| *** ssurana has quit IRC | 00:00 | |
| *** ssurana has joined #openstack-kolla | 00:03 | |
| *** salv-orlando has joined #openstack-kolla | 00:09 | |
| *** diogogmt has joined #openstack-kolla | 00:09 | |
| *** salv-orlando has quit IRC | 00:12 | |
| *** d_code has quit IRC | 00:16 | |
| *** godleon has joined #openstack-kolla | 00:20 | |
| *** ssurana has quit IRC | 00:20 | |
| *** ssurana has joined #openstack-kolla | 00:33 | |
| *** ssurana has quit IRC | 00:34 | |
| *** d_code has joined #openstack-kolla | 00:44 | |
| *** zhiwei has joined #openstack-kolla | 00:45 | |
| *** phuongnh has joined #openstack-kolla | 00:53 | |
| *** tfukushima has joined #openstack-kolla | 00:55 | |
| *** weiyu_ has joined #openstack-kolla | 01:07 | |
| *** salv-orlando has joined #openstack-kolla | 01:12 | |
| *** sacharya has joined #openstack-kolla | 01:12 | |
| *** sacharya has quit IRC | 01:17 | |
| *** salv-orlando has quit IRC | 01:17 | |
| *** britthouser has joined #openstack-kolla | 01:26 | |
| *** britthouser has quit IRC | 01:31 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Log the traceback in the kolla libraries https://review.openstack.org/289667 | 01:40 |
|---|---|---|
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: DO NOT MERGE: TEST MASTER BRANCH https://review.openstack.org/294906 | 01:41 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Pin the base distro release version https://review.openstack.org/319992 | 01:41 |
| *** weiyu_ has quit IRC | 01:46 | |
| *** Jeffrey4l has joined #openstack-kolla | 01:50 | |
| *** weiyu_ has joined #openstack-kolla | 01:51 | |
| *** weiyu_ has quit IRC | 01:55 | |
| *** weiyu_ has joined #openstack-kolla | 01:56 | |
| *** sacharya has joined #openstack-kolla | 01:59 | |
| *** weiyu_ has quit IRC | 02:03 | |
| *** weiyu_ has joined #openstack-kolla | 02:04 | |
| *** daneyon_ has joined #openstack-kolla | 02:06 | |
| *** daneyon_ has quit IRC | 02:11 | |
| *** d_code has quit IRC | 02:32 | |
| *** salv-orlando has joined #openstack-kolla | 02:33 | |
| *** weiyu_ has quit IRC | 02:37 | |
| openstackgerrit | Chris Ricker proposed openstack/kolla: Install missing package in spiceproxy container https://review.openstack.org/323054 | 02:37 |
| *** salv-orlando has quit IRC | 02:38 | |
| *** fragatin_ has joined #openstack-kolla | 02:41 | |
| *** fragatin_ has quit IRC | 02:44 | |
| *** fragatin_ has joined #openstack-kolla | 02:45 | |
| *** fragatina has quit IRC | 02:45 | |
| *** sdake_ has joined #openstack-kolla | 02:46 | |
| *** yuanying has quit IRC | 02:47 | |
| *** fragatin_ has quit IRC | 02:49 | |
| *** sdake has quit IRC | 02:50 | |
| *** yuanying has joined #openstack-kolla | 02:52 | |
| *** weiyu_ has joined #openstack-kolla | 02:52 | |
| *** fragatina has joined #openstack-kolla | 02:57 | |
| *** fragatina has quit IRC | 03:02 | |
| *** sdake_ has quit IRC | 03:07 | |
| *** d_code has joined #openstack-kolla | 03:10 | |
| *** yuanying has quit IRC | 03:43 | |
| *** yuanying has joined #openstack-kolla | 03:47 | |
| *** fragatina has joined #openstack-kolla | 03:48 | |
| openstackgerrit | Swapnil Kulkarni (coolsvap) proposed openstack/kolla: change keyfile to key https://review.openstack.org/320589 | 03:51 |
| *** neophy has joined #openstack-kolla | 03:54 | |
| *** d_code has quit IRC | 03:59 | |
| *** mdnadeem has joined #openstack-kolla | 04:01 | |
| *** mdnadeem has quit IRC | 04:02 | |
| *** mdnadeem has joined #openstack-kolla | 04:03 | |
| *** mdnadeem has quit IRC | 04:09 | |
| *** mdnadeem has joined #openstack-kolla | 04:10 | |
| *** mdnadeem has quit IRC | 04:13 | |
| *** mdnadeem has joined #openstack-kolla | 04:13 | |
| *** mdnadeem has quit IRC | 04:18 | |
| *** mdnadeem has joined #openstack-kolla | 04:18 | |
| *** mdnadeem has quit IRC | 04:19 | |
| *** d_code has joined #openstack-kolla | 04:19 | |
| *** sacharya has quit IRC | 04:19 | |
| *** mdnadeem has joined #openstack-kolla | 04:19 | |
| *** sacharya has joined #openstack-kolla | 04:20 | |
| *** sacharya has quit IRC | 04:21 | |
| *** sacharya_ has joined #openstack-kolla | 04:21 | |
| *** salv-orlando has joined #openstack-kolla | 04:39 | |
| *** salv-orlando has quit IRC | 04:44 | |
| *** zhiwei has quit IRC | 05:04 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Fix the cinder backup permission issue when enable ceph https://review.openstack.org/323121 | 05:04 |
| *** zhiwei has joined #openstack-kolla | 05:04 | |
| *** unicell has quit IRC | 05:05 | |
| *** tfukushima has quit IRC | 05:05 | |
| *** unicell has joined #openstack-kolla | 05:06 | |
| *** unicell1 has joined #openstack-kolla | 05:08 | |
| *** sacharya_ has quit IRC | 05:09 | |
| *** unicell has quit IRC | 05:10 | |
| *** salv-orlando has joined #openstack-kolla | 05:31 | |
| *** diogogmt has quit IRC | 05:31 | |
| *** diogogmt has joined #openstack-kolla | 05:32 | |
| *** neophy has quit IRC | 05:46 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: DO NOT MERGE: TEST LIBERTY BRANCH https://review.openstack.org/323133 | 05:54 |
| *** fragatina has quit IRC | 06:08 | |
| *** Mr_Broken has joined #openstack-kolla | 06:11 | |
| *** Mr_Broken has quit IRC | 06:15 | |
| *** Serlex has joined #openstack-kolla | 06:31 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: DO NOT MERGE: TEST LIBERTY BRANCH https://review.openstack.org/323133 | 06:34 |
| *** Serlex has left #openstack-kolla | 06:39 | |
| *** Serlex has joined #openstack-kolla | 06:39 | |
| *** salv-orlando has quit IRC | 06:49 | |
| *** neophy has joined #openstack-kolla | 06:53 | |
| *** Serlex has quit IRC | 06:54 | |
| *** salv-orlando has joined #openstack-kolla | 06:59 | |
| *** neophy has quit IRC | 07:04 | |
| *** ravig has quit IRC | 07:04 | |
| *** Mr_Broken has joined #openstack-kolla | 07:05 | |
| *** sacharya has joined #openstack-kolla | 07:09 | |
| *** sacharya has quit IRC | 07:14 | |
| *** Serlex has joined #openstack-kolla | 07:23 | |
| *** salv-orlando has quit IRC | 07:29 | |
| *** mikelk has joined #openstack-kolla | 07:30 | |
| *** Jeffrey4l has quit IRC | 07:34 | |
| *** sdake has joined #openstack-kolla | 07:34 | |
| sdake | evening folks | 07:35 |
| *** kklimonda_ has quit IRC | 07:47 | |
| *** kjelly has quit IRC | 07:57 | |
| *** Mr_Broke_ has joined #openstack-kolla | 08:02 | |
| coolsvap | i suspect its not evening for you sdake | 08:02 |
| coolsvap | past midnight i suppose | 08:02 |
| *** Mr_Broken has quit IRC | 08:05 | |
| *** sacharya has joined #openstack-kolla | 08:11 | |
| *** fragatina has joined #openstack-kolla | 08:12 | |
| *** sacharya has quit IRC | 08:16 | |
| *** mdnadeem has quit IRC | 08:16 | |
| *** mdnadeem has joined #openstack-kolla | 08:16 | |
| *** gfidente|afk is now known as gfidente | 08:16 | |
| *** dmk0202 has joined #openstack-kolla | 08:26 | |
| sdake | coolsvap whats up | 08:28 |
| *** jmccarthy has left #openstack-kolla | 08:28 | |
| *** Serlex has quit IRC | 08:28 | |
| *** tfukushima has joined #openstack-kolla | 08:31 | |
| *** jmccarthy has joined #openstack-kolla | 08:31 | |
| *** sdake_ has joined #openstack-kolla | 08:32 | |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implements elemental.yml DSL for openstack-base https://review.openstack.org/322350 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implements elemental.yml DSL for base https://review.openstack.org/321468 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implement elemental DSL parser https://review.openstack.org/322346 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implements elemental.yml DSL for keystone https://review.openstack.org/322533 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Documentation for elemental DSL https://review.openstack.org/322345 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implements elemental.yml DSL for aodh https://review.openstack.org/322336 | 08:33 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Implements elemental.yml DSL for mariadb https://review.openstack.org/322337 | 08:33 |
| coolsvap | sdake, nothing having lunch | 08:34 |
| *** sdake has quit IRC | 08:35 | |
| pbourke_ | sdake_: not so sure on this "elemental" name | 08:36 |
| coolsvap | sdake_, why are you working so late? | 08:36 |
| pbourke_ | whats the thinking behind that? | 08:36 |
| sdake_ | coolsvap because I choose to? | 08:37 |
| coolsvap | sdake_, hmm | 08:37 |
| sdake_ | pbourke_ elemental = play on words - elemental things were the foundation of everything | 08:38 |
| sdake_ | in greek | 08:38 |
| sdake_ | kolla a= greek word for glue | 08:39 |
| sdake_ | along with all the other container stuff (some type of greek word) | 08:39 |
| sdake_ | we are building hte foundation of the containers in a fail | 08:39 |
| sdake_ | ile | 08:39 |
| sdake_ | not fail | 08:39 |
| sdake_ | file | 08:39 |
| sdake_ | and that foundation is elemental ;-) | 08:39 |
| sdake_ | pbourke_ anwer yoru question? | 08:40 |
| sdake_ | it has to have some type of name | 08:40 |
| sdake_ | heat's dsl is called "HOT" | 08:40 |
| pbourke_ | sure :) | 08:40 |
| sdake_ | pbourke_ besides the name, what are your thoughts about its progress | 08:40 |
| pbourke_ | looking good. going through some of the reviews now | 08:41 |
| sdake_ | the parser in particular needs alot ore attention | 08:41 |
| sdake_ | as well as elemetnal.yml files for every container | 08:41 |
| sdake_ | at present it doesn't hose the correct output file to write teh containers to in teh case o faodh (or containers with -base in them) | 08:43 |
| pbourke_ | would you say its ready for review though | 08:43 |
| pbourke_ | theres stuff that clearly needs cleaned up Im sure you dont need me to tell you | 08:44 |
| sdake_ | here is what i'd like | 08:44 |
| sdake_ | the documentaiton ,please review | 08:44 |
| sdake_ | the yaml files, please review | 08:44 |
| sdake_ | the parser, give me some time to make it work properly | 08:44 |
| sdake_ | then we can bikeshet on the implementation | 08:45 |
| pbourke_ | sure thing | 08:45 |
| sdake_ | because the last python code i wrote as over 1 year ago | 08:45 |
| sdake_ | little rusty | 08:45 |
| sdake_ | the documentation in particular is what needs reviewing | 08:45 |
| sdake_ | so make sure the syntax of the elemental.yml file can handle every condition we through at it | 08:46 |
| sdake_ | throw | 08:46 |
| *** papacz has joined #openstack-kolla | 08:49 | |
| openstackgerrit | Merged openstack/kolla: change keyfile to key https://review.openstack.org/320589 | 08:52 |
| pbourke_ | sdake_: its neat to see all the packages neatly enumerated in the yml without all the RUN && \ jinja crap | 08:56 |
| sdake_ | i agree | 08:57 |
| sdake_ | as well as he commands | 08:57 |
| sdake_ | it offloads the coplexity of maintaing the dockerfiles into a parser | 08:57 |
| sdake_ | instead of across our development team | 08:57 |
| sdake_ | now, one could argue the parser is hard to write | 08:57 |
| *** efoley has joined #openstack-kolla | 08:57 | |
| sdake_ | but it can be written in a crappy way that works pretty quickly | 08:57 |
| sdake_ | reference my eparse.py review ;) | 08:57 |
| sdake_ | pbourke_ if you have suggestions for that python code, feel free to make them | 08:58 |
| sdake_ | pbourke_ my thinking was we would merge something that "works" and refactor it into nice tidy pythonic over time | 08:58 |
| pbourke_ | that seems to be the way its worked in the past | 08:58 |
| pbourke_ | someone comes out of nowhere with a massive refactor | 08:59 |
| *** efoley_ has joined #openstack-kolla | 08:59 | |
| sdake_ | i think we are nearing completion of the mssive refactors | 08:59 |
| sdake_ | unless at some point we want to refactor the ansible code to use parallel execution | 08:59 |
| pbourke_ | I would really like to see it merge with some level of basic unit tests though | 09:00 |
| sdake_ | i am not pro at writing unit tests | 09:00 |
| pbourke_ | I can try add something if you like | 09:00 |
| sdake_ | i really struggle | 09:00 |
| sdake_ | yes tha would be great! | 09:00 |
| sdake_ | 90% o fthe problem is I dont know how the tools work | 09:01 |
| sdake_ | so I have to copy and paste from other places | 09:01 |
| sdake_ | and try to figure out how to get it all to work | 09:01 |
| sdake_ | i had to change about 1500 test cases on introduction of oslo.messaging to heat | 09:01 |
