Saturday, 2015-06-27

*** dimsum__ has quit IRC00:11
*** dimsum__ has joined #kolla00:12
*** alisonh has left #kolla00:13
*** sdake has joined #kolla00:16
*** alisonh has joined #kolla00:19
sdakerhalliey you pinged me earlier00:37
rhalliseyoh I said cinder is having some issue00:37
sdakerhallisey you pinged me earlier00:37
sdakeya looking at outstanding bugs we could probably release with the dhcp bug fixed00:37
rhalliseyya00:37
rhalliseyfixed a ton of bugs today but stilll more to go00:38
sdakewell we dont haeto have a bugfree milestone reelase00:38
sdakebut it should work\00:38
*** vinkman has quit IRC00:39
*** dimsum__ has quit IRC00:39
rhalliseywell I was hoping for full cinder00:39
rhalliseyI'm going to try and revert to and old patch because I'm seeing an old error00:39
sdakei'm going back to bed00:39
rhalliseynight man00:40
sdakeapologies for being mia00:40
rhalliseyno problem00:40
sdakebut root canal pretty serious surgery00:40
rhalliseyit was handled00:40
rhalliseyya sorry to hear00:40
sdakeshit happens00:40
sdakeatlesat it was done well00:40
*** sdake has quit IRC00:40
*** dimsum__ has joined #kolla01:23
*** britthou_ has quit IRC01:36
*** britthouser has joined #kolla01:37
*** rhallisey has quit IRC01:39
*** erkules_ has joined #kolla01:57
*** erkules has quit IRC01:59
*** yapeng has joined #kolla02:26
*** yapeng has quit IRC02:31
*** sdake has joined #kolla02:32
*** dimsum__ has quit IRC02:32
*** britthou_ has joined #kolla02:36
*** dimsum__ has joined #kolla02:38
*** daneyon_ has quit IRC02:39
*** britthouser has quit IRC02:40
*** daneyon has joined #kolla02:40
*** daneyon has quit IRC02:44
*** dimsum__ has quit IRC02:46
sdakerhallisey re the rest of kolla, are services generally working well now?02:54
*** sdake_ has joined #kolla03:02
*** sdake has quit IRC03:06
*** absubram has joined #kolla03:27
*** yapeng has joined #kolla03:32
*** yapeng has quit IRC03:37
*** sdake_ has quit IRC03:41
*** sdake has joined #kolla04:05
*** sdake has quit IRC05:06
*** sdake has joined #kolla05:10
*** sdake has quit IRC06:08
*** nihilifer has joined #kolla06:39
openstackgerritMichal Rostecki proposed stackforge/kolla: Add base image for Ubuntu  https://review.openstack.org/19617206:46
*** nihilifer has quit IRC07:13
*** nihilifer has joined #kolla07:25
*** shardy has joined #kolla07:29
SamYaplemorning07:51
*** akwasnie has joined #kolla07:55
*** yapeng has joined #kolla08:22
*** yapeng has quit IRC08:27
SamYapleGODDAMMIT08:50
SamYaplefucking docker08:51
*** nihilifer has quit IRC08:57
*** akwasnie has quit IRC10:01
*** sdake has joined #kolla10:10
*** yapeng has joined #kolla10:11
sdakesamyaple about10:11
SamYapleyup10:13
SamYapleyoure up early/late10:13
SamYaplesdake: did you see the bug comment?10:13
*** sdake_ has joined #kolla10:13
*** yapeng has quit IRC10:16
*** sdake has quit IRC10:17
SamYaplesdake_: were you sleep typing again?10:19
sdake_nah i slept all day friday10:19
sdake_so awake now10:19
SamYapledid you see the bug comment/10:20
sdake_from 10am until about 10pm or so10:20
sdake_didn'tsee10:20
sdake_let me look10:20
sdake_thans for commenting - that is what i  wanted to talk about10:20
SamYaplebadnewsbears.jpg10:20
*** nihilifer has joined #kolla10:20
sdake_can you implement option #1, with wiating for docker 1.8 to fully thinify the containers10:23
sdake_why wa it reverted btw10:23
sdake_you wlll go to the degoba system10:24
SamYaplecommit message says it "breaks live migration", but it looks like it only breaks it when dealing with bind mounted volumes and live migration10:24
sdake_live migration of qemu?10:24
SamYapleno of docker containers10:25
SamYapleflocker10:25
sdake_do they have a plan to fix it10:25
sdake_or did they give up on the concept10:25
SamYapleconsidering there is no bug reports around, and this popped in and out between a release, the only way to "fix" this is from outside pressure10:26
SamYaplethe easiest way to get that is to have this be a daemon runtime option in a stable release10:26
sdake_i'll ping lk4d4 on monday about it10:26
SamYaplewe could also do a custom/signed docker build as another option10:27
SamYaplenot my favorite idea10:27
sdake_i dont want to be in the business of building doker10:27
SamYapleyea10:27
sdake_so re optoin #110:28
