Wednesday, 2015-04-29

*** hitalia has quit IRC00:36
*** Aish has quit IRC00:46
*** madhu_ak has quit IRC00:47
*** bharath has quit IRC01:09
*** woodster_ has quit IRC01:10
*** clev is now known as clev-away01:18
*** Aish has joined #openstack-lbaas01:18
*** clev-away is now known as clev01:20
*** freelinuxer has joined #openstack-lbaas01:45
*** fnaval has quit IRC01:48
*** mwang2_ has quit IRC01:52
*** madhu_ak has joined #openstack-lbaas01:55
*** nuxxer has joined #openstack-lbaas01:56
*** fnaval has joined #openstack-lbaas02:00
openstackgerritPhillip Toohill proposed stackforge/octavia: Add devstack plugin for octavia
*** clev is now known as clev-away02:07
*** woodster_ has joined #openstack-lbaas02:13
openstackgerritPhillip Toohill proposed stackforge/octavia: Update nova_driver to use amp_network
*** clev-away is now known as clev02:21
*** Guest95 has joined #openstack-lbaas02:27
*** clev is now known as clev-away02:37
Guest95Hi. new to lbaas channel. trying to make LBaaS scalable..   how to make multiple lbaas agents running on multiple nodes ?02:38
Guest95If that's not possible, recommended HW spec to serve 10000 pools ?02:46
Guest95any link / doc / pointer would be appreciated02:47
*** nuxxer has quit IRC02:54
*** freelinuxer has quit IRC03:11
*** Guest95 has quit IRC03:11
*** nuxxer has joined #openstack-lbaas03:17
rm_youwell, if Guest95 had stuck around, we could have told him about Octavia :P03:21
nuxxer@rm_you   Octavia .. Is it scalable ?03:23
rm_youThat is the plan03:23
nuxxervery curious.. want to know. what's best possible way to find out the proj ?03:24
rm_youAh I see your hostmask now, you just rejoined03:24
rm_youWell, this is the channel for it03:24
nuxxerI just bought textual5 .. paid $5 .. :- )03:24
rm_youThere will be presentations at OpenStack Summit in Vancouver next month03:24
rm_youyeah, it's worth it :)03:25
nuxxerI see.  Scalable .. in a sense that haproxy can scale ?03:26
rm_youIt uses VMs/Containers through Nova on the backend03:27
rm_youso it can scale as much as your Compute Cloud can scale03:27
rm_youeach LB ends up on its own VM/Container03:27
nuxxerThe limitation that I see with vanilla LBaaS is ... lbaas agent is stuck with haproxy processes in a same box. (whether it's VM or Bare Metal).03:28
rm_youso noisy-neighbor effect, fully scaling03:28
rm_youthis is not03:28
rm_youit should be the official default implementation for neutron-lbaas, by the end of Liberty cycle03:28
nuxxerok.. so one VM per one pool ? and that one VM runs one instance of ha proxy process ?03:29
rm_youone VM per LB03:29
rm_youLB is the new root object, not Pool03:29
rm_youand one VM runs one haproxy process per Listener03:29
nuxxerIs HAproxy HA'ed ?03:29
rm_youone LB can have multiple listeners03:29
rm_youand yes, it is HA03:29
rm_youeither active-passive or active-active03:30
rm_youdepending on deployment03:30
nuxxerbut you said one VM.03:30
rm_yousorry, VM-set03:30
rm_youwhatever it is03:30
rm_youI just think of it as one VM, with clones :P03:30
nuxxerI see. .. to make HA, probably two VMs required .. I guess.03:31
nuxxercorrect ?03:31
nuxxerreading.. what you mentioned above.. this is interesting.03:31
rm_youstick around and ask tomorrow during work hours03:31
rm_youmost of us work in US timezones03:32
nuxxerI am in Austin.. just a night OWL. :- )03:33
rm_youyeah, lots of us in SA at RAX03:33
nuxxerthis is cool.03:39
nuxxerwhat about lbaas agent?03:39
nuxxerany plan to make it HA'ed ?03:39
nuxxersay.. there are 1000 pools... having just one lbaas agent makes me nervous.03:39
nuxxerso far.. what I find is .. each LBaaS agent knows its own pools.03:39
rm_youdon't focus on pools03:39
rm_youpools are just a logical object now03:39
