| *** thorst has joined #openstack-watcher | 00:14 | |
| *** thorst has quit IRC | 00:15 | |
| *** thorst has joined #openstack-watcher | 00:17 | |
| *** jwcroppe has joined #openstack-watcher | 00:17 | |
| *** thorst has quit IRC | 00:19 | |
| *** Tamayo has joined #openstack-watcher | 00:19 | |
| *** sanfern has joined #openstack-watcher | 00:31 | |
| *** jwcroppe has quit IRC | 00:51 | |
| *** thorst has joined #openstack-watcher | 01:19 | |
| *** thorst has quit IRC | 01:24 | |
| *** hidekazu has joined #openstack-watcher | 01:33 | |
| *** ianychoi_ is now known as ianychoi | 01:35 | |
| *** sanfern has quit IRC | 01:59 | |
| *** thorst has joined #openstack-watcher | 02:13 | |
| *** thorst has quit IRC | 02:13 | |
| *** thorst has joined #openstack-watcher | 02:37 | |
| *** thorst has quit IRC | 02:38 | |
| *** sanfern has joined #openstack-watcher | 02:43 | |
| *** nicolasbock has quit IRC | 02:49 | |
| openstackgerrit | Hidekazu Nakamura proposed openstack/watcher master: Cinder model integration https://review.openstack.org/450642 | 03:01 |
|---|---|---|
| *** thorst has joined #openstack-watcher | 03:09 | |
| *** nicolasbock has joined #openstack-watcher | 03:20 | |
| *** thorst has quit IRC | 03:26 | |
| *** nicolasbock has quit IRC | 03:33 | |
| *** zhurong has joined #openstack-watcher | 04:15 | |
| *** zhurong has quit IRC | 04:19 | |
| *** Tamayo has quit IRC | 04:20 | |
| *** adisky__ has joined #openstack-watcher | 05:03 | |
| adisky__ | hidekazu: hii | 05:04 |
| hidekazu | adisky_: hi | 05:04 |
| adisky__ | regarding your comment in https://review.openstack.org/#/c/458492/4/watcher/applier/workflow_engine/base.py | 05:06 |
| adisky__ | for ABORT_TRUE | 05:07 |
| adisky__ | i have replied to it | 05:08 |
| adisky__ | what you feel now, we should add it or remove it?? | 05:08 |
| hidekazu | i want to here other reviewers opinion | 05:11 |
| hidekazu | but at least variables in base.py is not good. | 05:12 |
| adisky__ | hidekazu: ya for that i feel the same, for now i am moving to watcher/objects/action.py | 05:18 |
| adisky__ | and lets have other reviewrs opinion on that. | 05:18 |
| hidekazu | adisky__: BTW, could you help my unit test failure? | 05:20 |
| hidekazu | adisky__: https://review.openstack.org/#/c/450642/ | 05:21 |
| hidekazu | adisky__: in my env, this test ends with success, but failure in gate. | 05:22 |
| adisky__ | hidekazu: i will try | 05:22 |
| adisky__ | i am checking it | 05:22 |
| hidekazu | adisky__: please tox in your environment. | 05:22 |
| adisky__ | hidekazu: ok | 05:22 |
| hidekazu | adisky__: thanks. | 05:23 |
| *** thorst has joined #openstack-watcher | 05:23 | |
| *** thorst has quit IRC | 05:28 | |
| *** sanfern has quit IRC | 05:43 | |
| *** sanfern has joined #openstack-watcher | 05:44 | |
| adisky__ | hidekazu: py27 is succeded in my environment as well | 05:47 |
| hidekazu | adisky__: oh, thanks.. | 05:48 |
| adisky__ | hidekazu: :) | 05:49 |
| openstackgerrit | Hidekazu Nakamura proposed openstack/watcher master: Replace default cinder endpoint type https://review.openstack.org/469336 | 06:02 |
| *** thorst has joined #openstack-watcher | 06:24 | |
| *** thorst has quit IRC | 06:29 | |
| *** sanfern has quit IRC | 06:56 | |
| *** sanfern has joined #openstack-watcher | 06:57 | |
| *** vincentfrancoise has joined #openstack-watcher | 07:39 | |
| openstackgerrit | aditi sharma proposed openstack/watcher master: Cancel Action Plan https://review.openstack.org/458492 | 07:42 |
| *** thorst has joined #openstack-watcher | 07:46 | |
