/srv/irclogs.ubuntu.com/2013/08/29/#ubuntu-uds-client-1.txt

=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Currently no events are active in this room - http://summit.ubuntu.com/uds-1308/client-1/ - http://irclogs.ubuntu.com/2013/08/29/%23ubuntu-uds-client-1.html
=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Track: Foundations | Disccuss Critical Pieces of CI Infrastructure | Url: http://summit.ubuntu.com/uds-1308/meeting/21943/disccuss-critical-pieces-of-ci-infrastructure/
=== Ursinha is now known as Ursinha-afk
=== Ursinha-afk is now known as Ursinha
sil2100Hangout link for the current session: https://plus.google.com/hangouts/_/617b2d450bf059edef8675e873a4c84aead18077?authuser=0&hl=en13:57
sil2100asac: ^13:59
* rsalveti waves14:02
* sergiusens waves back14:02
zygahi14:02
looldoanac`: you're ok in terms of audio; video is a bit choppy but it's ok14:02
zygadoanac`: checkbox is used for sru testing and other kinds of "testing"14:04
loolwhat about autopkgtests?14:05
cjwatsonlool: it's mentioned there14:05
loolah sorry need to check the pad14:05
rfowlercan't see etherpad14:07
Ursinharfowler, try to "separate notes window" and it should ask to login (happened here a couple of times)14:08
loolrfowler: http://pad.ubuntu.com/uds-1308-disccuss-critical-pieces-of-ci-infrastructure14:09
rfowlerUrsinha: that worked14:09
Ursinhagood :)14:09
elmocanonistack is a known oversubscribed resource14:10
elmoyou shouldn't be using it for production services14:10
elmothat's what prodstack is for14:10
zygadoanac`: then use aws instead of canonistack14:11
plarsnot *too* much, but sometimes it's helpful to get at the config files directly14:15
plarswe do manually touch a lot of the nodes, especially those that are connected to devices14:16
plarsI'm probably behind where the conversation is at though...14:16
rfowlermost have remote management14:17
zygadoanac`, plars, asac: maybe we should run lava to control the phones?14:17
loolasac: right14:18
plarsthe public/private split is due to firewall restrictions14:18
zygaasac: we also have that issue (public/private) jenkins14:18
loolasac: I think the QA Lab was built separated from everything else14:18
loolasac: while public one was built alone just for this14:18
loolperhaps worth discussing stability of x86 machines used for testing graphics (e.g. Mir/Unity)?14:19
loolit's on the path to land Mir or some stacks, even if that's not ARM14:20
cjwatsonelmo: Yeah, I can understand the rationale.  I mostly just don't want to silently not mention that it's a practical problem for us, in this kind of session, although it is something we are living with14:21
sergiusensrfowler: do people in the EMEA or APAC timezone have access to remote management?14:23
Mirvsergiusens: yes, the problems we've had have been that remote power switching has not been enough14:23
rfowlersergiusens: i don't see why not14:23
Mirvor some of them, obviously many have been resolved with power switching remotely14:23
sil2100sergiusens: sometimes we could resolve it remotely, but there were moments when this was not enough14:23
sergiusensrfowler: yeah, we just need to have a goto list somewhere14:23
sergiusenssil2100: I know14:24
Mirvsergiusens: there's QA wiki page with quite a lot of details14:24
rfowlerhttps://wiki.canonical.com/UbuntuEngineering/QA/Lab14:24
rfowlersergiusens:14:24
looldoanac`, jibel: (ack)14:24
rfowlersergiusens: i update it constantly14:24
rfowlerthere are only two machines like that jibel...14:25
zygawe have the same class of issues in certification test pool14:25
rfowlerwith exception to the phones14:25
sergiusensrfowler: oh, I asked ones for something so I don't restart the wrong thing regarding that :-)14:25
zygawe're automatically booting about 50 machines for remote testing14:25
zygasome have issue with boot-from-lan or need manual interventions often14:25
zygabut we have a large lab / office where people have access to all hardware so it's easy to get around14:26
rfowleri can spin up one :)14:26
zygawe also look at moving to something like maas or maas derived to manage our pool14:26
plarsright14:26
zygawith extra features for less-than-ideal-hardware14:26
zygalike manual action console for people on site14:26
zygaor support for odd boot methods and one-off solutions14:26
zygaintegrated power management not based on IPMI14:27
zygaand stuff like that14:27
rfowlerasac: we are limited by space in this current lab for that14:27
zygaperhaps the same stuff can be reused to run other lbas14:27
zygalabs14:27
chilukor you could get a good remote control adapter14:27
zyga(and we're also going to use tablets/phones there)14:27