sdake_what kind of timeframe we taking about to get it implemented10:28
SamYaplei can redo it tonight, just get the plugin agent seperated and it wont be so bad10:29
SamYaplejust no namepsaces on the host :/10:29
SamYapleno*10:29
sdake_well lets assume that we will get that mount propogation feature back in at some point in the future10:29
sdake_so the namespace on the host prolem wont be a problem long term10:29
sdake_was that coherent :)10:30
SamYaplewe _might_ be able to get it back in Docker1.7.1 if they go for a quick daemon argument10:30
* SamYaple sighs10:31
SamYapleive been talking about this mount progogation thing :(10:31
sdake_what do you mean10:32
SamYapleto coworkers10:32
sdake_sorry just got up brain not booted10:32
SamYaplebecause this is a pain point for OSAD10:32
SamYaple"namespaces on host"10:32
sdake_you mean debugging purposes10:33
SamYapledebugging, troubleshooting, quality assurance10:33
SamYapleeverything really10:33
SamYaplehaving to enter into a container to enter the namespaces is a huge PITA for day to day life10:33
SamYaplewhen you talk to lk4d4 reference the commit i posted. I would be ok with a deamon option, environment variable, anything to make privatefs be configurable10:34
sdake_so looking at hat patch10:36
sdake_privatefs = true was it false previously?10:37
sdake_can ou explain what it does so i can talk with lk4d4 aout it intelligently10:40
sdake_or another option is we can grill him together ;)10:40
sdake_this woudl be my preferred approach but i'm not sure if your tz's match up10:40
sdake_because its an exxecdriver, I think a config option may be a nonstarter (cmd line option)10:43
sdake_envronment variable may be more acceptable10:44
sdake_so if we do option #1, is ovs agent going to work properly10:47
sdake_or is work needed there to fat-ify it a bit?10:47
sdake_samyaple awake bake up ;)10:47
sdake_back up that is10:47
SamYaplesdake_: sorry baby feeding10:58
sdake_ack10:58
SamYaplesometime after 1.6.0 the privatefs = true we removed so the container was created without private namepsaces10:59
SamYaplehttps://github.com/docker/docker/commit/ff770d33cdee5ab4dc0bdeb06dcfea49aa72ab8e10:59
SamYaplethat chagne reverts it back to the 1.6 behaviour10:59
SamYapleit needs to be toggleable/configurable10:59
sdake_why is it a config optioin internally if not confiurable?11:00
SamYapleat least on a per volume level ( docker run -v /run/netns:/run/netns:private or something like that)11:00
SamYaplecontainer is an object for namespaces and what not, not a docker construct11:00
sdake_so we want the volume created with a shared namespace then?11:01
SamYapleanyway. short and simple is if you want mountpoints to propogate down, the namespaces cant be private11:01
SamYaple"shared" is the wrong term11:01
SamYaplenot-private is correct11:02
sdake_it appears the docker cats think private should be the defautl11:02
sdake_so if we have a -v /ru/netns:/run/netns:not-private ?11:02
SamYaplei dont really care what the default is as long as we have a method to do it the way we want11:02
sdake_well what i'm aing is the default is probably going to be whatever it is in 1.7.011:03
sdake_so what shoud the volume flag be called?11:03
SamYaplethe initial commit that turned _off_ private namespaces planned to implement them as a flag on the volumes11:03
sdake_itw ould be hepful if I knew go so I could implement the patch ;)11:03
SamYaplethere is no reason private needs to be the default as was argued before the initial commit11:04
SamYaplethe _only_ reason this got turned back on was because it probably broke flocker or some other big names in docker before the 1.7 release11:04
sdake_well the arguemnt will be "it fucks up flocker" which is a docker thing and therefore there interest will be that flcoer is not busted11:04
sdake_meanning private will be the new default11:05
SamYaplei dont disagree. i also dont care what the defaults are, but this should be configurable11:05
sdake_got that part, where did the original commit come from where the person said "we are making it confiurable" and then didnt finish the job11:05
SamYapleit was a "we should do that if the need arises" comment in the initial commit thread for turning off private namepsace11:06
SamYaplecant find it at the moment11:06