| sdake_ | that took about 2 months | 09:01 |
| sdake_ | oslo.messaging = 3-4 days | 09:01 |
| sdake_ | test cases = 2 months | 09:01 |
| pbourke_ | it can be a bit of a black art | 09:02 |
| sdake_ | over time I'd like us to consider breaking our services out into pipelineable services that can run in jenkins or the like | 09:02 |
| *** efoley has quit IRC | 09:03 | |
| sdake_ | so the ockerfile producer would be a pipeline, the docker compiler wold be a pipeline | 09:03 |
| sdake_ | and a third component would glue them together | 09:03 |
| sdake_ | i dont like the structure of build.py | 09:03 |
| sdake_ | I think it should build the imagess | 09:03 |
| sdake_ | do one thing do it well so to speak | 09:03 |
| pbourke_ | im just glad we're putting the brakes on build.py | 09:04 |
| sdake_ | instead it does customization, image building, tagging, registry support, pushing, etc | 09:04 |
| sdake_ | i dont know anone hs put the brakes on build.py | 09:04 |
| sdake_ | but clearly it is out of control | 09:04 |
| *** Mech422 has joined #openstack-kolla | 09:04 | |
| sdake_ | i dont mind if we ship one tool to do the whole job | 09:05 |
| sdake_ | but it should be in separate files | 09:05 |
| sdake_ | with good separation o concerns in the code implementation | 09:05 |
| sdake_ | all of these basic things programmers should be doing were not done in build.py | 09:05 |
| sdake_ | which is ok, that is how software evolves | 09:05 |
| Mech422 | Morning ... can anyone tell me what my network config should look like given 2 nics bonded via ovs, with seperate vlans for internal,external,db, etc ? | 09:05 |
| Mech422 | I tried creating 'layer 3' ('linux' ) devices with the right settings... | 09:06 |
| Mech422 | but kolla wants to do stuff with the ovs setup ? | 09:06 |
| sdake_ | separte vlans? | 09:06 |
| sdake_ | the way kolla isolates networks now is through separate interfaces | 09:06 |
| Mech422 | sdake_: yes, I tried creating seperate 'linux' interfaces for each vlan via ovs and feeding that to kolla...but its wasn't happy | 09:07 |
| sdake_ | i honestly don't see how that would work in a kolla universe without seprate interfaces per network type | 09:08 |
| *** Serlex has joined #openstack-kolla | 09:08 | |
| sdake_ | maybe kolla should offer complete network isolatuion using vlans rather then physical nics | 09:08 |
| sdake_ | but it does not at this time | 09:08 |
| sdake_ | pbourke_ is that acurate? | 09:08 |
| Mech422 | sdake_: ahh - I set the interfaces up at 'ovsIntPorts' in the bare metal | 09:08 |
| Mech422 | s/at/as/ | 09:09 |
| *** SiRiuS has joined #openstack-kolla | 09:09 | |
| sdake_ | Mech422 i dont know details of ovs | 09:09 |
| sdake_ | I can barely get neutron to work ;) | 09:09 |
| sdake_ | we recommend separate networks | 09:09 |
| pbourke_ | im not familiar enough with vlans to advise how they could be used with kolla | 09:09 |
| sdake_ | let me show you a diagram | 09:09 |
| Mech422 | sdake_: LOL - me too... the 'IntPort' ones look like 'real' interface (eg eth0, em1, etc) | 09:09 |
| pbourke_ | im similarly struggling with neutron :/ | 09:10 |
| Mech422 | sdake_: so what would be your advise if you only have a single bonded pair of nics to work with ? | 09:10 |
| pbourke_ | you need a minimum of two interfaces to deploy kolla | 09:10 |
| Mech422 | pbourke_: heh - my next step after getting Kolla working is to try and get midonet working with it... | 09:10 |
| sdake_ | Mech422 http://www.gliffy.com/go/publish/image/10486755/L.png | 09:11 |
| pbourke_ | one for the management network and one for the external network | 09:11 |
| Mech422 | sdake_: thank you | 09:11 |
| *** tfukushima has quit IRC | 09:12 | |
| sdake_ | pbourke_ there is the diagram of our system | 09:12 |
| pbourke_ | sdake_: have you given much thought to how the "interit" functionality of the dsl will work | 09:13 |
| Mech422 | Hmm...so I could use my nics, without bonding -but lose HA... | 09:13 |
| *** pbourke_ has quit IRC | 09:13 | |
| Mech422 | what if I manually create br-ex ? is kolla smart enough not to mess with it ? | 09:13 |
| *** pbourke_ has joined #openstack-kolla | 09:13 | |
| sdake_ | Mech422 ou need more then 1 bonded interface :) | 09:13 |
| sdake_ | for example the controller node | 09:14 |
| sdake_ | needs 3 nics | 09:14 |
| sdake_ | one for he management network (unisolated, compeltely insecure) | 09:14 |
| sdake_ | the api network (locked down, only specific ports open, totally and completely firewalled) | 09:14 |
| Mech422 | sdake_: yeah - I don't have more... I have 2 10GigE, and a GigE for mgmt | 09:14 |
| sdake_ | and the storage network (not isolated, but shouldn't be on an externally routed network) | 09:14 |
| sdake_ | Mech422 thatis the same exact setup I use | 09:15 |
| Mech422 | sdake_: ahh - cool - so you just don't bond the 10GigE's then ? | 09:15 |
| sdake_ | pbourke_ I think inherit will take from one object and copy the contents to a different object | 09:15 |
| sdake_ | pbourke_ so fedora: | 09:16 |
| sdake_ | inherit: centos | 09:16 |
| sdake_ | would just pull in centos into fedora | 09:16 |
| pbourke_ | im just thinking we may not want to pull everything in | 09:16 |
| sdake_ | that way we dont hae to type the same thing a million times for distros which don't have much variance | 09:16 |
| sdake_ | what wouldn't you pull in? | 09:16 |
| pbourke_ | wondering would yaml's anchor feature be a more natural solution | 09:16 |
| sdake_ | i dont knwo yamls anchro feature | 09:17 |
| Mech422 | sdake_: heh - I do that with machine configs - I have a 'hardware' dict in ansible with a 'default' key | 09:17 |
| pbourke_ | you basically tag a block and can then reuse that tag elsewhere in the yaml | 09:17 |
| pbourke_ | so its explicit to the reader which parts are being reused | 09:17 |
| sdake_ | pbourke_ that sounds pretty good | 09:17 |
| sdake_ | can you make a proof of concept from the aodh or openstack-base or base containers | 09:17 |
| pbourke_ | https://gist.github.com/bowsersenior/979804#file-stooges-yml | 09:17 |
| sdake_ | openstack-base would probablybe best | 09:18 |
| pbourke_ | yeah ill hack on these a bit today | 09:18 |
| Mech422 | assuming I was dumb enough to want to play with the Kolla networking stuff - any idea which ansible roles are setting it up ? | 09:18 |
| pbourke_ | id like the derivative distro support to be solid before merging as that has proved a core issue to be solved in previous solutions | 09:18 |
| *** daneyon_ has joined #openstack-kolla | 09:20 | |
| sdake_ | pbourke_ if you modify any of my patches let me knwo so I can reintegrate the changes into my patch stream | 09:20 |
| sdake_ | pbourke_ agree that is why there a bi fat -2 on the first review in the queue | 09:20 |
| pbourke_ | Mech422: most of it is just in the config templates, where we bind the services to "api_interface" | 09:20 |
| sdake_ | that was the core issue we solved | 09:21 |
| sdake_ | but we have identified a whole slew of othe problems we want to solve | 09:21 |
| Mech422 | pbourke_: thanks | 09:21 |
| sdake_ | other then multidistro, jinja2 didn't solve much else | 09:21 |
| coolsvap | sdake_, i updated the parser patch yesterday seems like it is overwritten in patch 5, will update it again | 09:21 |
| patchbot | coolsvap: https://review.openstack.org/#/c/5/ - openstack-infra/system-config - Fix problem with jenkins known_hosts url. (MERGED) | 09:21 |
| sdake_ | coolsvap if you do that pull down he whole patch stream | 09:21 |
| *** matrohon has joined #openstack-kolla | 09:22 | |
| sdake_ | so when I pull down the whoel patch stream I have your changes in it | 09:22 |
| pbourke_ | Mech422: there's also stuff then in the neutron tasks to configure the neutron_external_interface for the external network. those are the two main variables to grep for | 09:22 |
| coolsvap | sdake_, ya | 09:22 |
| coolsvap | will do | 09:22 |
| *** berendt has quit IRC | 09:22 | |
| Mech422 | pbourke_: I think thats whats biting me atm - I try that first - thanks again! | 09:22 |
| sdake_ | Mech422 what you want is to use vlan's with 1 network interface to isolate traffic | 09:22 |
| sdake_ | what kolla offers is multiple network interfaces without vlan to isolate traffic | 09:22 |
| coolsvap | this is something new of patchbot, we don't need to provide complete url, didn't knew that | 09:22 |
| sdake_ | exception being in the neutron external network which uses vlanv/vxlan to isolate traffic iiuc | 09:23 |
| Mech422 | sdake_: yeah - I realize I'm going out in the weeds... | 09:23 |
| Mech422 | sdake_: and I want to try and get midonet into the mix as well | 09:23 |
| pbourke_ | sdake_: only if you use ovs or another driver that supports overlay networks | 09:23 |
| sdake_ | kolla implements ovs and linuxbridge | 09:23 |
| Mech422 | sdake_: I'm hoping if i manually configure br-ex, etc - that kolla won't touch it | 09:24 |
| pbourke_ | Mech422: it will | 09:24 |
| *** daneyon_ has quit IRC | 09:24 | |
| pbourke_ | :p | 09:24 |
| sdake_ | gah why would you want to manually configure your bridges? | 09:24 |
| Mech422 | pbourke_: well, crud puppies | 09:24 |
| Mech422 | sdake_: so I can go out in the weeds ?? :-) | 09:24 |
| sdake_ | Mech422 try hitting up vhosakot or dave_mccowan tomorrow | 09:24 |
| sdake_ | they are both networking nerds | 09:25 |
| sdake_ | sbezverk is a CCIE | 09:25 |
| Mech422 | sdake_: oh - thank you! | 09:25 |
| sdake_ | and even thugh I am not big on certifications, CCIE is a super hard certification | 09:25 |
| sdake_ | sbezverk can likely get you pointed in the right direction or translate to me so we can get work scheduled | 09:25 |
| sdake_ | or both :) | 09:25 |
| Mech422 | sdake_: thanks - I didn't really think anyone else would be interested... | 09:26 |
| Mech422 | sdake_: though the neutron midonet plugin looks interesting | 09:26 |
| sdake_ | if it means we need to implement some magic to add vlan tagging to specific networks, I thnk that may make sense | 09:26 |
| sdake_ | but i am not a networking genius | 09:26 |
| pbourke_ | Mech422: it would be pretty easy to disable it though. kolla is simple enough to hack on once you're familiar with the layout | 09:26 |
| Mech422 | cool :-) | 09:27 |
| Mech422 | rather then worrying about a bunch of wierd network configs - it might just be enough to have kolla not touch br-ex, etc. if they already exist | 09:28 |
| sdake_ | that is easy enough | 09:28 |
| Mech422 | then people can manually setup whatever they want/need without pestering you guys | 09:28 |
| sdake_ | just take out the task | 09:28 |
| sdake_ | we like being pestered - it lets us know we are loved :) | 09:29 |
| Mech422 | cool - I'll look at that once my coffee is ready :-) | 09:29 |
| sdake_ | but seriously, if kolla doesnt work correctly we want to make it work correctly | 09:29 |
| Mech422 | LOL - you're very loved - kolla is awesome :-) | 09:29 |
| sdake_ | well thanks for that - i'm sure the community appreciates the positive feedback | 09:29 |
| Mech422 | sdake_: I think networking will be one area where 'providers' will have all sorts of weird setups | 09:29 |
| sdake_ | it was a long painful road to get there | 09:29 |
| Mech422 | sdake_: storage too - sans, ceph, s3 buckets, etc | 09:30 |
| Mech422 | sdake_: I can imagine - it took me a month just to write the ansible stuff to configure kolla :-P | 09:30 |
| sdake_ | kolla iplements ceph | 09:30 |
| sdake_ | and we expect people to deploy with it too | 09:30 |
| Mech422 | sdake_: I saw that :-) ceph rocks! | 09:30 |
| sdake_ | what stuf did you write to confgiure kolla | 09:31 |
| sdake_ | that is probably stuff that needs to go upstream at some point :) | 09:31 |
| Mech422 | sdake_: the kolla side was just some simple stuff to pre-configure the config files... | 09:31 |
| sdake_ | you mean /etc/kolla/config? | 09:31 |
| Mech422 | sdake_: I also built users for runnign stuff and all that jazz | 09:31 |
| sdake_ | or /etc/kolla/globals.yml? | 09:31 |
| Mech422 | mostly globals.yml at this point - I _just_ got deploy working tonight | 09:32 |
| Mech422 | so I still have to configure a lot of stuff | 09:32 |
| sdake_ | nice- an AIO deploy? | 09:32 |
| Mech422 | nah - umm... 6 nodes ? my ansible stuff build multinode.yml dynamically | 09:32 |
