watcher18:31
*** alexchadin has quit IRC18:46
*** harlowja has quit IRC18:52
*** thorst_ has joined #openstack-watcher18:59
*** wootehfoot has joined #openstack-watcher19:32
*** wootehfoot has quit IRC19:33
*** wootehfoot has joined #openstack-watcher19:33
*** dtardivel has quit IRC19:54
*** jwcroppe has quit IRC20:27
*** jwcroppe has joined #openstack-watcher20:28
*** lamt has quit IRC20:29
*** jwcroppe has quit IRC20:31
*** jwcroppe_ has joined #openstack-watcher20:44
*** wootehfoot has quit IRC21:00
*** michaelgugino has quit IRC21:13
*** wootehfoot has joined #openstack-watcher22:11
*** thorst_ has quit IRC23:05
*** thorst_ has joined #openstack-watcher23:15
*** wootehfoot has quit IRC23:15
*** harlowja has joined #openstack-watcher23:58

Generated by 2.14.0 by Marius Gedminas - find it at!