rm_youthat's lbaasv103:40
rm_youwe're on lbaasv203:40
rm_youLoadBalancer is the root object03:40
rm_youa LoadBalancer can have 1-n Listener objects03:40
rm_you(Neutron-LBaaS v2)03:41
rm_youOctavia is using the v2 model03:41
nuxxerI started playing with lbaas 2 days ago. so. please understand if I ask dumb questions. :- )03:41
rm_youit's just a lot of people are stuck in v1 terminology03:41
rm_youwhich uses Pools as the root03:41
rm_youwhich makes zero sense03:41
nuxxerlistener == VIP ?03:41
rm_youit is kinda what the v1 VIP object was03:42
rm_youbut the v1 VIP object was badly named03:42
nuxxer(sort of ?_03:43
nuxxeroh... having VM... you already decoupled haproxy with lbaas agent.03:44
nuxxerstill ... only one lbaas agent knows about the LB VM .. right ?03:44
rm_youthe controllers are still "in communication with" the VMs03:44
rm_youthe controller layer will be HA/scaled in v1.003:45
rm_youv0.5 is the current goal tho03:45
rm_youthat we are close to03:45
nuxxerso. in current version of octavia .. If I run multiple lbaas agent.. pools won't be distributed evenly on them.  correct ?03:54
nuxxerlbaas agent ------------ VM (that runs haproxy) --------------- backend servers (like web servers)03:54
nuxxerThe reason I am asking about multiple lbaas agent is ... again about scalability..03:54
rm_youi am going to start pretending i don't know what a pool is :P03:54
nuxxerhaha ..03:55
nuxxerthing is .. openstack document does not give me good definition of v2.03:55
rm_youyeah, by "LBaaS Agent" you mean our "Controller Layer", and right now it is single-host, but v1.0 of Octavia will have it HA/scaling03:55
rm_youv0.5 though it is just a single host03:55
rm_you might help some03:56
nuxxera few problems .. I can think of... (hope I am wrong).03:58
nuxxerif becomes a VM (where you run haproxy)... that VM needs security group based on the protocol being handled.03:58
nuxxerif VM being HA'ed, 2 VM per LB.03:58
nuxxerprobably VM needs several interfaces.. then ~ 2 ports per VM...03:58
nuxxerx2 if HA'ed03:59
nuxxer1000 LB , then 2000 VMs.03:59
*** rm_work is now known as rm_work|away04:00
rm_youyeah just about right04:00
rm_youit's worked out04:00
rm_youthe exact network architecture depends on the deployer04:01
rm_youbut there are some examples04:01
nuxxerI am glad to find this proj.04:01
nuxxerI really want to know the details about Octavia . will read the terms.04:01
rm_youbest bet is to ask tomorrow04:01
nuxxernext time... hopefully I stop talking about pools . LOL04:02
rm_youthere are a lot of us here online04:02
rm_youduring the day04:02
nuxxerappreciate it .  I will let you rest. !04:02
nuxxerglad to meet you . "- )04:02
rm_youyeah, hopefully it works for you04:03
rm_youany chance you will be at the Vancouver summit?04:04
rm_youor are you getting into Openstack too late for that to be workable?04:04
*** kiran-r has joined #openstack-lbaas04:13
*** rm_work|away is now known as rm_work04:13
nuxxer:- ) I was unix kernel developer for a while and just started to use openstack about 4 month ago.04:17
nuxxerand 2 + days for lbaas04:17
nuxxerit's been fun.04:17
rm_workwell, this is the channel for it04:18
rm_workI think most of us at RAX will be available from about 1pm+ CST04:18
rm_workbut there may be people from HP/A10/other online before that04:19
nuxxerI see.04:20
*** clev-away is now known as clev04:20
nuxxerafter playing with lbaas , my first question was .. why only one lbaas agent ...04:20
*** kiran-r has quit IRC04:22
nuxxerlbaas agent should be scaled and HA'ed. hopefully soon in octavia . :- )04:24
nuxxeractive/acitve  way of HA. that would be awesome.04:24
nuxxerso that n mumber of lbaas agent can serve multiple lbaas calls from multiple tenants .04:24