| *** thorst has quit IRC | 07:50 | |
| *** efoley__ has joined #openstack-watcher | 08:09 | |
| *** alexchadin has joined #openstack-watcher | 08:09 | |
| *** hidekazu has quit IRC | 08:36 | |
| openstackgerrit | aditi sharma proposed openstack/watcher master: Watcher official install-guide https://review.openstack.org/463281 | 08:40 |
| *** thorst has joined #openstack-watcher | 08:47 | |
| *** vincentfrancoise has quit IRC | 08:47 | |
| *** vincentfrancoise has joined #openstack-watcher | 08:48 | |
| openstackgerrit | Merged openstack/watcher master: Replace oslo_utils.timeutils.isotime https://review.openstack.org/468826 | 08:49 |
| openstackgerrit | Merged openstack/watcher master: Deleted audit record still get by 'audit list'cmd https://review.openstack.org/468271 | 08:54 |
| openstackgerrit | Merged openstack/watcher master: Trivial fix typos https://review.openstack.org/469017 | 08:54 |
| *** vincentfrancoise has quit IRC | 09:01 | |
| *** andreaf has quit IRC | 09:03 | |
| *** thorst has quit IRC | 09:06 | |
| *** andreaf has joined #openstack-watcher | 09:06 | |
| *** alexchadin has quit IRC | 09:46 | |
| *** nicolasbock has joined #openstack-watcher | 09:56 | |
| *** vinaykns has joined #openstack-watcher | 10:01 | |
| *** vinaykns has quit IRC | 10:05 | |
| *** vincentfrancoise has joined #openstack-watcher | 10:12 | |
| *** sanfern has quit IRC | 10:15 | |
| *** sanfern has joined #openstack-watcher | 10:17 | |
| openstackgerrit | licanwei proposed openstack/watcher master: Add action description https://review.openstack.org/450720 | 10:21 |
| openstackgerrit | licanwei proposed openstack/watcher master: Add action description notification https://review.openstack.org/454638 | 10:21 |
| *** nicolasbock has quit IRC | 10:27 | |
| *** nicolasbock has joined #openstack-watcher | 10:27 | |
| *** nicolasbock has quit IRC | 10:28 | |
| *** nicolasbock has joined #openstack-watcher | 10:28 | |
| *** sanfern has quit IRC | 10:51 | |
| *** thorst has joined #openstack-watcher | 11:04 | |
| *** thorst has quit IRC | 11:08 | |
| *** thorst has joined #openstack-watcher | 11:30 | |
| *** sanfern has joined #openstack-watcher | 12:17 | |
| openstackgerrit | YumengBao proposed openstack/watcher master: please ignore just for test https://review.openstack.org/469455 | 12:31 |
| openstackgerrit | YumengBao proposed openstack/watcher master: just for test please ignore https://review.openstack.org/469458 | 12:39 |
| *** efoley_ has joined #openstack-watcher | 13:03 | |
| *** dpawlik has joined #openstack-watcher | 13:05 | |
| *** efoley__ has quit IRC | 13:07 | |
| *** danpawlik has quit IRC | 13:09 | |
| *** jwcroppe has joined #openstack-watcher | 13:10 | |
| *** lakerzhou has joined #openstack-watcher | 13:14 | |
| *** efoley_ has quit IRC | 13:19 | |
| *** efoley_ has joined #openstack-watcher | 13:20 | |
| *** efoley has joined #openstack-watcher | 13:21 | |
| *** efoley_ has quit IRC | 13:21 | |
| *** efoley_ has joined #openstack-watcher | 13:39 | |
| *** zhurong has joined #openstack-watcher | 13:42 | |
| *** efoley has quit IRC | 13:42 | |
| *** alexchadin has joined #openstack-watcher | 13:51 | |
| *** Yumeng__ has joined #openstack-watcher | 13:55 | |
| alexchadin | we start in 3 minutes on #openstack-meeting-4 channel | 13:57 |
| vincentfrancoise | alexchadin: I'll be a few minutes late | 13:57 |
| alexchadin | vincentfrancoise: ok | 13:58 |
| sballe_ | morning | 14:00 |