| Mech422 | 7 nodes if you count the deployment host | 09:33 |
| sdake_ | ok past my bedtime | 09:33 |
| sdake_ | its 2:30 | 09:33 |
| sdake_ | night all | 09:34 |
| sdake_ | Mech422 sbezverk is on the esat coast in candara iirc | 09:34 |
| sdake_ | so he should be around in 4-5 hours | 09:34 |
| Mech422 | sdake_: thanks! Gnite! | 09:34 |
| Mech422 | its 2:30 AM here...so I'll prolly be heading to bed myself | 09:34 |
| sdake_ | pbourke_ ancor thing looks ok - kind of hard to read | 09:34 |
| sdake_ | pbourke_ but if there is already a cut and repace feature of yml we should probably use it | 09:35 |
| pbourke_ | sure | 09:35 |
| sdake_ | pbourke_ re eparse, at present its broken on the adoh image | 09:35 |
| sdake_ | pbourke_ if you get into it and decide it needs major rewriting, knock yurself out | 09:36 |
| Mech422 | oh btw - I had to manually upgrade the pip and virtualenv/tox on ubuntu 14.04...I finally found your messages about them in the irc logs sdake_ :-) | 09:36 |
| sdake_ | i knwo you youngins have your opinions about python | 09:36 |
| *** SiRiuS has quit IRC | 09:36 | |
| sdake_ | just like i do with my old daddy school lanuage C:) | 09:36 |
| pbourke_ | :) | 09:37 |
| pbourke_ | sdake_: by the way, any idea how we can get your infra patch merged to add the oraclelinux gates | 09:38 |
| pbourke_ | poked a few times on #openstack-infra, no luck | 09:38 |
| sdake_ | pbourke_ they are all asleep atm | 09:39 |
| pbourke_ | ah, maybe wrong timezone for me | 09:39 |
| sdake_ | or alternatively in some foreign country doing something related to openstack | 09:39 |
| sdake_ | I have pinged them a few times over the last week | 09:39 |
| sdake_ | and that got 1 +2 :) | 09:39 |
| sdake_ | so I'll hunt down a second one, if you remind me hte morning it might work best | 09:39 |
| sdake_ | patch sitting around with 1 +2 for 2 weeks, logn lag time | 09:40 |
| sdake_ | after summit, openstack lags or about 2-4 weeks | 09:40 |
| sdake_ | or/for | 09:40 |
| *** SiRiuS has joined #openstack-kolla | 09:40 | |
| sdake_ | pbourke_ probably the best one to go with is bse, then that is the most complexx | 09:40 |
| sdake_ | also I didn't iplement any ubuntu rendering of the images | 09:40 |
| sdake_ | only centos | 09:40 |
| pbourke_ | right | 09:41 |
| sdake_ | and there is that major glaring bug that images don't gte created if they have a -base in their catalog | 09:41 |
| sdake_ | but i am not superman and need to hit the rack :) | 09:41 |
| sdake_ | pbourke_ what does this do https://gist.github.com/bowsersenior/979804#file-stooges-yml-L10 | 09:43 |
| *** weiyu_ has quit IRC | 09:45 | |
| *** weiyu_ has joined #openstack-kolla | 09:46 | |
| openstackgerrit | Merged openstack/kolla: Fix the incompatibility for ansible 2.1.0 https://review.openstack.org/321754 | 09:47 |
| *** phuongnh has quit IRC | 09:47 | |
| openstackgerrit | Merged openstack/kolla: Fix the cinder backup permission issue when enable ceph https://review.openstack.org/323121 | 09:48 |
| *** phuongnh has joined #openstack-kolla | 09:49 | |
| pbourke_ | sdake_: actually not familiar with that line. I think it will take the content of default and override the URL and stooges keys | 09:50 |
| pbourke_ | the merging stuff is a little funky Im not sure we'll need it | 09:50 |
| sdake_ | i have a use case for merging i need now | 09:50 |
| sdake_ | are yo utalkingabout the elemental-merge.yml file i talked about in the docs? | 09:51 |
| pbourke_ | no i like that | 09:51 |
| pbourke_ | I mean the yaml anchor merging | 09:51 |
| sdake_ | the reaon I added that inherit thing is i didn't wnt to repeat oraclelinux almost word for word from the entos section | 09:52 |
| sdake_ | for like 50 docker file groupings | 09:56 |
| *** sdake_ is now known as sdake | 09:57 | |
| pbourke_ | so how do we handle for example something like the part where ol requires rpm --nodeps for the packages section | 09:59 |
| *** rhallisey has joined #openstack-kolla | 10:01 | |
| pbourke_ | i guess we could specify that section in the yml and the parser will do the right thing | 10:01 |
| pbourke_ | its the same as what Im proposing with anchors really | 10:01 |
| *** mgoddard has joined #openstack-kolla | 10:04 | |
| *** pbourke_ has quit IRC | 10:09 | |
| *** pbourke_ has joined #openstack-kolla | 10:10 | |
| *** sacharya has joined #openstack-kolla | 10:12 | |
| Mech422 | I'm an idgit... | 10:12 |
| Mech422 | looks like I can use the bonded interfaces just by creating a couple of pairs of veth interfaces, tagged with the right vlans and pass those to kolla | 10:13 |
| *** sacharya has quit IRC | 10:16 | |
| *** mdnadeem has quit IRC | 10:16 | |
| *** mdnadeem has joined #openstack-kolla | 10:17 | |
| *** weiyu_ has quit IRC | 10:17 | |
| *** mdnadeem has quit IRC | 10:17 | |
| *** mdnadeem has joined #openstack-kolla | 10:17 | |
| *** zhiwei has quit IRC | 10:18 | |
| *** mdnadeem has quit IRC | 10:19 | |
| *** mdnadeem has joined #openstack-kolla | 10:19 | |
| openstackgerrit | Merged openstack/kolla: Compile the messages in the horizon https://review.openstack.org/322020 | 10:22 |
| sdake | pbourke_ for that case (nodeps) nodeps will always be the cae | 10:25 |
| sdake | I was thinking we would just have a oraclelinux parser for that part of the yaml markup | 10:25 |
| sdake | pbourke_ with the nodeps stuff | 10:25 |
| sdake | alternativelly, the parser knows which section its parsing | 10:26 |
| sdake | if its parsing anything oraclelinux and there is a repo section, it could use the special parser for le that prints out --nodeps | 10:26 |
| *** rhallisey has quit IRC | 10:27 | |
| *** weiyu_ has joined #openstack-kolla | 10:28 | |
| *** berendt has joined #openstack-kolla | 10:28 | |
| *** weiyu_ has quit IRC | 10:30 | |
| *** weiyu_ has joined #openstack-kolla | 10:31 | |
| *** weiyu_ has quit IRC | 10:33 | |
| *** SiRiuS has quit IRC | 10:42 | |
| *** SiRiuS has joined #openstack-kolla | 10:42 | |
| pbourke_ | sdake: still around? | 10:49 |
| sdake | shoot | 10:50 |
| pbourke_ | do we support per node augment files? | 10:50 |
| pbourke_ | see line 183 here https://review.openstack.org/#/c/322334/1/doc/pci-passthrough.rst | 10:50 |
| patchbot | pbourke_: patch 322334 - kolla (stable/mitaka) - Adds documentation of how to perform SR-IOV | 10:50 |
| pbourke_ | actually its in the docs so I guess we do :) | 10:53 |
| pbourke_ | didnt know that though | 10:53 |
| *** tfukushima has joined #openstack-kolla | 10:53 | |
| coolsvap | but we dont have docs backported to stable branche | 10:54 |
| coolsvap | i saw this patchset today | 10:54 |
| openstackgerrit | Merged openstack/kolla: Adding Upgrade to iscsid/tgtd container https://review.openstack.org/314096 | 10:56 |
| pbourke_ | coolsvap: ah just noticed its targeted for mitaka | 10:58 |
| coolsvap | i think both doc and associated change should go through master to stable branches | 10:59 |
| *** tfukushima has quit IRC | 10:59 | |
| pbourke_ | +1 | 11:01 |
| *** rhallisey has joined #openstack-kolla | 11:06 | |
| *** phuongnh has quit IRC | 11:07 | |
| *** dmk0202 has quit IRC | 11:07 | |
| openstackgerrit | Merged openstack/kolla: Fix kolla-build return codes https://review.openstack.org/322115 | 11:08 |
| openstackgerrit | Merged openstack/kolla: Update Documentation https://review.openstack.org/319899 | 11:08 |
| openstackgerrit | Merged openstack/kolla: Compile the messages in the horizon https://review.openstack.org/322021 | 11:08 |
| *** godleon has quit IRC | 11:09 | |
| openstackgerrit | Merged openstack/kolla-kubernetes: Add more details to the quickstart and usage guide. https://review.openstack.org/321905 | 11:12 |
| openstackgerrit | Merged openstack/kolla-kubernetes: Adding kuebctl setup task to point to local hyperkube https://review.openstack.org/321886 | 11:20 |
| openstackgerrit | Merged openstack/kolla-kubernetes: Add bootstrap directory to usr/local/share/kolla-kubernetes https://review.openstack.org/321897 | 11:22 |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Move developer related docs into their own section https://review.openstack.org/323298 | 11:31 |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Retitle quick start to make it more obvious in index https://review.openstack.org/323304 | 11:35 |
| *** dmk0202 has joined #openstack-kolla | 11:42 | |
| *** weiyu has joined #openstack-kolla | 11:42 | |
| sdake | coolsvap we don't backport documentation becausae our docs are published on the internets | 11:46 |
| sdake | and for some godawful reason people want to write the docs separately from the patch | 11:46 |
| sdake | which means, they dont get rwritten | 11:46 |
| *** mdnadeem has quit IRC | 11:46 | |
| sdake | but its to late now, we dont knwo what doc patches were made to cherrypick | 11:46 |
| sdake | because theywere nt tracked with bug ids | 11:47 |
| *** mdnadeem has joined #openstack-kolla | 11:47 | |
| sdake | if you want to make the docs sync, sort out how to do the backporting | 11:47 |
| sdake | and we can do that | 11:47 |
| sdake | otherwise, I think we have more important things to concern ourselves with | 11:47 |
| *** mdnadeem has quit IRC | 11:47 | |
| sdake | nobody will actually read the documenation shipped witht eh software | 11:47 |
| sdake | they will go to the rendered documentation | 11:48 |
| coolsvap | sdake, yes | 11:48 |
| *** mdnadeem has joined #openstack-kolla | 11:48 | |
| *** mdnadeem has quit IRC | 11:48 | |
| sdake | coolsvap yes to what, I said 10 different things there | 11:48 |
| *** mdnadeem has joined #openstack-kolla | 11:48 | |
| *** mdnadeem has quit IRC | 11:48 | |
| coolsvap | i just wanted to make pbourke_ aware that the docs and related fix is targeted to mitaka not master | 11:49 |
| *** mdnadeem has joined #openstack-kolla | 11:49 | |
| *** mdnadeem has quit IRC | 11:49 | |
| sdake | patches should go in master first obviously | 11:49 |
| *** mdnadeem has joined #openstack-kolla | 11:49 | |
| coolsvap | i know the reasoning behind the doc thing | 11:49 |
| *** mdnadeem has quit IRC | 11:49 | |
| coolsvap | just wanted to make sure he's also aware that we need to ask it to submit to master not to mitaka directly | 11:50 |
| *** mdnadeem has joined #openstack-kolla | 11:50 | |
| *** Mr_Broke_ has quit IRC | 11:50 | |
| *** mdnadeem has quit IRC | 11:50 | |
| *** mdnadeem has joined #openstack-kolla | 11:50 | |
| *** Mr_Broken has joined #openstack-kolla | 11:50 | |
| *** mdnadeem has quit IRC | 11:51 | |
| coolsvap | nothing else | 11:51 |
| sdake | cool | 11:51 |
| coolsvap | :) | 11:51 |
| *** mliima has joined #openstack-kolla | 11:51 | |
| *** mdnadeem has joined #openstack-kolla | 11:52 | |
| *** Mr_Broken has quit IRC | 11:55 | |
| *** weiyu has quit IRC | 12:01 | |
| *** mdnadeem has quit IRC | 12:01 | |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Implement elemental DSL parser https://review.openstack.org/322346 | 12:04 |
| *** Mr_Broken has joined #openstack-kolla | 12:05 | |
| *** mgoddard has quit IRC | 12:08 | |
| *** mgoddard has joined #openstack-kolla | 12:08 | |
| *** JoseMello has joined #openstack-kolla | 12:12 | |
| *** sacharya has joined #openstack-kolla | 12:13 | |
| openstackgerrit | Merged openstack/kolla: Move developer related docs into their own section https://review.openstack.org/323298 | 12:13 |
| openstackgerrit | Merged openstack/kolla: Retitle quick start to make it more obvious in index https://review.openstack.org/323304 | 12:13 |
| openstackgerrit | Merged openstack/kolla: Use file module instead of running rm to remove a directory https://review.openstack.org/321832 | 12:15 |
| *** d_code has quit IRC | 12:16 | |
| *** sacharya has quit IRC | 12:17 | |
| *** dwalsh has joined #openstack-kolla | 12:17 | |
| *** Jeffrey4l has joined #openstack-kolla | 12:20 | |
| *** d_code has joined #openstack-kolla | 12:23 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Fix the cinder backup permission issue when enable ceph https://review.openstack.org/323330 | 12:23 |
| sbezverk | rhallisey ping | 12:24 |
| rhallisey | sbezverk, hey | 12:24 |
| *** dave-mccowan has joined #openstack-kolla | 12:25 | |
| *** cu5 has joined #openstack-kolla | 12:26 | |