chilukor machines that come with a good one.14:27
chiluklike hp ilo etc.14:27
Mirvrfowler: btw that page being up-to-date has been very appreciated when we've been remote debugging14:27
plarsyeah, I'm lost... I think the conversation took a very different turn14:27
chilukin this day and age there is no excuse to need someone to go push buttons on a server14:28
loolev: it's not jenkins, it's the physical slaves14:28
loolev: like e.g. some machine with ATI graphics14:28
evah, right14:28
rfowlerMirv: np14:28
zygawe also have two physical labs so we have all-globe coverage pretty much14:28
plarsI think it's more like needing to add lots more devices and move towards scheduling them more dynamically14:28
rfowlerMirv: keeps me sane too when I bring up all these boxes14:28
loolor nexus 4 / nexus 714:28
cjwatsonchiluk: A lot of the stuff here is inherently not server-class, surely (e.g. mobile)14:29
* zyga feels totally ignored here ;)14:29
sergiusenszyga: i'm reading you14:29
chilukcjwatson, I understand mobile...14:29
sergiusensbut the conversation seems hard to interject14:29
loolzyga: (I'm reading you but mostly throwing questions at the conversation from the hangout :-)14:29
zygathanks!14:30
zygaI think we have the _exact_ same problems but you have a bit larger scale, I'm just interested in getting a good solution for both teams14:30
rfowlerthere are already 25+ phones/devices in the lab and I'm generally on top of keeping them up14:30
sergiusenszyga: join the hangout!14:30
loolzyga: https://plus.google.com/hangouts/_/617b2d450bf059edef8675e873a4c84aead18077?authuser=0&hl=en14:30
zygasergiusens: it's full and I don't want to interrupt you that much but ok14:31
loolzyga: isn't it up to 15?14:31
chilukdoanac++14:31
zygait's full14:31
plarsiirc there were difficulties making our test VMs and physical machines provision with maas because it lent itself more to knowing ahead of time what you want, where we want the crack of the day14:31
UrsinhaI thought there was a 10 people limit14:31
plarsretoaded: you looked into that a lot I think?14:31
sergiusenszyga: I'm dropping so you can join14:31
plarsbut I still think we're having lots of different conversations in parallel here14:31
plarsah14:32
zygathanks14:32
plarsthat pc is not a huge thing to replace14:32
chilukcan't we hack phones with arduinos to push buttons?14:32
sergiusenszyga: get in!14:32
chiluki know the unity/mir team has something like that for their display hotplug testing14:32
plarsthere's not much to charm for nodes like phoenix14:32
rfowlerchiluk: I'm working on that14:32
retoadedplars, we're still stuck with what maas can provide on precise until the lab gets update to a more recent version anyway14:32
chilukrfowler nice14:32
plarsthe only "special" piece it needs is the jenkins slave, and a bit of udev rules that we scp from just about anywhere else14:32
plarsthat's it14:32
plarswhere would we put our VMs if we did that? prodstack?14:36
plarsdoes it support us doing something like this and can it handle us spinning up lots of VMs in this way?14:37
* cjwatson drops off so there's more space for people contributing more14:37
* ogra_ agrees with cjwatson and leaves too 14:38
zygaasac: how do you plan on using charms on the various test machines?14:44
zygaasac: using openstack on virtual machines and use a juju provider for that14:44
zygaasac: what about physical machines? maas?14:44
jamespagehere to answer questions re jenkins/java in the archive if need be...14:45
jamespageelmo, either14:45
jamespageelmo, fastpath = image based14:45
jamespageelmo, or d-i based installed14:46
nuclearbobwe can still do that with pxe without cobbler14:47
zygawe don't use cobbler14:47
zygaand we also use maas to test servers and cloud setups lately14:47
sergiusensrfowler: asac: doanac`: can we electrically open all the devices for power cycling initially?14:49
sergiusensdevices == phones, tablets14:49
doanac`sergiusens: currently they just attached via USB to the host14:49
doanac`rfowler was looking into adding the ability to "push the power button". but its pretty hard and requires soldering14:50
zygadoanac`: lava lmp boards :)14:50
sergiusensasac: there's the RT bucket14:51
sergiusensretoaded: ^^14:51
sergiusensfginther: ^^14:51
doanac`zyga: thought about it, but I'm not sure if that works well with consumer products like an N414:51
retoadedsergiusens, that works for the instances that lead to an RT being filed. Many we handled via E-mail or IRC14:52
sergiusensretoaded: yeah, that sort of sucks i retrospect...14:53
retoaded:-)14:53
sergiusens*in14:53
elmojamespage: is the image specific to MAAS or was I making that up?14:54