| openstackgerrit | Merged openstack/watcher-specs master: Add spec for Supporting HA for background jobs in Watcher https://review.openstack.org/447996 | 14:17 |
| openstackgerrit | Merged openstack/watcher-specs master: Add spec for services-versioned-notifications-api bp https://review.openstack.org/445817 | 14:23 |
| *** vinaykns has joined #openstack-watcher | 14:31 | |
| *** zhurong has quit IRC | 14:35 | |
| *** vincentfrancoise has quit IRC | 14:48 | |
| *** vincentfrancoise has joined #openstack-watcher | 14:53 | |
| *** alexchadin has quit IRC | 14:59 | |
| lakerzhou | sballe_, will a new strategy be added for noisy neighbor, or there is already one under review? | 15:03 |
| aspiers | hi guys, has anyone investigated linear programming as a way of implementing a strategy for consolidating VMs onto a minimal number of compute hosts? | 15:05 |
| aspiers | I've done quite a bit of linear programming recently for a completely different problem space (totally unrelated to work) but somehow I didn't notice the similarity between the two until my friend pointed it out just now :) | 15:06 |
| lakerzhou | aspiers, what are the constraints you consider when consolidating the VMs? | 15:08 |
| aspiers | lakerzhou: mostly RAM / CPU / disk capacity and overcommit ratios on the host | 15:13 |
| aspiers | they're extremely straightforward linear constraints! | 15:13 |
| aspiers | I can't believe I didn't realise this earlier | 15:13 |
| lakerzhou | those are very basic ones | 15:14 |
| lakerzhou | there are also affinity/anti-affinity, NUMA, and etc | 15:14 |
| aspiers | lakerzhou: sure, those could easily be added | 15:14 |
| lakerzhou | sounds interesting | 15:15 |
| aspiers | I already implemented affinity/anti-affinity in my other (non-cloud) solution, in fact | 15:15 |
| aspiers | in that context, it was assignment of people to teams | 15:15 |
| lakerzhou | I am not sure if ppl have been done similar work in this project | 15:15 |
| aspiers | acabot told me that some Ph.D. mathematicians or similar had looked at this kind of stuff in the past | 15:16 |
| *** Yumeng__ has left #openstack-watcher | 15:16 | |
| *** vincentfrancoise has quit IRC | 15:17 | |
| lakerzhou | I am interested in the work personally. I implemented VM placement algorithms using heuristic and Genetic Algorithms | 15:18 |
| aspiers | a few days ago I did some research into the various F/OSS MIP solvers out there | 15:18 |
| *** vincentfrancoise has joined #openstack-watcher | 15:18 | |
| aspiers | the TL;DR conclusion was: lp_solve < glpk < COIN-OR cbc | 15:18 |
| aspiers | cbc can't compete with commercial MIP Solvers, but it's still pretty good | 15:19 |
| aspiers | even glpk is usable for smaller problems | 15:19 |
| aspiers | the number of variables =~ (num of compute nodes) * (num of VMs) | 15:19 |
| aspiers | so for a small cloud it should be a small problem | 15:20 |
| aspiers | when it gets to CERN size, maybe a commercial solver would be needed :) | 15:20 |
| lakerzhou | sound reasonable to me :) | 15:20 |
| aspiers | the objective function could not only minimise the number of compute hosts used, but also the number of VM migrations required to reach that placement | 15:21 |
| aspiers | hmm, that would require more variables | 15:21 |
| aspiers | but still, this sounds WELL worth investigation, if it wasn't already done | 15:21 |
| lakerzhou | minimizing migrations is important | 15:22 |
| aspiers | yes for sure | 15:22 |
| aspiers | http://blog.adamspiers.org/2015/05/17/cloud-rearrangement/#unnumbered-3 | 15:22 |
| aspiers | oh, optimisation for KSM would be a big win | 15:23 |