| sbezverk | rhallisey I figured out the issue with dns.. it is HostNetwork: True, when I start busybox with this setting, it mounts host's resolv.conf, without HostNetwork true, it uses kubelet's provided resolv.conf | 12:26 |
| sbezverk | rhallisey it looks like this issue will be fixed in 1.12 docker.. currently specifying dns and using host net are mutually exclusive.. | 12:27 |
| *** sdake has quit IRC | 12:27 | |
| rhallisey | og | 12:28 |
| rhallisey | oh | 12:28 |
| rhallisey | sbezverk, interesting | 12:28 |
| sbezverk | rhallisey yeah, unfortunately it is interesting with -1 ;-) | 12:29 |
| rhallisey | sbezverk, :) no worries. I'm glad you tracked this down | 12:31 |
| rhallisey | can you comment on this in your spec | 12:31 |
| rhallisey | just so we track the issue | 12:31 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Fix the cinder backup permission issue when enable ceph https://review.openstack.org/323338 | 12:33 |
| *** ppowell has joined #openstack-kolla | 12:37 | |
| sbezverk | rhallisey will do | 12:40 |
| *** zhiwei has joined #openstack-kolla | 12:47 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Bump the ansible version to 2.0.2.0-1 in the kolla-toolbox https://review.openstack.org/311625 | 12:47 |
| *** zhiwei has quit IRC | 12:51 | |
| *** SiRiuS has quit IRC | 12:54 | |
| *** absubram has joined #openstack-kolla | 12:54 | |
| *** absubram_ has joined #openstack-kolla | 12:55 | |
| *** daneyon_ has joined #openstack-kolla | 12:56 | |
| *** diogogmt has quit IRC | 12:56 | |
| *** efoley_ is now known as efoley | 12:57 | |
| *** diogogmt has joined #openstack-kolla | 12:57 | |
| *** jtriley has joined #openstack-kolla | 12:58 | |
| *** absubram has quit IRC | 12:59 | |
| *** absubram_ is now known as absubram | 12:59 | |
| *** dwalsh has quit IRC | 13:00 | |
| *** daneyon_ has quit IRC | 13:01 | |
| *** SiRiuS has joined #openstack-kolla | 13:03 | |
| *** stvnoyes has joined #openstack-kolla | 13:03 | |
| *** dwalsh has joined #openstack-kolla | 13:03 | |
| *** dwalsh has quit IRC | 13:08 | |
| *** fragatina has quit IRC | 13:09 | |
| *** fragatina has joined #openstack-kolla | 13:10 | |
| *** fragatina has quit IRC | 13:11 | |
| *** dwalsh has joined #openstack-kolla | 13:12 | |
| *** dwalsh has quit IRC | 13:13 | |
| *** dwalsh has joined #openstack-kolla | 13:14 | |
| *** fragatina has joined #openstack-kolla | 13:15 | |
| *** ayoung has joined #openstack-kolla | 13:22 | |
| *** diogogmt has quit IRC | 13:25 | |
| *** absubram has quit IRC | 13:33 | |
| *** diogogmt has joined #openstack-kolla | 13:36 | |
| *** mgoddard_ has joined #openstack-kolla | 13:45 | |
| *** mgoddard has quit IRC | 13:45 | |
| *** d_code has quit IRC | 13:47 | |
| *** mgoddard__ has joined #openstack-kolla | 13:55 | |
| *** mgoddard_ has quit IRC | 13:58 | |
| openstackgerrit | Merged openstack/kolla: Log the traceback in the kolla libraries https://review.openstack.org/289667 | 13:59 |
| *** diogogmt has quit IRC | 14:01 | |
| *** mdnadeem has joined #openstack-kolla | 14:02 | |
| *** d_code has joined #openstack-kolla | 14:03 | |
| openstackgerrit | Serguei Bezverkhi proposed openstack/kolla-kubernetes: Kolla Kubernetes Service Discovery BP https://review.openstack.org/317537 | 14:03 |
| Mr_Broken | Hey guys | 14:04 |
| Mr_Broken | using kolla-ansible reconfigure in stable/mitaka | 14:04 |
| Mr_Broken | works great, but how do i remove the line from existing config ? | 14:04 |
| Mr_Broken | or at least comment it ? | 14:04 |
| Mr_Broken | trying to change glance-api.conf, and remove ( or comment ) #filesystem_store_datadir = /var/lib/glance/images/ | 14:05 |
| Mr_Broken | because i want only swift to handle glance backend | 14:05 |
| rhallisey | Mr_Broken, when you comment it out what happens? Does it get added back in? | 14:05 |
| Mr_Broken | nothing happens when i comment it | 14:06 |
| Mr_Broken | like it getting ignored | 14:06 |
| rhallisey | so reconfigure never picks it up? | 14:06 |
| berendt | in this case it should be sufficient to overwrite the value of default_store in [glance_storage] | 14:06 |
| openstackgerrit | Mauricio Lima proposed openstack/kolla: Manila driver session is needed only in manila-share https://review.openstack.org/311232 | 14:06 |
| sbezverk | Mr_Broken when you change default store type to swift, all file related config becomes irrelevant. | 14:06 |
| sbezverk | Mr_Broken I do not think you need to worry about it | 14:07 |
| Mr_Broken | i did change it | 14:07 |
| Mr_Broken | but now i am having a problem when trying to create image : | 14:07 |
| Mr_Broken | let ma paste sec | 14:07 |
| sbezverk | Mr_Broken have you tested your swift to make sure that it is working? | 14:08 |
| Mr_Broken | yes | 14:08 |
| Mr_Broken | its working | 14:08 |
| Mr_Broken | can access it with swift client itself | 14:08 |
| Mr_Broken | please have a look : | 14:10 |
| Mr_Broken | http://paste.openstack.org/show/506590/ | 14:10 |
| berendt | have you configure swift inside glance? | 14:10 |
| Mr_Broken | want me to paste glance-api.conf ? and glance-swift.conf ? | 14:10 |
| Mr_Broken | i tried configuring it for like 6 hours now ;] | 14:10 |
| berendt | I think it makes sense to paste the generated glance-api.conf file of the controller node | 14:11 |
| *** ravig has joined #openstack-kolla | 14:12 | |
| sbezverk | Mr_Broken please paste those config files.. Cannot say I saw this error before.. | 14:13 |
| Mr_Broken | http://paste.openstack.org/show/506591/ | 14:13 |
| Mr_Broken | sbezverk: same here - didnt have those issues in Liberty #JSYK ;] | 14:13 |
| sbezverk | Mr_Broken man, I think you are missing bunch of config for swift | 14:17 |
| sbezverk | in glance-api.conf | 14:17 |
| sbezverk | unless you did not paste all info | 14:17 |
| Mr_Broken | i didnt config all | 14:18 |
| Mr_Broken | i didnt paste all sorry | 14:18 |
| Mr_Broken | there were some passwords and logins :> | 14:18 |
| Mr_Broken | didnt bother to edit :] | 14:18 |
| Mr_Broken | i pasted only sections that are responsible for glance | 14:18 |
| Mr_Broken | you want it all ? | 14:18 |
| sbezverk | Mr_Broken, just remove passwords | 14:19 |
| Mr_Broken | kk | 14:19 |
| Mr_Broken | sec | 14:19 |
| Mr_Broken | here You go: | 14:20 |
| Mr_Broken | http://paste.openstack.org/show/506592/ | 14:20 |
| Mr_Broken | any ideas ? | 14:25 |
| Mr_Broken | need to switch locations will be offline for like 45mins. | 14:26 |
| *** Mr_Broken has quit IRC | 14:27 | |
| *** b_bezak has joined #openstack-kolla | 14:27 | |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Implement elemental DSL parser https://review.openstack.org/322346 | 14:27 |
| *** Mr_Broken has joined #openstack-kolla | 14:28 | |
| berendt | does it makes sense to interpret the include_footer/include_header files as jinja templates? at least for me this would be useful. | 14:29 |
| *** Mr_Broken has quit IRC | 14:32 | |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Implements elemental.yml DSL for base https://review.openstack.org/321468 | 14:32 |
| *** cu5 has quit IRC | 14:33 | |
| berendt | It looks like the merge of https://review.openstack.org/#/c/322115/ introduced a new bug, I mentioned this in the already closes bug report. Should the bug report be re-opened? | 14:34 |
| patchbot | berendt: patch 322115 - kolla - Fix kolla-build return codes (MERGED) | 14:34 |
| coolsvap | berendt, no open a new bug report | 14:35 |
| *** inc0 has joined #openstack-kolla | 14:35 | |
| berendt | coolsvap ok | 14:35 |
| berendt | coolsvap reported as https://bugs.launchpad.net/kolla/+bug/1587501 | 14:37 |
| openstack | Launchpad bug 1587501 in kolla "TypeError exception when running kolla-build with --template-only" [Undecided,New] | 14:37 |
| Lyncos | inc1 ping | 14:37 |
| coolsvap | berendt, will you be working on this? | 14:38 |
| inc0 | good morning | 14:39 |
| inc0 | I'm back to the land of the living | 14:39 |
| *** mgoddard__ has quit IRC | 14:39 | |
| *** mgoddard has joined #openstack-kolla | 14:39 | |
| inc0 | Lyncos, wassup? I'm still 0 :( | 14:39 |
| berendt | coolsvap no I think it the author of the merged patch should have a look on it | 14:39 |
| Lyncos | inc0 lol sorry | 14:39 |
| Lyncos | inc0 we figured out our millions mount points problem | 14:39 |
| coolsvap | maybe we can revert it for time being then | 14:39 |
| coolsvap | does not look like a big fix | 14:40 |
| berendt | I think a revert is not necessary, normal builds are working | 14:40 |
| inc0 | Lyncos, do tell please | 14:40 |
| Lyncos | inc0 it was because in ubuntu by default LXC is installed .. we removed it and now the when we restart our DHCP container it's stable | 14:41 |
| inc0 | oh? interesting | 14:41 |
| Lyncos | it is some sort of incompatibility with LXC / Docker in ubuntu | 14:41 |
| inc0 | maybe something we should do in our base image | 14:41 |
| Lyncos | it's on the host system | 14:41 |
| inc0 | ahh | 14:41 |
| berendt | Lyncos a note in the getting started documentation would be nice | 14:41 |
| inc0 | well, truth be told, docker did state it in documentation | 14:41 |
| coolsvap | lets see, I dont know will miller lets see will wait for some time | 14:41 |
| Lyncos | inc0 yes | 14:42 |
| inc0 | but sure, pop a patch to make node in our quickstart | 14:42 |
| inc0 | I'm glad that you got this figured out | 14:42 |
| coolsvap | mgoddard, wmiller ping can you guys have a look at https://bugs.launchpad.net/kolla/+bug/1587501 | 14:42 |
| openstack | Launchpad bug 1587501 in kolla "TypeError exception when running kolla-build with --template-only" [High,New] | 14:42 |
| *** diogogmt has joined #openstack-kolla | 14:43 | |
| mgoddard | coolsvap, sure | 14:43 |
| coolsvap | seems like aftereffect of patch 322115 | 14:43 |
| patchbot | coolsvap: https://review.openstack.org/#/c/322115/ - kolla - Fix kolla-build return codes (MERGED) | 14:43 |
| openstackgerrit | Merged openstack/kolla: Fix the cinder backup permission issue when enable ceph https://review.openstack.org/323338 | 14:49 |
| *** Lyncos has quit IRC | 14:50 | |
| *** ayoung has quit IRC | 14:51 | |
| mgoddard | coolsvap: I think you're right. wmiller is on it | 14:52 |
| openstackgerrit | Merged openstack/kolla: Set ServerTokens parameter to 'Prod' for Ubuntu Apache config https://review.openstack.org/321798 | 14:54 |
| openstackgerrit | Merged openstack/kolla: Manila driver session is needed only in manila-share https://review.openstack.org/311232 | 14:55 |
| *** mliima has quit IRC | 15:06 | |
| *** d_code has quit IRC | 15:11 | |
| Mech422 | anyone happen to know a trick for dynamically querying ansible vars? I'm trying to get say ansible_eth0 when 'eth0' varies at runtime | 15:11 |
| pbourke_ | hostvars | 15:12 |
| Mech422 | pbourke_: problem is ansible_eth0 isn't a dict? so I can't do ansible[ interface] to get it... | 15:13 |
| Mech422 | pbourke_: ahh...I see...ansible_eth0 is a key in hostvars.... | 15:14 |
| Mech422 | never thought of it that way - thanks! | 15:14 |
| *** d_code has joined #openstack-kolla | 15:14 | |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Implements elemental.yml DSL for base https://review.openstack.org/321468 | 15:14 |
| openstackgerrit | Paul Bourke proposed openstack/kolla: Implement elemental DSL parser https://review.openstack.org/322346 | 15:14 |
| *** zhiwei has joined #openstack-kolla | 15:15 | |
| *** ravig has quit IRC | 15:18 | |
| *** zhiwei has quit IRC | 15:19 | |
| *** vhosakot has joined #openstack-kolla | 15:21 | |
| Mech422 | huh - ansible doesn't notice veth interfaces... no ansible_foo.ipv4.address for me :-P | 15:22 |
| *** mliima has joined #openstack-kolla | 15:23 | |
| *** berendt has quit IRC | 15:31 | |
| *** openstackgerrit has quit IRC | 15:33 | |
| *** openstackgerrit has joined #openstack-kolla | 15:33 | |
| *** b_bezak has quit IRC | 15:36 | |
| *** b_bezak has joined #openstack-kolla | 15:36 | |
| *** sacharya has joined #openstack-kolla | 15:38 | |
| *** matrohon_ has joined #openstack-kolla | 15:38 | |
| *** matrohon_ has quit IRC | 15:38 | |
| *** b_bezak has quit IRC | 15:41 | |
| *** dmk0202 has quit IRC | 15:42 | |
| *** ayoung has joined #openstack-kolla | 15:45 | |
| *** dwalsh has quit IRC | 15:48 | |
| inc0 | Mech422, hmm, that's strange | 15:48 |
| Mech422 | inc0: actually - it looks like thats an ordering problem on my side | 15:48 |