SamYaplethey refactored the files making history go *poof*11:06
sdake_if you can find it it would be helpful with communication wiht lk4d411:06
sdake_i see11:06
sdake_so basically they added a bunch of code to do the prviate/nonprivate implementation11:07
sdake_and the made the default something that busted flocker11:07
sdake_so that person should fix it ideally :)11:07
SamYapleno. that commit is literally the only thing that changed11:07
sdake_something added the config option right?11:08
SamYaplethere was never a config option11:08
SamYaple11:06 < SamYaple> it was a "we should do that if the need arises" comment in the initial commit thread for turning off private namepsace11:08
sdake_in the object11:08
SamYapleno in the PR11:08
sdake_where is that thread11:08
SamYapledude i just said i dont know. im looking. they refactored so history is gone11:08
sdake_i see11:09
sdake_ok :)11:09
sdake_4am, slack plz :)11:09
SamYaplehttps://github.com/docker/libcontainer/pull/52011:09
SamYaplehttps://github.com/docker/docker/pull/1385411:10
sdake_cool dan walsh11:10
SamYaplelooks like the daemon flag is already being discussed in that latest pull thread11:10
sdake_maybe he can fix it11:10
SamYapleyea this was a quick reaction to fix a regression before release of 1.711:12
SamYaplethis will be re-reverted in 1.8 in some form probably, if we keep pressure on it11:12
sdake_it would help if someone had a pull request ready to go11:14
sdake_I'll talk to dwalsh monday11:14
sdake_maybe he cando that work11:14
sdake_or find soemone at rht to do it11:14
sdake_I'd do it but I don't know go11:14
SamYaplei dont even understand what the CRIU issue is, but it appears to be a kernel bug. thats a bit beyond my level11:14
SamYaplemakign it configurable, that I can do11:14
sdake_configurable via -v option?11:22
sdake_that seems like the best option to make all parties happy11:22
sdake_i'll start off with a private email11:25
sdake_dwalsh doesn't respond on irc super well but tends to read his email11:25
sdake_he has alot of channels to watch after11:25
sdake_brb reconnecting to vpn11:26
sdake_oh already connected nm :)11:27
sdake_samyaple would -v work with -v private and -v non-private in same container?11:30
sdake_or is he container all or nothing11:30
SamYaplei honestly dont know the way they do volume mounts11:30
SamYapleI want to say yes, but i guess it really depends11:30
SamYaplei cant say im 100% sure that a mount namespace set to private doesnt force that option down the chain11:31
SamYapleso ill revert11:31
SamYaplefor simplicities sake well have 2 containers. one for neutron-agents (dhcp l3 metadata), the other for hte plugin agents (openvswitch linuxbridge)11:32
SamYapleI am very disappoint11:34
SamYaplesdake_ I just had a brilliant idea! why dont we just force people to use docker 1.7.0-rc2 forever!?!11:40
sdake_rc2 has the working code?11:41
*** sdake_ has quit IRC11:41
SamYapleyea11:41
SamYaplebut it twas a joke11:41
SamYaplewe may yet have a chance to fix this before liberty11:42
SamYaple1.8 will be out in 2-3 months11:43
SamYapleoh oh oh i know! restart the container between each neutron network creation or deletion!11:45
SamYapleyea thatll work11:45
* SamYaple cries a little11:45
*** dimsum__ has joined #kolla11:53
*** dims_ has joined #kolla11:57
*** erkules_ is now known as erkules11:58
*** erkules has joined #kolla11:58
*** dimsum__ has quit IRC12:00
*** yapeng has joined #kolla12:00
*** sdake has joined #kolla12:02
sdakesamyaple sorry power problems12:03
sdakei thought about whaty ou aid though re rc2 forcing12:03
sdakeI think if we were to do that we would need a group vote on that concept (at wednesdays'meeting)12:04
sdakeI want to release prior to that12:04
sdakeSo lets proceed with a partial revert/rework to get the reelase out12:05
*** yapeng has quit IRC12:05
sdakeand we can mae a deicisoin on wednesday for long term solution12:05
sdakemy concern is that the docer cats may never ccept that featrure12:06
sdakead we would be stuc o rc2 foreer12:06
sdakeso my gut says the oeast risy way is to just do the rewori ajd get that problem solved upstream first12:08
sdakeso I am not totally supportive of that model12:08
sdakeso please proceed with the plan you have of making fat containers except for linuxbridge and ovs12:10