*** clev is now known as clev-away04:42
openstackgerritPhillip Toohill proposed stackforge/octavia: Removes include_members from jinja_cfg
*** ajmiller__ has joined #openstack-lbaas05:09
*** madhu_ak has quit IRC05:10
*** ajmiller_ has quit IRC05:12
rm_worknuxxer: that's the plan, yes05:22
rm_workI am hoping our default deployment will be 2-3 haproxy nodes per LB, but it could scale up significantly05:23
*** kiran-r has joined #openstack-lbaas05:35
*** Aish has quit IRC05:43
*** Aish has joined #openstack-lbaas05:43
*** pcaruana has joined #openstack-lbaas06:09
*** nuxxer has quit IRC06:20
*** evgenyf has joined #openstack-lbaas06:46
*** kobis has joined #openstack-lbaas06:56
*** Aish has quit IRC08:03
*** woodster_ has quit IRC08:40
*** smcgough has left #openstack-lbaas09:06
*** Aish has joined #openstack-lbaas09:09
*** apuimedo has quit IRC09:09
*** xgerman has quit IRC09:38
*** apuimedo has joined #openstack-lbaas10:20
*** kiran-r has quit IRC11:53
*** woodster_ has joined #openstack-lbaas11:56
*** openstackgerrit has quit IRC12:07
*** openstackgerrit has joined #openstack-lbaas12:07
*** chlong has joined #openstack-lbaas12:44
*** clev-away is now known as clev13:08
*** fnaval has quit IRC14:05
*** apuimedo has quit IRC14:12
*** fnaval has joined #openstack-lbaas14:19
*** ajmiller__ has quit IRC14:25
*** kobis has quit IRC14:53
*** kobis has joined #openstack-lbaas14:54
*** kobis has quit IRC14:57
*** kobis has joined #openstack-lbaas14:57
*** kobis has quit IRC14:58
*** kobis has joined #openstack-lbaas14:58
*** jorgem has joined #openstack-lbaas15:20
*** jorgem has quit IRC15:27
*** jorgem has joined #openstack-lbaas15:30
*** evgenyf has quit IRC15:32
*** kiran-r has joined #openstack-lbaas15:39
*** Aish has quit IRC15:53
*** rm_work is now known as rm_work|away15:53
*** jorgem has quit IRC16:06
*** jorgem has joined #openstack-lbaas16:08
*** jorgem has quit IRC16:08
*** jorgem has joined #openstack-lbaas16:08
*** mlavalle has joined #openstack-lbaas16:12
*** nuxxer has joined #openstack-lbaas16:13
*** madhu_ak has joined #openstack-lbaas16:14
*** chlong has quit IRC16:19
*** pcaruana has quit IRC16:22
madhu_akblogan: Just wondering whether the patch is okay? because, I am thinking to write couple more tests based on testscenario that the work is not duplicated :)16:32
*** kobis has quit IRC16:34
*** Aish has joined #openstack-lbaas16:36
*** Aish has quit IRC16:45
*** rm_you has quit IRC16:52
*** bharath has joined #openstack-lbaas16:52
*** rm_you has joined #openstack-lbaas16:52
*** rm_you has quit IRC16:52
*** rm_you has joined #openstack-lbaas16:52
*** mwang2 has joined #openstack-lbaas16:53
nuxxerHi, Is there architecture diagram that show octavia topology ?  .... that shows end to end configuration ?17:00
*** BeardyMcBeards has left #openstack-lbaas17:01
*** bharath has quit IRC17:01
*** jorgem has quit IRC17:03
*** jorgem has joined #openstack-lbaas17:04
*** mwang2_ has joined #openstack-lbaas17:10
*** mwang2 has quit IRC17:11
*** Aish has joined #openstack-lbaas17:12
*** jorgem has quit IRC17:14
*** sbalukoff has quit IRC17:19
*** clev is now known as clev-away17:25
*** sbalukoff has joined #openstack-lbaas17:37
openstackgerritMichael Johnson proposed stackforge/octavia: Fixes a config load bug for keystone v3 domains
bloganmadhu_ak: was testscenario doing the same thing as ddt?17:40
madhu_akthe only difference is, we dont need to install thrid party package for ddt17:40
bloganddt is standard?17:41
blogani.e. is a built in python lib?17:41
*** hitalia has joined #openstack-lbaas17:42
bloganah interesting, wonder why the neutron folks use testscenarios over ddt then17:42
madhu_akI think I misunderstood your question, for ddt when running the tests, we have to include ddt in test-requirements.txt17:43