| lakerzhou | The problem might be solved in the industry but I don't see publications | 15:23 |
| aspiers | I'll ping Martin the compsci professor who briefly visited this channel | 15:23 |
| lakerzhou | sure, please feel free to ping me at lakerzhou@gmail if you want more discussion | 15:28 |
| sballe_ | lakerzhou: the noisy neighbor code is already under review and the spec is also under review | 15:43 |
| *** vincentfrancoise has quit IRC | 15:46 | |
| lakerzhou | sballe_, will a new doc be added here: https://github.com/openstack/watcher/tree/master/doc/source/strategies ? | 15:59 |
| *** jwcroppe has quit IRC | 16:16 | |
| *** vincentfrancoise has joined #openstack-watcher | 16:44 | |
| *** vincentfrancoise has quit IRC | 16:49 | |
| *** efoley_ has quit IRC | 16:59 | |
| *** jwcroppe has joined #openstack-watcher | 17:04 | |
| *** wootehfoot has joined #openstack-watcher | 17:22 | |
| *** vinaykns has quit IRC | 17:23 | |
| aspiers | lakerzhou: ok thanks! | 17:35 |
| lakerzhou | aspiers, look forward to hearing more from you. Nice blog BTW. | 17:36 |
| aspiers | thanks :) | 17:37 |
| aspiers | lakerzhou: do you or anyone else in the Watcher team have access to real-world cloud arrangement metrics at a reasonable scale? a good starting point would simply be the amount of RAM and CPU for each VM and compute host in a cloud, plus maybe the current placement of VMs in that cloud | 17:38 |
| aspiers | I'm guessing Watcher already collects this stuff and could make it easily available? | 17:38 |
| aspiers | I would need some sample data to feed into a LP model | 17:38 |
| aspiers | and then if the LP model found a good solution, I could feed it through my algorithm detailed in that blog, and we'd have a complete server consolidation strategy :) | 17:39 |
| aspiers | lakerzhou: just watched your batch scheduling lightning talk from Boston. I like it! makes a lot of sense to me. Presumably the same algorithms could be used not just for batch scheduling but also for rearrangement | 17:46 |
| aspiers | Ah, you covered that in the last slide :) | 17:46 |
| aspiers | lakerzhou: I have no idea how genetic / simulated annealing algorithms would perform relative to MIP | 17:47 |
| aspiers | lakerzhou: but I think it's so easy to try MIP that it's worth it anyway | 17:47 |
| lakerzhou | I can provide GA implementation if you are ready to make a comparison | 17:47 |
| aspiers | sounds good! | 17:50 |
| aspiers | OK I emailed you | 17:51 |
| aspiers | I have to go now | 17:51 |
| lakerzhou | aspiers, Nice speak with you. talk to you late. | 17:51 |
| aspiers | later :) | 17:51 |
| *** nicolasbock has quit IRC | 17:58 | |
| *** sanfern has quit IRC | 18:12 | |
| *** adisky__ has quit IRC | 19:01 | |
| *** wootehfoot has quit IRC | 20:16 | |
| *** wootehfoot has joined #openstack-watcher | 20:16 | |
| *** wootehfoot has quit IRC | 20:54 | |
| *** wootehfoot has joined #openstack-watcher | 21:11 | |
| *** thorst has quit IRC | 21:17 | |
| *** lakerzhou has quit IRC | 21:19 | |
| *** harlowja has quit IRC | 21:21 | |
| *** thorst has joined #openstack-watcher | 21:56 | |
| *** harlowja has joined #openstack-watcher | 22:06 | |
| *** wootehfoot has quit IRC | 22:10 | |
| *** Tamayo has joined #openstack-watcher | 22:34 | |
| *** jwcroppe has quit IRC | 22:48 | |
| *** thorst has quit IRC | 23:09 | |
| *** nicolasbock has joined #openstack-watcher | 23:16 | |
| *** thorst has joined #openstack-watcher | 23:20 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!