sdakeand we can make a longer term decision on wednesday12:10
sdakei am doubtful the team will support that model12:10
sdake(the rc2 lockin foreer)12:11
sdakebut worth a discussion and I'll add to agenda?12:11
sdakesound good?12:11
sdakeI'll try to get lk4d4 to port into 1.7.1 the change if he can12:11
sdakeso we have a viable solution quickly and are't waiting 2 months for 1.8 of docker12:12
openstackgerritSam Yaple proposed stackforge/kolla: Partially revert thin neutron containers  https://review.openstack.org/19637412:15
SamYaple11:41 < sdake_> rc2 has the working code?12:15
SamYaple11:41 < SamYaple> yea12:15
SamYaple11:42 < SamYaple> but it twas a joke12:15
SamYaple11:43 < SamYaple> we may yet have a chance to fix this before liberty12:15
SamYaple11:43 < SamYaple> 1.8 will be out in 2-3 months12:15
SamYaple11:45 < SamYaple> oh oh oh i know! restart the container between each neutron network creation or deletion!12:15
SamYaple11:45 < SamYaple> yea thatll work12:15
SamYaple11:45  * SamYaple cries a little12:15
* sdake cries alot12:16
sdakefetch me some anestaphine from the medlock ;)12:16
SamYapletesting the patch I just pushed now12:16
sdakeok if you dont want me to bring it up wed I wont :)12:16
SamYapleits not too much rework12:16
sdakecool12:16
SamYaplei dont want to require a dev branch, and an old RC at that to make this work12:16
sdakei'll test once you give it a once-over12:16
SamYaplewell never be viewed as stable12:16
sdakeagree12:17
sdakethat is my concern as well12:17
SamYaplethe fact is most people wont even notice what I am talking about with the namepsaces.... until they start REALLY using it12:17
SamYaplejust trying to head of the storm and get this production ready now12:17
sdakedid mordred get the relicense wrapped up per chance?12:18
SamYaplenot that he has told me12:18
sdakewouldl you ping him monday12:19
sdakesince that seems to be our main blocker atm12:19
sdakemaybe we can release today if this change works12:19
SamYaplesdake: im sure we can release today12:28
SamYapleits not a big blocker for me. ive got my local branch wit hthe modules and im working on rabbit and mysql12:28
sdakeyup i'm good with releasing with a broken horizon12:28
SamYaplehorizon is broken?12:28
sdakesampyaple I want to distribute the work12:28
SamYapleI was always going to be the one going mariadb and rabbit12:28
SamYaplethose are... unique12:29
sdakeso other popele can do it and learn on it as well12:29
SamYaplethe openstack stuff is quick and easy12:29
sdakeyup would still like to distribute the work12:29
SamYapleitll be more conversation about "why" than "how"12:29
SamYapleyea i dont plan on doing it all12:29
SamYapleafter i get this stuff done im on to source releases12:29
sdakehopefully yo ucan fit the single nic thing in there too :)12:30
SamYaplei plan on having a talk about source kolla stuff at the summit12:30
sdakesince that is a blocker for liberty 212:30
SamYapledepends, i can do it with openvswitch12:30
sdakeyup12:30
SamYaplebut that doesnt really have config-internal support12:30
sdakewith ovs12:30
sdakesomeone can add it i'm sure12:31
sdakewe will probably get requests for internal support for ovs12:31
sdakei want to not get in the business of doing all the internal work for everything12:31
sdakebut for ovs i think we should make an effort12:31
sdakesince that is how tripleo deploys12:31
sdakei'll find someone else to do that part12:31
sdakemaybe jpeeler or rhallisey can tackle it12:32
sdakeor inc012:32
sdakethere are alot of folks that do want internal support for tripleo12:32
*** dimsum__ has joined #kolla13:11
*** dims_ has quit IRC13:12
*** dims_ has joined #kolla13:24
*** dimsum__ has quit IRC13:24
openstackgerritMichal Rostecki proposed stackforge/kolla: Add base image for Ubuntu  https://review.openstack.org/19617213:26
*** dimsum__ has joined #kolla13:35
*** dims_ has quit IRC13:35
openstackgerritSam Yaple proposed stackforge/kolla: Partially revert thin neutron containers  https://review.openstack.org/19637413:45
*** dimsum__ has quit IRC13:48
*** dimsum__ has joined #kolla13:48
*** yapeng has joined #kolla13:49
*** yapeng has quit IRC13:53
*** dims_ has joined #kolla14:03
*** dimsum__ has quit IRC14:03
*** rhallisey has joined #kolla14:05
*** yapeng has joined #kolla15:38
*** pbourke has quit IRC15:39
*** pbourke has joined #kolla15:39
*** yapeng has quit IRC15:43
sdakemorning15:45