bloganohh, testscenarios you do not?17:43
*** jorgem has joined #openstack-lbaas17:43
madhu_akyep..its already mentioned in test-requirements.txt17:43
bloganokay in that case testscenarios is better :)17:45
madhu_aksince testscenario is included in test-requirements.txt, I am just importing testscenarios with its subclass TestWithScenarios to run the scenarios :)17:45
bloganthough it does look like some openstack projects do use ddt, while most use testscenarios (based off a quick google search)17:45
madhu_akto be honest, ddt is yet to be used in openstack projects17:46
bloganlooked like marconi was using it17:46
blogani guess its zaqar now17:46
madhu_akoh yeah17:47
madhu_akI didnt hear that name thoug ;)17:47
madhu_akyes, they are using ddt17:48
bloganmadhu_ak: reviewed, minor complaints :)17:50
madhu_akawesome, thanks for the comments :)17:50
bloganmadhu_ak: thanks for doing the testscenarios research17:51
madhu_akbtw when using scenario tests, not sure whether there is a way to skip17:51
madhu_akwill do research though17:51
madhu_akno problem17:51
blogantagging with @skip wouldn't work (or @unnittest.skip)17:52
madhu_aknot sure, will check though17:52
madhu_akOnce this is addressed and merged, I would like to rearrange the existing tests to reflect it?17:52
bloganmadhu_ak: yeah that'd be great17:53
madhu_akone more question, I am not sure of this error when running the test against my devstack:17:56
madhu_ak    tempest.exceptions.Conflict: An object with that identifier already exists17:56
madhu_ak    Details: An object with that identifier already exists17:56
madhu_ak    Details: {u'type': u'OverQuota', u'detail': u'', u'message': u"Quota exceeded for resources: ['network']"}17:56
madhu_akmath domain error17:56
madhu_akSlowest Tests17:56
*** SumitNaiksatam has joined #openstack-lbaas17:56
madhu_akblogan ^^ any idea about that?17:56
bloganhow many networks do you haev under that account?17:58
madhu_akwhen I do neutron net-list, I could see 1118:00
madhu_akincluding public and provate18:00
*** woodster_ has quit IRC18:00
blogani bet your quota is at 1018:01
bloganlook in /etc/neutron/neutron.conf at the [quotas] section18:02
*** bharath has joined #openstack-lbaas18:02
madhu_akyes it is 1018:03
madhu_akbut it is commented..18:03
madhu_akI think its default18:04
*** davidlenwell has quit IRC18:05
*** davidlenwell has joined #openstack-lbaas18:05
bloganyeah its the default18:05
*** bharath has quit IRC18:06
*** nuxxer has quit IRC18:06
madhu_akI see..18:06
madhu_akanyways, I need to clear up then..18:06
*** rm_work|away is now known as rm_work18:07
bloganyeah the only two options are that your tests stay under that limit by deleting networks it doesn't need (since its parallel im not sure thats always possible) or up the quota18:07
*** bharath has joined #openstack-lbaas18:17
*** clev-away is now known as clev18:21
*** jorgem1 has joined #openstack-lbaas18:27
*** jorgem has quit IRC18:29
bloganjohnsom: you changed the agenda page while i was editing!18:33
johnsomOpps, sorry.18:33
bloganjohnsom: ha no worries, i really just added on to what you had18:34
johnsomI asked German if he wanted me to update it as he is in meetings all day18:34
johnsomblogan Are you driving today?18:35
*** hitalia has quit IRC18:37
johnsomI guess I should clarify, are you driving the meeting vs. your truck18:39
*** bharath has quit IRC18:41
*** jorgem1 has quit IRC18:43
*** bharath has joined #openstack-lbaas18:44
*** hitalia has joined #openstack-lbaas18:57
*** jorgem has joined #openstack-lbaas19:04
bloganjohnsom: lol yeah ill drive it19:07
openstackgerritTrevor Vardeman proposed stackforge/octavia: Amphora SSH Driver