| Mech422 | inc0: I referenced it before the playbook runs that creates it :-P | 15:49 |
| inc0 | yeah, having veth created helps with using it | 15:49 |
| inc0 | ;) | 15:49 |
| Mech422 | inc0: nah...after my perpetual motion machines, use before definition should be cake!! :-) | 15:50 |
| inc0 | are you running cloud on your perpetum mobile? And it doesn't break? | 15:51 |
| *** ravig has joined #openstack-kolla | 15:52 | |
| pbourke_ | inc0: I think the main value in the DSL is it's easier to override/extend than using includes | 15:55 |
| pbourke_ | it gives us more control over the API than currently | 15:55 |
| inc0 | pbourke_, well, I disagree. We will be limited to options we thought of | 15:56 |
| inc0 | it won't give us more control than we have | 15:56 |
| pbourke_ | its not so much a true DSL as it is a parser | 15:56 |
| inc0 | it will generate dockerfile at the end, so it won't be more than Dockerfile API offers | 15:56 |
| pbourke_ | or a more structured way of representing what we already have | 15:56 |
| inc0 | yeah, but that should be done by refactoring of what we have | 15:57 |
| inc0 | and I think we have all the toolset we need | 15:57 |
| inc0 | we can do macro in jinja2 and such | 15:57 |
| openstackgerrit | Merged openstack/kolla: Do not use bare variables in the ceph role https://review.openstack.org/321834 | 15:57 |
| inc0 | I just don't want to lose opportunity to easily edit dockerfile itself | 15:57 |
| pbourke_ | I can see the worry of that being restictive | 15:58 |
| inc0 | most of abstraction layers are at some point | 15:58 |
| inc0 | and adding yet another one isn't good patter imho | 15:59 |
| pbourke_ | we need to enumerate the issues with the current approach and see if they can be resolved in jinja | 15:59 |
| inc0 | let's first rule out refactoring stuff without adding new layers of complexity to it | 15:59 |
| inc0 | yeah, if it can't be solved, then we should think of solutions | 16:00 |
| inc0 | not before | 16:00 |
| pbourke_ | the big one is customisations | 16:00 |
| pbourke_ | say we you want to install one more package in base, how do you do that | 16:00 |
| pbourke_ | (specific to your environment) | 16:01 |
| pbourke_ | i.e. wont be going upstream | 16:01 |
| *** mdnadeem_home has joined #openstack-kolla | 16:01 | |
| inc0 | well, you'd need to fork stuff now | 16:02 |
| inc0 | and that's good imho | 16:02 |
| inc0 | forking dockerfiles isn't bad, it's not logic per se | 16:02 |
| inc0 | which also adds to my argument, if you want to fork Dockerfile, its less likely to bite you later | 16:03 |
| inc0 | as Dockerfile API will be more stable than DSL one | 16:03 |
| inc0 | jinja2 was created to build super-complicated html pages and it does it well | 16:05 |
| inc0 | it's a tool to simplify building large documents, we don't need to reinvent it | 16:05 |
| pbourke_ | hmm maybe | 16:06 |
| inc0 | so, how about that, I'll write demo of what I'd have in mind | 16:06 |
| inc0 | and we'll confront both ideas | 16:06 |
| pbourke_ | I agree we need to be sure this can't be solved/improved with the current methods | 16:06 |
| inc0 | my point exactly | 16:07 |
| Mech422 | inc0: heh - I like that 'perpetum mobile' :-) | 16:07 |
| inc0 | well it's latin for perpetual motion machine | 16:07 |
| pbourke_ | what i had in mind originally was something like the yaml but using jinja include blocks | 16:08 |
| inc0 | yeah that's how I'd do it too | 16:08 |
| inc0 | like {% include base_packages_install_{{ install_base }}_{{ install_type }} %} | 16:09 |
| pbourke_ | yeah | 16:09 |
| inc0 | this won't work, but I'll figure out how to make it works | 16:09 |
| inc0 | and we'll have base_packages_install_centos_binary | 16:09 |
| inc0 | a file | 16:09 |
| *** ravig has quit IRC | 16:09 | |
| pbourke_ | i think one thing sdake has an issue with also, which I agree with is the proliferation of options going into build.py | 16:10 |
| inc0 | with RUN yum -i install ... | 16:10 |
| inc0 | what do you mean? | 16:10 |
| *** fragatina has quit IRC | 16:10 | |
| pbourke_ | so any time people want to customise something it ends up becoming a jinja2 variable and another arg gets added build.py | 16:11 |
| pbourke_ | not saying those couldn't be moved elsewhere though | 16:12 |
| inc0 | well, that won't be solved in DSL | 16:12 |
| inc0 | if we need additional input, then we need it | 16:12 |
| inc0 | however we'll turn it into Dockerfile at the end | 16:12 |
| *** d_code has quit IRC | 16:12 | |
| pbourke_ | well with the DSL people would supply chunks of yaml that would be easily mergable | 16:12 |
| pbourke_ | but I agree it's essentially the same as jinja includes just solved with yaml | 16:13 |
| inc0 | and now they'll supply chunks of dockerfile | 16:13 |
| inc0 | you'll have 2 languages to learn instead of 3 (jinja, dockerfile, DSL) | 16:13 |
| inc0 | and 0 parsers to support instead of 1;) | 16:13 |
| pbourke_ | the jinja would go away | 16:13 |
| inc0 | then I'm total -2 to it | 16:14 |
| pbourke_ | lol | 16:14 |
| inc0 | I don't want reimplementing jinja | 16:14 |
| inc0 | 1 backwards compatibility | 16:14 |
| inc0 | 2 we'll reimplement jinja... | 16:14 |
| inc0 | only in yamls | 16:14 |
| inc0 | I agree yamls are cute, but really? | 16:14 |
| pbourke_ | ha | 16:14 |
| pbourke_ | i dunno | 16:14 |
| inc0 | also I don't really see problem it solves | 16:15 |
| inc0 | I agree, Dockerfiles we have are ugly | 16:15 |
| *** vhosakot has quit IRC | 16:15 | |
| inc0 | but that's because we made a call of only using "if" statement from jinja | 16:15 |
| inc0 | I was opposed to this limitation, but that what we agreed in Santa Clara | 16:15 |
| inc0 | now we can move beyond that and refactor our dockerfiles using full power jinja2 offers | 16:16 |
| pbourke_ | i like the abstraction of a list of data without all the RUN && \ && curl blah blah | 16:16 |
| inc0 | but what if you have to make a RUN that wasn't considered in DSL? | 16:16 |
| inc0 | like add some option to RUN or whatever? | 16:16 |
| inc0 | pip install with proxy settings | 16:16 |
| *** vhosakot has joined #openstack-kolla | 16:17 | |
| inc0 | or private mirror of repo | 16:17 |
| inc0 | you know...all the things we didn't think of | 16:17 |
| *** matrohon has quit IRC | 16:17 | |
| inc0 | with jinja it's just adding few things to RUN in one of files | 16:17 |
| inc0 | with DLS is to reverse engineer whole thing and hope it will never change | 16:17 |
| *** thumpba has joined #openstack-kolla | 16:18 | |
| pbourke_ | one will probably tell you we've arrived on a stable API that shouldn't change for the forseeable future | 16:18 |
| *** mikelk has quit IRC | 16:18 | |
| inc0 | until it does... | 16:18 |
| inc0 | also how long it would take to get this "stable API"? | 16:18 |
| inc0 | and people like Lyncos, they already fork our dockerfiles | 16:19 |
| inc0 | to enable their own stuff in it | 16:19 |
| inc0 | and thats cool! | 16:19 |
| inc0 | we should make it easier, not harder | 16:19 |
| pbourke_ | so do oracle | 16:19 |
| inc0 | I expect many people to do it | 16:19 |
| inc0 | snowflakes are there in a plenty | 16:19 |
| pbourke_ | if forking is so great why do we support config merging? | 16:22 |
| inc0 | well there are things you can only do with merging | 16:23 |
| inc0 | like chaniging config per node | 16:23 |
| *** thumpba has quit IRC | 16:23 | |
| *** thumpba has joined #openstack-kolla | 16:24 | |
| inc0 | and I'm not saying it's great, I'm saying sometimes it's required | 16:24 |
| inc0 | for us more than for usual projects | 16:24 |
| pbourke_ | ah good point | 16:25 |
| inc0 | and adding layer of complexity will make it harder | 16:25 |
| *** ayoung has quit IRC | 16:25 | |
| *** tyrola has joined #openstack-kolla | 16:27 | |
| pbourke_ | oh well, heading home. see what sdake thinks :) | 16:29 |
| inc0 | sure, have a good one | 16:29 |
| *** Mr_Broken has joined #openstack-kolla | 16:29 | |
| inc0 | I'll tinker up some demo of what I have in mind | 16:29 |
| pbourke_ | cool | 16:29 |
| *** unicell1 has quit IRC | 16:30 | |
| *** dwalsh has joined #openstack-kolla | 16:32 | |
| *** daneyon_ has joined #openstack-kolla | 16:33 | |
| *** dwalsh has quit IRC | 16:36 | |
| *** dwalsh has joined #openstack-kolla | 16:37 | |
| *** daneyon_ has quit IRC | 16:37 | |
| *** d_code has joined #openstack-kolla | 16:41 | |
| *** ayoung has joined #openstack-kolla | 16:42 | |
| *** ssurana has joined #openstack-kolla | 16:42 | |
| *** mark-casey has joined #openstack-kolla | 16:47 | |
| mark-casey | hey all. how are server reboots handled in multinode? do you just run kolla-ansible reconfigure? | 16:52 |
| mark-casey | and, I guess, is this any different if only one server got rebooted | 16:52 |
| *** berendt has joined #openstack-kolla | 16:52 | |
| inc0 | mark-casey, well, you shouldn't have to run anything | 16:54 |
| inc0 | stuff should just work | 16:54 |
| inc0 | problem is with all controllers reboot as you need to rescue mariadb | 16:55 |
| inc0 | we have play for that | 16:55 |
| *** ppowell has quit IRC | 16:58 | |
| mark-casey | inc0: ok. so docker comes up and the restart=always takes care of it except galera? | 16:58 |
| inc0 | well, in theory | 16:58 |
| mark-casey | hehe, sure | 16:58 |
| inc0 | if you have aio you will run into race conditions | 16:58 |
| inc0 | but with multinode it should work | 16:58 |
| mark-casey | inc0: is that because aio is lacking the HA bits so won't retry as much if out of order? | 17:00 |
| inc0 | well, with aio nova will stand before keystone or mariadb and such | 17:00 |
| inc0 | order is random | 17:00 |
| inc0 | with multinode you still have haproxy dealing with that | 17:00 |
| inc0 | now when I think about it if keepalived migrate floating ip before haproxy stands we might have race | 17:01 |
| inc0 | but that's unlikely and it will cause subsecond downtime | 17:01 |
| mark-casey | :) | 17:01 |
| *** diogogmt has quit IRC | 17:02 | |
| *** diogogmt has joined #openstack-kolla | 17:03 | |
| *** dwalsh has quit IRC | 17:05 | |
| openstackgerrit | Mark Goddard proposed openstack/kolla: Don't always unpack image build results tuple, as it may be None https://review.openstack.org/323505 | 17:19 |
| *** dwalsh has joined #openstack-kolla | 17:20 | |
| *** ppowell has joined #openstack-kolla | 17:21 | |
| *** dwalsh has quit IRC | 17:26 | |
| *** unicell has joined #openstack-kolla | 17:26 | |
| *** d_code has quit IRC | 17:27 | |
| *** zhiwei has joined #openstack-kolla | 17:31 | |
| *** vhosakot has quit IRC | 17:34 | |
| *** zhiwei has quit IRC | 17:36 | |
| *** Serlex has quit IRC | 17:36 | |
| *** vhosakot has joined #openstack-kolla | 17:37 | |
| *** dwalsh has joined #openstack-kolla | 17:40 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Adding the password generation step to the docs. https://review.openstack.org/322362 | 17:43 |
| *** ravig has joined #openstack-kolla | 17:45 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: [doc] change Ansible version to exactly 2.0.x in quickstart. https://review.openstack.org/322361 | 17:46 |
| *** ravig has quit IRC | 17:47 | |
| *** Lyncos has joined #openstack-kolla | 17:49 | |
| *** ravig has joined #openstack-kolla | 17:54 | |
| mag009_ | inc0: i found my problem btw... the mount issue that I had | 17:57 |
| mag009_ | apparently it was caused by the lxd installed on my ubuntu server | 17:57 |
| mag009_ | after uninstalling anything to do with lxd and lxc docker work like a charm | 17:57 |
| inc0 | yeah Lyncos already told me, be a good community people and make a note in quickstart please;) | 17:58 |
| *** fragatina has joined #openstack-kolla | 18:03 | |
| *** d_code has joined #openstack-kolla | 18:04 | |
| vhosakot | yes, please add this info in quickstart | 18:05 |
| openstackgerrit | Marc-Andre Gatien proposed openstack/kolla: adding Note for ubuntu 16.04 https://review.openstack.org/323519 | 18:13 |
| mag009_ | done | 18:14 |
| *** mark-casey has quit IRC | 18:17 | |
| *** ravig has quit IRC | 18:17 | |
| openstackgerrit | Marc-Andre Gatien proposed openstack/kolla: adding Note for ubuntu 16.04 https://review.openstack.org/323519 | 18:20 |
| *** SiRiuS has quit IRC | 18:23 | |
| mliima | mag009_, add . after ) | 18:24 |