sdakesamyaple any word on test status?15:45
SamYaplesdake: still testing. containers are building in gate, not on my host15:47
SamYaplereally need a new build system15:47
SamYapleoh well would you look at that15:47
SamYapleSuccessfully built image in /tmp/kolla-workdir.Fdhwo7PK2L/docker/centos/rdo/neutron/neutron-agents15:47
SamYapleyoure a lucky charm sdake15:47
sdakemost people think I'm a headache15:51
*** sdake is now known as steakache15:51
steakacherhallisey asuming that patch wors, can you make yourself available today to approve it15:52
*** steakache is now known as sdake15:52
sdakeneutron-agents are dhcp and friends?15:53
rhalliseywhich15:53
rhalliseyok15:53
sdakethe neutron one15:53
sdakedont approve yet15:53
rhalliseykk15:53
sdakeI need to verify it as well15:53
sdakeafter sam tests it15:53
rhalliseyil be around for another hour15:53
rhalliseythen later today15:53
sdakelater today is good15:53
sdakeit will probably take me 2-3 hours to get my env going15:53
sdakeeven though we are two days past deadline date, its not terrible15:54
sdakeI hate to be a week late with software tho (monday0 :(15:54
*** dimsum__ has joined #kolla16:18
*** dims_ has quit IRC16:19
*** sdake_ has joined #kolla16:19
SamYaplealright how messed with keystone16:20
SamYaplekeystone wont start because keystone is waiting for keystone16:20
SamYaplewho*16:20
SamYaplewaiting for Keystone @ http://192.168.31.12:35357/v2.016:22
SamYaple  File "/usr/bin/keystone-all", line 6, in <module>16:22
SamYaple    from keystone.cmd.all import main16:22
SamYaple  File "/usr/lib/python2.7/site-packages/keystone/cmd/all.py", line 33, in <module>16:22
SamYaple    from keystone.server import eventlet as eventlet_server16:22
SamYaple  File "/usr/lib/python2.7/site-packages/keystone/server/eventlet.py", line 23, in <module>16:22
SamYaple    from oslo_service import service16:22
SamYapleImportError: No module named oslo_service16:22
SamYaplejeez these packages are far from stable16:22
*** sdake has quit IRC16:22
openstackgerritSam Yaple proposed stackforge/kolla: Partially revert thin neutron containers  https://review.openstack.org/19637416:40
SamYaplesdake_ rhallisey harmw jpeeler  ^^ please review https://review.openstack.org/19637416:41
SamYapleit is all functional now, but before you commit please ensure it passes the gate16:41
SamYapleheading to sleep. if it needs another patch onto, feel free to add onto that patch. make sure to add the co-authored-by line so you get credit16:43
sdake_samyaple file bug for keystone oslo_service16:54
sdake_oh off to sleep16:54
sdake_I'll file a bug and finish testing16:54
sdake_hey guys I'm heading to the hifi shop for a couple hours, bbiaf16:55
rhalliseyya something changed in cinder17:28
rhalliseythe package...17:28
rhalliseygrr... broke my container17:29
*** rhallisey is now known as rhallisey|brb17:29
*** dimsum__ has quit IRC17:50
*** jruano has joined #kolla18:09
*** jruano has quit IRC18:34
*** shardy has quit IRC18:47
*** shardy has joined #kolla18:47
*** yapeng has joined #kolla19:15
*** yapeng has quit IRC19:20
*** jruano has joined #kolla19:31
*** jruano has quit IRC20:05
*** dimsum__ has joined #kolla20:16
*** dimsum__ has quit IRC20:19
*** dimsum__ has joined #kolla20:20
*** dimsum__ has quit IRC20:31
*** dimsum__ has joined #kolla20:31
*** jruano has joined #kolla20:38
*** sdake_ has quit IRC20:40
*** dimsum__ has quit IRC20:58
*** dimsum__ has joined #kolla20:59
*** yapeng has joined #kolla21:04
*** yapeng has quit IRC21:09
*** dims has joined #kolla21:10
*** dimsum__ has quit IRC21:10
*** dims_ has joined #kolla21:13
*** dims has quit IRC21:15
*** dims_ has quit IRC21:22
*** nihilifer has quit IRC21:35
*** dims has joined #kolla22:23
*** jruano has quit IRC22:33
*** jruano has joined #kolla22:39
*** dims has quit IRC22:46
*** dims has joined #kolla22:47
*** yapeng has joined #kolla22:53
*** dims has quit IRC22:57
*** yapeng has quit IRC22:58
*** jtriley has joined #kolla23:04
*** dims has joined #kolla23:11
*** jtriley has quit IRC23:24
*** dims has quit IRC23:27

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