*** hitalia has quit IRC19:20
*** bharath has quit IRC19:21
*** bharath has joined #openstack-lbaas19:24
*** woodster_ has joined #openstack-lbaas19:31
*** bharath_ has joined #openstack-lbaas19:46
*** bharath has quit IRC19:47
openstackgerritMerged stackforge/octavia: Removes include_members from jinja_cfg
*** clev is now known as clev-away19:48
*** bharath has joined #openstack-lbaas19:50
*** bharath_ has quit IRC19:52
dougwigblogan: i'll have a few announcements at the meeting today.19:57
*** ajmiller has joined #openstack-lbaas19:58
blogandougwig: are they already on the agenda?19:58
dougwigjust added19:58
*** ajmiller_ has joined #openstack-lbaas19:59
*** jorgem has quit IRC19:59
*** nuxxer has joined #openstack-lbaas20:09
openstackgerritmin wang proposed openstack/neutron-lbaas: Admin_state_up tempest test for health monitor
*** hitalia has joined #openstack-lbaas20:18
*** kiran-r has quit IRC20:19
*** hitalia has quit IRC20:23
*** jorgem has joined #openstack-lbaas20:33
*** ajmiller_ has quit IRC20:33
*** ajmiller has quit IRC20:34
johnsomblogan When do we want to merge?20:43
ptoohillI suppose I should take a look but, why are others not needed that Pecan req blogan? Do you know what we are doing differently?20:45
*** hitalia has joined #openstack-lbaas20:46
bloganjohnsom: i'd like to get it merged after teh ssh driver, so i'm actaully ready to merge the ssh driver20:46
bloganjohnsom: and then the controller worker would have some merge conflicts to fix probably20:46
bloganjohnsom: and then we cna get that merged when you are comfortable enough with it being merged and allowing others to post reviews to fix bugs20:46
johnsomCan we merge the keystone one too?20:47
johnsomblogan: I have +2 ssh20:47
bloganjohnsom: yep20:47
johnsomSo, land controller worker in maybe 30?  Give time for ssh to get through and me to rebase, etc.20:48
bloganjohnsom: sounds good20:49
johnsomblogan: cool20:49
openstackgerritMerged stackforge/octavia: Fixes a config load bug for keystone v3 domains
johnsomDooh, keystone conflict with ssh20:54
*** hitalia has quit IRC20:59
ptoohillI think the latest Pecan pulls in singledispatch, blogan. Could you verify when you get a chance?21:00
bloganptoohill: sure21:00
ptoohill0.8.3 is latest for me21:00
*** hitalia has joined #openstack-lbaas21:03
*** clev-away is now known as clev21:05
ptoohillhmm, it seems to be loading singledispatch even at 0.8.0 for me21:06
bloganwell i dont know what happened with my testing of it21:06
bloganare you doing it from a fresh install?21:06
bloganand do python install21:06
ptoohillI purge everything pecan/singledispatch and attempt to install by hand21:06
ptoohilli did not do it that way21:06
bloganmight be an issue with pbr then21:06
ptoohillthe global-reqs version is 0.8.0 for pecan, i would have assumed others would need this also21:07
bloganyeah me too21:07
johnsomTrevorV blogan Are you guys rebasing the ssh patch?21:07
bloganTrevorV is giong to21:07
johnsomOk, cool21:08
bloganah screw it ill do it21:08
johnsomI am ready to go for the controller21:08 install seems to work for me, blogan. So maybe this isnt a needed req?21:09
ptoohillI purged packages, removed req from requirements.txt and ran it to be clear21:09
bloganptoohill: let me run it on a fresh machine and see first21:10
bloganand then we can just remove them21:10
ptoohillgood deal, thank you21:10
ptoohillwhere is 'ipaddr' being used?21:10
ptoohilleh, ill dig21:10
blogani thought it was pecan as well21:10
bloganoh ptoohill try to run octavia after setup.py21:11
ptoohilli see the singledispatch req get install during setup21:11
ptoohillyou mean run api?21:11
bloganshould jsut be octavia-api from cli21:11