| openstackgerrit | Marc-Andre Gatien proposed openstack/kolla: adding Note for ubuntu 16.04 https://review.openstack.org/323519 | 18:25 |
| mag009_ | are you guys going to focus on reducing the size of the images ? like using alpine | 18:26 |
| mag009_ | mliima thanks! | 18:26 |
| *** ravig has joined #openstack-kolla | 18:26 | |
| *** Mr_Broken has quit IRC | 18:28 | |
| *** Mr_Broken has joined #openstack-kolla | 18:29 | |
| *** mdnadeem_home has quit IRC | 18:33 | |
| *** nilk has joined #openstack-kolla | 18:36 | |
| nilk | Hey guys | 18:36 |
| nilk | so... we've tried to deploy an OpenStack cluster with stable/mitake | 18:36 |
| nilk | but all we got is this error in production: https://bugs.launchpad.net/kolla/+bug/1587226 | 18:37 |
| openstack | Launchpad bug 1587226 in kolla "gate failed with the error publicURL endpoint for compute service not found" [Undecided,New] | 18:37 |
| nilk | two controller nodes, two network nodes, one compute node (for now) | 18:37 |
| nilk | stable/mitake | 18:37 |
| nilk | source-based | 18:37 |
| nilk | CentOS 7 | 18:37 |
| nilk | the MariaDB deployment doesn't work either since the timeouts are too short, it gets stuck in a funny state where both instances hang in a loop and write multiple MB of logfiles | 18:41 |
| nilk | we fixed that one, but no idea about the compute node issues | 18:42 |
| nilk | the endpoint simply isn't created (neither is the service) | 18:42 |
| *** SiRiuS has joined #openstack-kolla | 18:47 | |
| *** fragatina has quit IRC | 18:51 | |
| *** tyrola has quit IRC | 18:51 | |
| *** d_code has quit IRC | 18:52 | |
| *** fragatina has joined #openstack-kolla | 18:52 | |
| inc0 | hmm | 19:05 |
| inc0 | nilk, so deployment succeeded? | 19:05 |
| inc0 | I mean ansible ran correctly? | 19:05 |
| *** mdnadeem has quit IRC | 19:08 | |
| *** mdnadeem has joined #openstack-kolla | 19:09 | |
| *** ravig has quit IRC | 19:11 | |
| *** efoley has quit IRC | 19:11 | |
| *** ravig has joined #openstack-kolla | 19:15 | |
| *** daneyon_ has joined #openstack-kolla | 19:15 | |
| Daviey | nilk: Actually, i agree with the timeout... I've seen that a few times... I upped it to 120 seconds. What worked for you? | 19:17 |
| *** ravig has quit IRC | 19:18 | |
| *** daneyon_ has quit IRC | 19:20 | |
| *** ravig has joined #openstack-kolla | 19:21 | |
| *** tyrola has joined #openstack-kolla | 19:22 | |
| *** inc0 has quit IRC | 19:23 | |
| *** Mr_Broken has quit IRC | 19:29 | |
| *** Mr_Broken has joined #openstack-kolla | 19:30 | |
| *** Jeffrey4l_ has joined #openstack-kolla | 19:32 | |
| *** Jeffrey4l has quit IRC | 19:34 | |
| *** Mr_Broken has quit IRC | 19:34 | |
| *** ravig has quit IRC | 19:34 | |
| *** mdnadeem has quit IRC | 19:38 | |
| *** mdnadeem has joined #openstack-kolla | 19:39 | |
| *** dwalsh has quit IRC | 19:39 | |
| *** mdnadeem has quit IRC | 19:39 | |
| *** dwalsh has joined #openstack-kolla | 19:39 | |
| *** tyrola has quit IRC | 19:39 | |
| *** mdnadeem has joined #openstack-kolla | 19:40 | |
| *** mdnadeem has quit IRC | 19:40 | |
| *** mdnadeem has joined #openstack-kolla | 19:41 | |
| nilk | inc0: yes, it deployed, Horizon works, it's just that the endpoint did not get created | 19:42 |
| nilk | Daviey: 60s | 19:42 |
| *** ravig has joined #openstack-kolla | 19:44 | |
| *** sdake has joined #openstack-kolla | 19:45 | |
| *** sdake_ has joined #openstack-kolla | 19:49 | |
| nilk | inc0: what actually happens is that register.yml is skipped for the controller nodes | 19:49 |
| *** sdake has quit IRC | 19:51 | |
| *** inc0 has joined #openstack-kolla | 19:52 | |
| *** mdnadeem has quit IRC | 19:53 | |
| inc0 | Texas... | 19:53 |
| inc0 | storm came in and power went out | 19:54 |
| *** mdnadeem has joined #openstack-kolla | 19:54 | |
| inc0 | it's been storming every day for few days now | 19:54 |
| *** ayoung has quit IRC | 19:54 | |
| *** ayoung has joined #openstack-kolla | 19:55 | |
| nilk | inc0, weather has been a bit rough in Germany, too | 19:56 |
| inc0 | I've seen photos from Czech with hail | 19:56 |
| nilk | we got lucky, a few villages got destroyed | 19:56 |
| inc0 | we had similar hais few months back | 19:56 |
| nilk | but people have been really lucky | 19:57 |
| inc0 | nilk, which part of germany? | 19:57 |
| nilk | South | 19:57 |
| inc0 | yeah...that was crazy | 19:57 |
| inc0 | global warming all right | 19:57 |
| inc0 | nilk, you mentioned some issues you were having? | 19:57 |
| inc0 | I lost it due to well...storm | 19:58 |
| *** tyrola has joined #openstack-kolla | 19:58 | |
| nilk | http://www.tagesschau.de/multimedia/bilder/unwetter-383.html | 19:58 |
| nilk | np I mostly figured it out anyway | 19:58 |
| nilk | what happens is that it only attempts to run nova's register.yml on the compute nodes | 19:58 |
| inc0 | ok | 19:58 |
| nilk | which is pointless since inventory_hostname in groups['nova-api'] does not match | 19:59 |
| inc0 | hmm, it should do this on controller nodes | 19:59 |
| nilk | yeah that's what I thought, too | 19:59 |
| nilk | no idea why this happens but if I manually patch it, it works | 19:59 |
| nilk | the controller nodes are in nova-api, the compute node isn't | 20:00 |
| nilk | http://paste.openstack.org/show/506649/ | 20:00 |
| inc0 | which ansible are you using? | 20:01 |
| inc0 | 1.9.4 for stable/mitaka? | 20:02 |
| nilk | yes | 20:02 |
| inc0 | it sometimes does that, there is bug for it | 20:02 |
| inc0 | hold o | 20:02 |
| inc0 | n | 20:02 |
| *** tyrola has quit IRC | 20:02 | |
| inc0 | it was about ordering of hosts, bug in ansible | 20:02 |
| nilk | fun fact: it worked a few times and we have no idea why it no longer does | 20:03 |
| inc0 | well, can't find bug now, but it was apearing from time to time in neutron | 20:03 |
| inc0 | anyway move to ansible 2 should help, but we just merged ansible 2 support | 20:03 |
| nilk | we're scared of master | 20:04 |
| nilk | also we don't want to rebuild all images | 20:04 |
| inc0 | https://github.com/openstack/kolla/blob/master/ansible/site.yml#L100 try to move nova-controller above compute | 20:04 |
| inc0 | nilk, wise | 20:04 |
| inc0 | mitaka should be good, I don't advice moving to master | 20:05 |
| inc0 | just saying it should be better | 20:05 |
| nilk | I think Ansible 2 is only in master? | 20:05 |
| *** dmk0202 has joined #openstack-kolla | 20:05 | |
| inc0 | yeah | 20:05 |
| inc0 | anyway, try hack I shown you | 20:05 |
| inc0 | change order of hosts in this file | 20:06 |
| inc0 | and see if it helps | 20:06 |
| inc0 | nova-api above compute ** | 20:06 |
| nilk | it worked | 20:07 |
| nilk | thank you so much | 20:07 |
| nilk | but... WTF | 20:07 |
| inc0 | no problem | 20:07 |
| inc0 | yeah, this is one of *these* bugs | 20:07 |
| nilk | that one cost me a lot of time :P | 20:07 |
| inc0 | anyway, should be fixed after move to ansible2 | 20:07 |
| inc0 | glad I could help | 20:07 |
| inc0 | you can publish a patch with order change if you want | 20:08 |
| *** mdnadeem has quit IRC | 20:08 | |
| nilk | we undid everything we did since we believed that something we did broke it | 20:08 |
| inc0 | so others won't catch the same issue | 20:08 |
| nilk | since it worked in the beginning | 20:08 |
| nilk | nice one | 20:08 |
| inc0 | yeah I know the pain | 20:08 |
| inc0 | this is a bug in ansible itself | 20:08 |
| *** mdnadeem has joined #openstack-kolla | 20:08 | |
| inc0 | we hit that in neutron and we change the ordering there | 20:09 |
| *** SiRiuS has quit IRC | 20:09 | |
| inc0 | issue is, it kinda happen randomly | 20:09 |
| nilk | yeah I realize, I'm not even complaining | 20:09 |
| inc0 | I deployed mitaka multiple times and never seen it | 20:09 |
| nilk | just... WTF | 20:09 |
| nilk | everything else has been smooth so far, except for that MariaDB issue | 20:10 |
| inc0 | yeah, we have issues with galera from times to times | 20:10 |
| inc0 | what did you change? | 20:10 |
| inc0 | which timeout? | 20:10 |
| nilk | actually, I added pause statements after starting each container | 20:11 |
| nilk | it works in master | 20:11 |
| nilk | not in mitaka | 20:11 |
| nilk | as in, it works out of the box in master but we had to fix in Mitaka | 20:11 |
| nilk | I'm really impressed with Kolla, great job | 20:12 |
| nilk | we're using it for our new production cloud | 20:12 |
| inc0 | I'm glad to hear that:) | 20:12 |
| inc0 | wonderful, we need people deploying it so we'll test ourselves in fire so to say | 20:12 |
| nilk | docs are a bit lacking at times, I hope that I'll find time to commit a few fixes | 20:13 |
| inc0 | let us know if you run into any issues, we're happy to help | 20:13 |
| inc0 | docs are one of big priorities this release | 20:13 |
| inc0 | they're horrible now, but we're getting there | 20:13 |
| nilk | also it's easy to make mistakes that result in non-descript errors | 20:13 |
| inc0 | yeah, ansible 2 helps a bit | 20:13 |
| inc0 | but really we need better prechecks and sanity checks | 20:14 |
| Daviey | inc0: Thinking of adding support for ansible --diff|--check (dry run stuff?) | 20:14 |
| nilk | yes | 20:14 |
| nilk | it shouldn't even run with SELinux enabled or wrong Ansible or Docker version | 20:14 |
| inc0 | Daviey, well we have playbook with some check stuff, but that's a good idea | 20:14 |
| inc0 | we need lots of infrastructure like that | 20:15 |
| inc0 | for now we were focused on main features, now user experience | 20:15 |
| Daviey | inc0: Call me a scaredy cat, but i'm really worried about making a fat fingered typo and flushing all the data away :) | 20:15 |
| inc0 | I'm working on refactoring of dockerfiles as we speak;) | 20:15 |
| inc0 | Daviey, I totally get that, fortunately it'll be hard to flush data with kolla | 20:16 |
| nilk | we made a nice script which tears down the entire cluster in a few seconds | 20:16 |
| nilk | pretty cool | 20:16 |
| inc0 | as we have cleanup tools completely separate from our plays | 20:16 |
| nilk | also extremely scary | 20:16 |
| inc0 | nilk, tools/cleanup-host and tools/cleanup-containers | 20:16 |
| inc0 | these are scary scripts | 20:16 |
| inc0 | but do what you just said | 20:17 |
| nilk | yes we use those | 20:17 |
| inc0 | ansible -m script and go ahead; | 20:17 |
| sdake_ | kolla-ansible cleanup does the job too :) | 20:17 |
| nilk | we made sure that no customer data survives either | 20:17 |
| Daviey | yeah, i wrapped scp'ing them along with flushing the parition table and renaming the ceph block device | 20:17 |
| Daviey | and executing them over ssh | 20:17 |
| inc0 | Daviey, use ansible, ssh is passe;) | 20:18 |
| Daviey | sdake_: not in stable/mitaka it doesn't | 20:18 |
| inc0 | sdake_, btw, I'm working on providing alternative to DSL | 20:18 |
| Daviey | inc0: fair point :) | 20:18 |
| inc0 | I really dislike this idea, I'll publish demo of refactored dockerfiles to make use of more advanced jinja | 20:18 |
| nilk | so. fix deployed, I'll re-deploy the cluster | 20:19 |
| nilk | if it works I'll get drunk | 20:19 |
| inc0 | haha | 20:19 |
| nilk | spent three days fighting this | 20:19 |
| inc0 | drink one for us;) | 20:19 |
| Daviey | nilk: 3 days? n00b. | 20:20 |
| inc0 | god I miss European beer | 20:20 |
| sdake_ | inc0 jinja2 = not convinced | 20:20 |
| nilk | Daviey, yes, for sure | 20:20 |
| nilk | lots of auxiliary stuff unrelated to Kolla | 20:20 |
| inc0 | sdake_, well, we already have it | 20:20 |
| sdake_ | inc0 but if you can meet the numerous requreiments I have - then we can consider it :) | 20:20 |
| inc0 | what are the requirements? | 20:20 |
| Daviey | nilk: (i'm only about 3 x more than you) | 20:20 |
| inc0 | also we already use jinja2 | 20:21 |
| nilk | :P | 20:21 |
| inc0 | it's your DSL we should consider not the other way around:P | 20:21 |
| inc0 | you're one making a revolution | 20:21 |
| sdake_ | I need to be able to customize any part of any dockerfile by 1) merging one section into another 2) overridding whats there | 20:21 |
| inc0 | well, you can change dockerfiles... | 20:21 |
| sdake_ | inc0 i will make surer there is no feature loss | 20:21 |
| sdake_ | without changing dockerfiles | 20:21 |
| inc0 | well you need to change something | 20:22 |
| sdake_ | i am good with changing a user provided file | 20:22 |
| inc0 | either it's dockerfile, jinja tpl or your dsl tpl | 20:22 |