jamespageelmo, for fast path - yes14:54
jamespageelmo, for d-i its just the stock mini.iso installer stuff thats used14:54
rfowlerthere needs to be more discussions like this14:54
sil2100Definitely, this session is too short for this14:55
elmojamespage: cool, ta14:56
elopiothank you.14:57
sil2100Thanks guys14:57
asacsergiusens: the problem with powercycling is that we also need to push the button14:57
asacto start the phone14:57
asacunless we keep the boot partition free i think14:58
asac(on galaxy nexus)14:58
sergiusensasac: I tried the free boot partition thing, it won't pan out with image based upgrade image15:00
=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Track: Client | Mir performance improvements | Url: http://summit.ubuntu.com/uds-1308/meeting/21940/client-1308-mir-performance/
asacsergiusens: right. thats the problem15:00
ogra_relais ...15:00
sil2100https://plus.google.com/hangouts/_/34e6e37dab70881ef67b20d0acc417a82fed18e0?authuser=0&hl=en <- hangout link for Mir preformance improvements15:01
ogra_usb driver ... to glue onto the case15:01
ogra_*driven15:01
sil2100(brb in front of the cam)15:01
sil2100Let's spin this session15:06
sil2100https://plus.google.com/hangouts/_/34e6e37dab70881ef67b20d0acc417a82fed18e0?authuser=0&hl=en <- in case anyone wants to join in, hangout link15:07
ollithe mentioned URL to compare Nexus and Nexus 4 - http://gfxbench.com/compare.jsp?D1=Google+Galaxy+Nexus+%28Samsung+I9250%29&D2=Google+Nexus+4&cols=215:13
john-mcaleelydo you have a feel for how to separate Mir performance from Unity8's? ie, how does UNity8 work on destop GL + Xmir vs UNity 8 on GLES/Touch?15:14
john-mcaleelyYes - I 'd like to be able to separate Unity 8 and Mir performance.15:20
john-mcaleelyagreed on screen size changes!15:21
john-mcaleely(I made up the config to try to isolate Mir. It may make no sense - so thanks olli)15:21
john-mcaleelyaha - so on desktop GL drivers, etc. Nice15:22
john-mcaleely(sorry, I didn't join the hangout - multitasking at my end in another hangout)15:23
john-mcaleelywhere's the link?15:28
john-mcaleely:-)15:28
olliin launchpad ;)15:28
john-mcaleelyha!15:28
ollikg is pulling it up15:29
kgunnhttps://blueprints.launchpad.net/ubuntu/+spec/client-1303-mir-converged15:29
john-mcaleelythx15:29
john-mcaleelythanks kgunn olli sil2100 !15:31
kgunnjohn-mcaleely: your welcome...thanks for the questions15:31
sil2100Thanks everyone!15:32
sil2100Sorry for not being too active, so much things to do!15:32
=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Currently no events are active in this room - http://summit.ubuntu.com/uds-1308/client-1/ - http://irclogs.ubuntu.com/2013/08/29/%23ubuntu-uds-client-1.html
=== alecu is now known as alecu_lunch
=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Track: Client | Mir roadmap | Url: http://summit.ubuntu.com/uds-1308/meeting/21938/client-1308-mir-roadmap/
sil2100https://plus.google.com/hangouts/_/638e314fae56e6bde09d491276807326e3b48472?authuser=0&hl=en <- hangout for people who want to participate18:01
=== alecu_lunch is now known as alecu
olliis there any input re Mir roadmap from the audience18:44
chilukis there any plan to backport 14.10 features into 14.04?18:45
chilukin order to decrease support burden on xmir?18:45
chilukyou probably heard it from me.. I don't think it's really been answered18:47
TheMusoIs there going to be any form of multi-display output on our supported touch devices between now and 14.04?18:47
TheMusoi.e I'll be using a Nexus 10 for a11y development, and I'm  hoping to use the HDMI out on that with a monitor for easier development and workflow.18:47
chilukthanks olli..18:48
argesany talk about wireless display technologies? such as intel widi? It would be pretty cool to be able to beam my phone's screen to a large monitor.18:48
seepa+1 arges18:49
argesobviously intel widi might not be a phone, but something similar.18:49
chilukyeah that would be nice ..18:49
argesunderstood. cool18:50
sil2100Any other remarks?18:50
chilukwe are already partnered with intel18:50
chilukright?18:50
kgunnsee miracast18:51
sil2100Thanks everyone!18:53
sil2100olli: thankfully it wasn't a door or a door-frame ;) Bad luck, just had to damage myself right before UDS during the wekend18:54
=== udsbotu changed the topic of #ubuntu-uds-client-1 to: Currently no events are active in this room - http://summit.ubuntu.com/uds-1308/client-1/ - http://irclogs.ubuntu.com/2013/08/29/%23ubuntu-uds-client-1.html
=== marrusl is now known as marrusl_afk
=== marrusl_afk is now known as marrusl

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!