*** nuxxer is now known as nuxxer-away21:12
ptoohillno module named pecan, im clearly doing something wrong21:12
openstackgerritBrandon Logan proposed stackforge/octavia: Amphora SSH Driver
bloganthat was an odd conflict, it actaully removed a close ]21:13
ptoohilloh, nvm derp. I purged after i installed last time. It runs for me21:13
bloganokay then21:13
openstackgerritTrevor Vardeman proposed stackforge/octavia: Update PUT methods to send update through queue
ptoohillpecan doesnt install ipaddr though21:18
ptoohillAlso, on first pass im not seeing it being used etiher21:21
ptoohilli take that back21:22
johnsomblogan: ssh through the gate21:22
openstackgerritBrandon Logan proposed stackforge/octavia: Amphora SSH Driver
bloganjohnsom: missed a newline21:23
bloganptoohill: ImportError: No module named simplegeneric21:24
bloganName: pecan21:24
bloganVersion: 0.8.321:24
ptoohill:/ its not breaking for me and what I would believe the rest of the OS community otherwise that would be in global reqs21:25
ptoohilland were not using the ipaddr directly, I see it used by other libs like requests21:26
*** nuxxer-away is now known as nuxxer21:28
bloganim fine with removing it21:28
bloganthough i don't think it will work on a fresh install21:28
blogani think its something else21:28
bloganwhat is that freeze from?21:29
ptoohillmy box21:29
bloganso you already have simplegeneric21:29
bloganuninstall that and reinstall pecan21:29
ptoohillah, i thought it was on dispatch21:30
bloganit was, this is different21:30
blogannow pip show pecan has singledispatch as a requirement21:30
ptoohillit tells me its not installed21:30
ptoohilloh, cause i uninstalled dispatch. one sec21:30
bloganwell pecan should install21:30
openstackgerritMerged stackforge/octavia: Amphora SSH Driver
bloganuninstall pecan and simplegeneric21:31
bloganthen try to install pecan21:31
bloganjohnsom: ^^ megeed21:31
johnsom(if that is a word...)21:31
blogani do believe ti is in our world21:31
ptoohillremoved those three, it seems to have just installed dispatch21:31
ptoohillchecking freeze21:31
bloganjust run octavia-api21:32
ptoohillah, theres the ipaddr req21:32
ptoohillok, so i do get the simplegeneric missing21:34
openstackgerritMichael Johnson proposed stackforge/octavia: Implements Octavia Controller Worker
ptoohillSo, if anything we need that and ipaddr. Though, we have to be doing something different then others. These should be reqs already I would have thought21:35
ptoohillno, WSME installs ipaddr and simplegeneric21:37
ptoohillIll spin up a box and test this but I believe we dont need these extra reqs and can remove them.21:38
johnsomController worker passed Jenkins.  Do we want to land consumer worker now as well?21:39
*** jorgem has quit IRC21:42
*** SumitNaiksatam has quit IRC21:43
*** mwang2_ has quit IRC22:00
*** bharath has quit IRC22:06
bloganjohnsom: yes!22:06
*** jorgem has joined #openstack-lbaas22:06
*** jorgem has quit IRC22:17
*** bharath has joined #openstack-lbaas22:18
*** mwang2 has joined #openstack-lbaas22:30
*** jorgem has joined #openstack-lbaas22:31
openstackgerritTrevor Vardeman proposed stackforge/octavia: Update PUT methods to send update through queue
*** bharath has quit IRC22:48
*** mwang2 has quit IRC22:56
*** rm_work is now known as rm_work|away22:58
*** jorgem has quit IRC23:00
openstackgerritMerged openstack/neutron-lbaas: Updated from global requirements
*** nuxxer has quit IRC23:19
*** madhu_ak has quit IRC23:35
openstackgerritPhillip Toohill proposed stackforge/octavia: Remove extra requirements
openstackgerritPhillip Toohill proposed stackforge/octavia: Update nova_driver to use amp_network
*** chlong has joined #openstack-lbaas23:50

Generated by 2.14.0 by Marius Gedminas - find it at!