| inc0 | conditional includes, there you go.. | 20:22 |
| inc0 | I don't like DSLs, will make stuff super hard | 20:22 |
| inc0 | not to mention that we'll have bunch of code to support | 20:23 |
| sdake_ | ok first up - need a separate repository per container | 20:23 |
| inc0 | eee? why? | 20:23 |
| sdake_ | how do you do that with conditions in a sane way? | 20:23 |
| sdake_ | $$dayjob wants that | 20:23 |
| inc0 | repository per container? We specifically didn't want that | 20:23 |
| sdake_ | we need flexibility to be able to handle it | 20:24 |
| sdake_ | and we didn't want it by default, we never said we didn't want it at all | 20:24 |
| inc0 | but git clone --recursive is answer for you | 20:24 |
| *** mliima has quit IRC | 20:24 | |
| inc0 | well we either have all it kolla repo or create shitload of repos | 20:24 |
| sdake_ | inc0 have you tried my dsl implementation? | 20:25 |
| Daviey | (or, move the files to the source project) | 20:25 |
| *** mark-casey has joined #openstack-kolla | 20:25 | |
| *** jtriley has quit IRC | 20:25 | |
| inc0 | I did look at it | 20:25 |
| inc0 | and I didn't really like it tbh | 20:25 |
| inc0 | all sorts of things may go wrong | 20:25 |
| sdake_ | inc0 what would need to happen for include per repo is you would have to do a #include nova-compute-repo.txt | 20:25 |
| sdake_ | yo uwould need that for very type of includeable thing | 20:25 |
| inc0 | also I'd rather start by stating a problem, then discuss how can we solve it with what we already have | 20:25 |
| sdake_ | and it would have to do the right thing when the file is not present | 20:26 |
| sdake_ | inc0 did you run the tool that generates parsed dockerfiles? | 20:26 |
| inc0 | no, I believe it works | 20:26 |
| inc0 | I just don't like idea itself | 20:26 |
| sdake_ | its so simple and straightforward | 20:26 |
| inc0 | until it stops being simple | 20:26 |
| sdake_ | build.py is out of hand | 20:26 |
| sdake_ | even pbourke_ agrees | 20:27 |
| inc0 | people already started to create their own tailored dockerfiles | 20:27 |
| inc0 | and we're screwing them off | 20:27 |
| *** ayoung has quit IRC | 20:27 | |
| inc0 | and I'm -2 to screwing people off | 20:27 |
| sdake_ | tailored dockerfiles are not the answer | 20:27 |
| sdake_ | but lets assume someone has a tailored dockerfile | 20:27 |
| *** dmk0202 has quit IRC | 20:28 | |
| sdake_ | I'll personally write a conversion to the new technique for them - so no screw off for them | 20:28 |
| sdake_ | if there are such people | 20:28 |
| sdake_ | if you had tried the tool | 20:28 |
| inc0 | sdake_, please, let's start with problems we want to solve | 20:28 |
| sdake_ | yo uwould see its super simple | 20:28 |
| sdake_ | i told you the problem I want to solve | 20:28 |
| inc0 | multiple repos, got it | 20:28 |
| inc0 | can do with jinja | 20:28 |
| sdake_ | not just multiple repos | 20:29 |
| sdake_ | i want to be able to customize anything in the dockerfile | 20:29 |
| sdake_ | in a programmatic way | 20:29 |
| inc0 | conditional includes, can do with jinja | 20:29 |
| inc0 | dockerfiles are DSL | 20:29 |
| sdake_ | ok let me give you a scenario | 20:29 |
| inc0 | I don't like idea of DSL for DSL | 20:29 |
| sdake_ | i have custom repo | 20:29 |
| sdake_ | I have two binary rpms that must be installed for plugins | 20:29 |
| sdake_ | I want to override all of the custom repositories provided by CentOS because I have some sort of business relationship with Mirantis and want to ship their distro instead | 20:30 |
| sdake_ | can that be done in j2? | 20:31 |
| inc0 | so, first one is easy | 20:31 |
| inc0 | --include-header << dockerfile with setup of repos and install packages >> | 20:31 |
| inc0 | you can do this now | 20:31 |
| sdake_ | yes but that is per build | 20:31 |
| sdake_ | we need it per repo! | 20:31 |
| sdake_ | rather per conttainer | 20:32 |
| inc0 | second one, with refactoring I'm doing now you'll have separate file called "redhat-dockerfile" with block "setup-repo" | 20:32 |
| inc0 | all you need to do is create your own new myown-dockerfile | 20:32 |
| Daviey | How are you planning to have known good version compatability with multiple external branches? | 20:32 |
| kklimonda | inc0: I have local ubuntu mirrors, and I want to use them instead of archive.ubuntu.com - I can't do that easily right now | 20:32 |
| inc0 | with block setup_repo | 20:32 |
| inc0 | and point build.py7 to it | 20:32 |
| sdake_ | inc0 here is where we are in disagreement | 20:33 |
| sdake_ | I think build.py has too much logic | 20:33 |
| sdake_ | I think it muddles how the dockerfiles are built | 20:33 |
| inc0 | well, let's refactor build.py | 20:33 |
| sdake_ | and you are proposing addding even more logic | 20:33 |
| inc0 | I agree | 20:33 |
| inc0 | well, you want to replace this logic with new logic of a DSL | 20:33 |
| Daviey | kklimonda: I used to rewrite dns for to easily solve that. | 20:33 |
| inc0 | kklimonda, for apt packages you have this file: | 20:34 |
| nilk | w00t | 20:34 |
| nilk | it worked! | 20:34 |
| sdake_ | inc0 if you can deliver the 3 things above in some consistent way using j2, by providing all those files ahead of time, then i guess that would wfm, although I think it would be clunky | 20:34 |
| inc0 | https://github.com/openstack/kolla/blob/stable/mitaka/docker/base/sources.list | 20:34 |
| nilk | 0 to OpenStack in a few minutes | 20:34 |
| *** fragatina has quit IRC | 20:34 | |
| nilk | I really like this project :) | 20:34 |
| sdake_ | nilk kolla pretty sweeet huh :) | 20:34 |
| inc0 | sdake_, I can provide it and I disagree | 20:34 |
| inc0 | that it would be clunky | 20:34 |
| inc0 | remember that I was opposet to limit ourselves to "only using if statement in jinja" | 20:35 |
| inc0 | that was discussion we had in Santa Clara | 20:35 |
| inc0 | but we decided to only use if's so here we are | 20:35 |
| Daviey | nilk: Why did you choose to roll your own images, rather than use kollaglue's? | 20:35 |
| sdake_ | that is so we could get a basic implementation down before people started rocket scientisting | 20:35 |
| *** godleon has joined #openstack-kolla | 20:35 | |
| inc0 | now we can move ouit of this limit and write proper temnplates | 20:35 |
| sdake_ | inc0 i am by no means a inj2 template expert | 20:36 |
| inc0 | sdake_, I know, now we can refactor and include some syntactic sugars from jinja | 20:36 |
| sdake_ | but imo this problem cannot be solved by teplates | 20:36 |
| nilk | Daviey: easier to implement custom fixes that way | 20:36 |
| nilk | also, security considerations | 20:36 |
| inc0 | it can by templates AND refactor of build | 20:36 |
| Daviey | nilk: Such as? I'm genuinely interested | 20:36 |
| inc0 | and we should do both, I agree | 20:36 |
| kklimonda | inc0: good point, although I also had to provide my own gpg keys. | 20:36 |
| sdake_ | inc0 also the dsl parser produces more optimized output:) | 20:37 |
| nilk | Daviey: say we want to backport something from OpenStack master, or fix a bug related to our setup | 20:37 |
| inc0 | kklimonda, --include-header << path to file with dockerfile code that includes them >> | 20:37 |
| sdake_ | but you think docker.inc would sort that problem out on their own | 20:37 |
| inc0 | during build | 20:37 |
| nilk | this way, we know that the build process works and we can just rebuild it from Git | 20:37 |
| sdake_ | inc0 here is what I think you are proposing | 20:37 |
| sdake_ | in every dockerfile | 20:37 |
| sdake_ | follow a structure like | 20:37 |
| kklimonda | inc0: yes, that's doable but I just don't think it's the right way to extend templates - I consider it a hack tbh | 20:38 |
| sdake_ | {include my own bins} | 20:38 |
| sdake_ | {include external bins} | 20:38 |
| sdake_ | {include my own srcs} | 20:38 |
| sdake_ | {include external srcs} | 20:38 |
| inc0 | kklimonda, it is, and extending dockerfiles is what we're discussing | 20:38 |
| Daviey | kklimonda: Why would you need to do your own gpg? | 20:38 |
| sdake_ | {include my own repos} | 20:38 |
| sdake_ | {include external repos} | 20:38 |
| inc0 | sdake_, no, that's not what I propose | 20:38 |
| inc0 | let me polish idea a bit but essentially what I want | 20:38 |
| inc0 | is | 20:38 |
| kklimonda | Daviey: I have a custom repository with some additional packages | 20:38 |
| nilk | Daviey: many serious deployments end up carrying their own patches on top of stable OpenStack and this way, we know that it won't hurt us in the future | 20:39 |
| inc0 | in main Dockerfile we have block {% block include_repos %} | 20:39 |
| nilk | also, the fewer external dependencies, the better | 20:39 |
| Daviey | nilk: perfect, thanks | 20:39 |
| inc0 | then we have ubuntu-dockerfile.j2 which has {% block include_repos %}<<ubuntu logic with repos | 20:39 |
| inc0 | centos will be centos-dockerfile.j2 | 20:40 |
| inc0 | and you can add myown-dockerfile | 20:40 |
| inc0 | wich, if present will be used | 20:40 |
| inc0 | and you can override blocks on your own | 20:40 |
| sdake_ | what about per container repos? | 20:40 |
| sdake_ | I totallly get how you can include one file in another in jinja2 ;) | 20:40 |
| inc0 | it's not just file inclusion | 20:40 |
| Daviey | nilk: Are you using insecure docker registry? Did you HA it all? | 20:40 |
| inc0 | ehh, I'll show it by example | 20:41 |
| inc0 | just give me a few | 20:41 |
| nilk | Daviey: secured registry | 20:41 |
| nilk | HA is unnecessary since it's not production critical | 20:42 |
| sdake_ | inc0 to me, what wins is what offers the easiest user interaction | 20:43 |
| Daviey | nilk: you aren't worried about losing images? | 20:43 |
| inc0 | sdake_, devil is in the details as always | 20:43 |
| inc0 | it will work until we find some edge case and need to code around it | 20:43 |
| inc0 | this is why build.py is big and ugly now | 20:43 |
| inc0 | all the edge cases | 20:43 |
| sdake_ | yes - dsl solves this :) | 20:44 |
| inc0 | and you want to add another layer of complexity above it | 20:44 |
| inc0 | no, it will add layer of complexity | 20:44 |
| sdake_ | i'd like to make build.py just build raw dockerfiles with dependency management | 20:44 |
| kklimonda | Daviey: is there a reason to keep images around? rebuilding them is pretty fast | 20:44 |
| sdake_ | I'd like this eparse.py code to produce the raw dockerfiles | 20:44 |
| *** dmk0202 has joined #openstack-kolla | 20:44 | |
| sdake_ | and yes I'd like a cli on top that integrates these two things togetheer | 20:44 |
| sdake_ | I'd like to eject all the speciial snowflakes from build.py | 20:45 |
| inc0 | we can solve it with making build.py a better tool | 20:45 |
| inc0 | and optimize jinja2 | 20:45 |
| nilk | Daviey: backups for sure, but apart from that, no worries | 20:45 |
| inc0 | sdake_, agree...this is refactoring | 20:45 |
| sdake_ | i haven't passed judgement on jinja2 | 20:45 |
| inc0 | instead of reinventing wheel let's try to make ones we have more round | 20:45 |
| sdake_ | other then to think I thought it couldn't be done | 20:45 |
| sdake_ | again if it works better for the operator, thats a win in my booko | 20:45 |
| inc0 | well, more things can be done than you think | 20:46 |
| sdake_ | like I said not a jinj2 rocket scientist | 20:46 |
| inc0 | operators likes to have insight into what's happening | 20:46 |
| inc0 | obscuring it by DSL is not that | 20:46 |
| inc0 | dockerfiles are little more than shell scripts | 20:46 |
| sdake_ | the buld process produces the exact dockerfiles needed for the build | 20:46 |
| inc0 | all commands are there and it's visible | 20:46 |
| sdake_ | that doesn't obscure anything | 20:46 |
| inc0 | well until you want to change something | 20:46 |
| Daviey | kklimonda: I want to have a consistent infra'. Ie, know that my cloud has all the same version of the image | 20:46 |
| sdake_ | inc0 right - end users should never have to change something in a dockerfile! | 20:47 |
| *** wmiller has quit IRC | 20:47 | |
| inc0 | ofc they should, they could and they will | 20:47 |
| inc0 | we are deployment tool | 20:47 |
| inc0 | changing repo is an example | 20:48 |
| sdake_ | no, dockerfiles are what kolla provides to buid limages | 20:48 |
| sdake_ | user provides their own customization logic | 20:48 |
| inc0 | only one example among 100s | 20:48 |
| sdake_ | not logic data | 20:48 |
| inc0 | ehh, trust me we'll have snowflakes deployed by kolla | 20:49 |
| inc0 | and you can't code for every single one of them | 20:49 |
| inc0 | especially if you want to keep k8s and such working | 20:49 |
| kklimonda | Daviey: yes, but in that case you'd probably want to use frozen repository and frozen base image anyway? otherwise you can't be sure that image you are building to patch a service will have all the same dependencies anyway | 20:49 |
| inc0 | for me it's better to keep API close to source of normal dockerfiles | 20:49 |
| inc0 | and allow people to change it easily withpout us limiting them by features of DSL | 20:50 |
| sdake_ | inc0 i'd like to measure success on objective use cases | 20:50 |
| inc0 | and time to get them in place | 20:50 |
| inc0 | and stability of resulting code | 20:51 |
| sdake_ | some ide athat someone will come alog and change the dockerfile is not objective, I can't see it being done if we do the customization job correctly | 20:51 |
| sdake_ | if we don't do the customization job correctly, then yes of course the dockerfiles will have to change | 20:51 |
| sdake_ | this is what led to the dsl development in the first place | 20:51 |
| inc0 | so we should strive to accomodate as much use cases as possible without destabilize what we have | 20:51 |
| inc0 | and that's what you're proposing | 20:51 |
| inc0 | destabilization | 20:51 |
| Daviey | kklimonda: agreed | 20:51 |
| sdake_ | inc0 i'm not sure how you jumped to destabilization from new code ;) | 20:51 |
| inc0 | well you want to chagne architecture of what we have | 20:52 |
| inc0 | that's always destabilization | 20:52 |
| sdake_ | two 6 week windows under which to fix bugs | 20:52 |
| inc0 | also I care about upgrades with as much backward compatibility as possible | 20:52 |
| sdake_ | its the same thing we went through when we ejected the symlinks | 20:52 |
| * kklimonda wishes that docker tags other than :latest were immutable.. | 20:52 | |
| sdake_ | fwiw I had thought a dsl would work at that time | 20:52 |
| inc0 | and backward compatibility means if people used something in mitaka it should work in liberty | 20:52 |
| inc0 | newton** | 20:53 |
| sdake_ | but I didn't think we could technically pull it off correctly because we had too many unknowns in our dockerfiles | 20:53 |
| sdake_ | now all the unknowns are knowns | 20:53 |
| inc0 | by newton people will have cicd systems based on build.py | 20:53 |
| inc0 | which you'll break | 20:53 |
| inc0 | so before we'll break stuff like that | 20:53 |
| inc0 | let's try to solve problems without breaking them | 20:53 |
| sdake_ | you jsut said above the fundamental structure of bulild.py is unsound | 20:54 |
| inc0 | which I'm certain we can, just lets have a discussion not jump to conclusions | 20:54 |
| inc0 | I agree, refactor it | 20:54 |
| sdake_ | refactoring a bad design leads to another bad design | 20:54 |
| inc0 | we need to break build.py into more managable chunks of code | 20:54 |
| sdake_ | the breaking of build.py apart could potentiallly fix it | 20:54 |
| inc0 | but it's not bad design, it evolved from "Another simple script" | 20:55 |
| sdake_ | however jinja2 and the logic are all mixed up in the build tool | 20:55 |
| inc0 | and you want to replace it with "Another simple script" | 20:55 |
| sdake_ | i want to keep most of build.py instact | 20:55 |
| sdake_ | minus the jinja2 parts | 20:55 |
| inc0 | build.py is baded on jinja | 20:55 |
| inc0 | that's why it's in python | 20:55 |
| sdake_ | ok instead of circular arguments | 20:56 |
| inc0 | you want to remove a year of work we put into it | 20:56 |
| sdake_ | that is sunk cost thinking! | 20:56 |
| inc0 | no, I'm trying to make you realize that PoCs are nice | 20:56 |
| inc0 | but reality often breaks stuff | 20:56 |
| sdake_ | avoidance of sunk cost thinking is what got us to where we are today - which is from dead to emergent | 20:56 |
| inc0 | tell me that, how do you add proxy setting to your dsl? | 20:57 |
| sdake_ | i am totally willing to admit this dsl coudl have bugs | 20:57 |
| sdake_ | but we have an etire 6 months to fix it | 20:57 |
| inc0 | it's not about bugs, it's about edge use cases | 20:57 |
| sdake_ | inc0 can you leave that question in the review | 20:57 |
| inc0 | we will spend a lot of time to arrive to feature parity we have now | 20:57 |
| sdake_ | yes plus a whole slew of new features with a radically simpler implementation | 20:58 |
| inc0 | and then we'll have another bunch of code that's hard to maintain | 20:58 |
| inc0 | I prefer starting from "let's try to refactor what we have" than "let's throw it away and start over" | 20:58 |
| sdake_ | inc0 figuring out what *I* want customized is easy | 20:58 |
| sdake_ | its anything with a top level DSL tag in my dsl documentation | 20:58 |
| *** prithiv has joined #openstack-kolla | 20:59 | |
| inc0 | ehh, let me provide counter proposal | 20:59 |
| sdake_ | i want the customizations done once (on disk) but applied to ALL dockerfiles during build | 20:59 |
| inc0 | also please let's have a spec for this one | 20:59 |
| *** dwalsh has quit IRC | 20:59 | |
| inc0 | this one requires a spec, too much controversy | 20:59 |
| inc0 | and start a spec with stating a problem you're solving | 20:59 |
| sdake_ | inc0 your the only one controversing but if you want a spec you got it :) | 20:59 |
| inc0 | including all the use cases you want to solve | 20:59 |
| inc0 | well, I am controversing, that's my job as "guardian of the gate" right? | 21:00 |
| sdake_ | i wasn't making insult | 21:00 |
| inc0 | so yes, please spec it up, formalize a problems and we will discuss how to solve them best | 21:00 |
| sdake_ | i was pointing out your the only one not on board out of the cores I've spoken with :) | 21:00 |
| inc0 | I didn't feel insulted | 21:00 |
| inc0 | well, sorry | 21:01 |
| inc0 | :P | 21:01 |
| sdake_ | so a spec vote will be a waste of time | 21:01 |
| inc0 | but I'm a stubborn person | 21:01 |
| *** pbourke-home has joined #openstack-kolla | 21:01 | |
| sdake_ | but whatever - wfm :) | 21:01 |
| inc0 | I don't want a vote, I want problem statement | 21:01 |
| inc0 | and discussion before jumping to conclusions | 21:01 |
| sdake_ | inc0 the conclusion of a spec is a vote | 21:01 |
| sdake_ | inc0 i am just jumping to hte fact there will be a vote and it will pass | 21:02 |
| sdake_ | hence waste of time | 21:02 |
| sdake_ | but if yo uwant further docs, I am happy to provide ;) | 21:02 |
| inc0 | cool, in the menatime I'll work out some demo of jinja2 refactor | 21:03 |
| *** thumpba has quit IRC | 21:03 | |
| inc0 | and an idea how to make it super extensible with allowance of user-provided stuff | 21:03 |
| kklimonda | if it were possible to cover all usecases you're designing DSL for, just by making better use of jinja2 templates, then not adding additional layer of complexity is a goal worth pursuing imo. | 21:03 |
| sdake_ | cool i've already proven a dsl can work for this use case in princple | 21:03 |
| inc0 | kklimonda, +2 to that | 21:04 |
| inc0 | dsl can work, it's cost of making it working and making old stuff stop working that I don't want | 21:04 |
| inc0 | people already build around our API | 21:05 |
| inc0 | and API shouldn't change that much | 21:05 |
| inc0 | and build.py is our API | 21:05 |
| sdake_ | i had not tried jinja2 because I didn't think it offered programmatic manipulation | 21:05 |
| sdake_ | but i guess that can be done in hacky fashion which is aok for a ci/cd gate | 21:06 |
| sdake_ | which is typically full of hacking | 21:06 |
| sdake_ | inc0 you recognize if I had not written prototype, I could not write spec :) | 21:07 |
| inc0 | no, it can be done with pretty elegant way | 21:07 |
| openstackgerrit | Christian Berendt proposed openstack/kolla: Add passwords parameter to LONG_OPTS in kolla-ansible script https://review.openstack.org/323585 | 21:07 |
| inc0 | prototypes are cool | 21:07 |
| sdake_ | in that vein, can you rpovide a prototype of base, openstack-base, mariadb, and neutron | 21:08 |
| sdake_ | this is what I intend to provide - we can compare the implementations | 21:08 |
| inc0 | working on it | 21:09 |
| rhallisey | sdake_, so I'm not sure about dsl yet either. Since I think we can do plugins with j2 | 21:10 |
| rhallisey | but we can have that discussion tmr. I'm on my way out | 21:10 |
| sdake_ | one thing I really hate in build.py very much, is it includes repository names | 21:10 |
| sdake_ | that is just plain wrong | 21:10 |
| sdake_ | inc0 also yml has an inclusion thing - called anchor blocks | 21:11 |
| *** rhallisey has quit IRC | 21:11 | |
| sdake_ | inc0 acnhor blocks are one way to solve the include header and footer | 21:12 |
| sdake_ | the beuty of the dsl is it includes only what is necessary and nothing else | 21:12 |
| inc0 | until it starts being necessary and then we need to extend dsl | 21:14 |
| inc0 | so we extend dsl and it becomes uglier in time | 21:14 |
| inc0 | that's what happend with build.py | 21:14 |
| *** ppowell has quit IRC | 21:14 | |
| sdake_ | this is beuty of dsl, its in yaml it can easily be converted from one format to another with a few python calls | 21:15 |
| sdake_ | vs jinja2 which requires significant human hand holding | 21:15 |
| sdake_ | the main thing I am after is that the base dockerfiles *we* provide in the kolla project are completely unmodified | 21:16 |
| sdake_ | (I get they need to be modified to support the work you speak of) | 21:16 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Refactor of base Dockerfile to allow custom repo https://review.openstack.org/323589 | 21:18 |
| inc0 | just to demo up what jinja2 can do | 21:19 |
| inc0 | we can break this up even more, as much as we want | 21:19 |
| inc0 | working on openstack-base with all it's long installs | 21:20 |
| sdake_ | inc0 the immediate problem I see with your implementation is twofold - 1) extending build.py 2) we will end up with a custom file for every type - how to handle that for every container? | 21:21 |
| sdake_ | oh and 3 ) all the infroatmion is now spread out throug hthe system instead of in one place | 21:21 |
| sdake_ | there is logic for processing the jinja2 files, logic to including inside jinja2 in separate locations | 21:22 |
| sdake_ | but don't just cusstomize repositories, customize *everything* | 21:22 |
| inc0 | extending build.py won't be an issue if we refactor it | 21:22 |
| inc0 | which we should do | 21:22 |
| inc0 | we can customize just repos if we want to, by adding more blocks | 21:23 |
| *** ayoung has joined #openstack-kolla | 21:23 | |
| inc0 | this is effect of 2hrs of work, bear with me, it will take more to refactor all the stuff | 21:24 |
| sdake_ | i want customization of everything | 21:25 |
| sdake_ | not just repos | 21:25 |
| sdake_ | and i want to be able to delete stuff that is already there and stock replace it | 21:25 |
| sdake_ | or altneratively merge with what is therre | 21:25 |
| sdake_ | these are my objectives for the dsl | 21:25 |
| inc0 | so, I can do that | 21:26 |
| inc0 | you can't | 21:26 |
| inc0 | because | 21:26 |
| inc0 | how will you add custom option to apt-get install run? | 21:26 |
| sdake_ | not custom options, custom packages | 21:27 |
| inc0 | like for example proxy or sth | 21:27 |
| inc0 | no, I want custom option | 21:27 |
| inc0 | impossible with your dsl | 21:27 |
| inc0 | easy with my macro | 21:27 |
| sdake_ | not impossible would require another key | 21:27 |
| sdake_ | i dont see the use case for that though | 21:27 |
| inc0 | self.output.append('RUN yum -y install %s \\' % value) << from eparse | 21:28 |
| sdake_ | why would you possibly want to customize what is passed to apt or yum? | 21:28 |
| inc0 | sdake_, proxy config | 21:28 |
| inc0 | -o HttpProxy::whatever | 21:28 |
| sdake_ | inc0 could you expand | 21:28 |
| sdake_ | can you provide an example where that is used today? | 21:28 |
| inc0 | instead of apt-get -y install I want apt-get -y -o HttpProxy::whatever install | 21:28 |
| inc0 | in custom dockerfiles people might have | 21:29 |
| sdake_ | ok, where is that used in current dockerfiles? | 21:29 |
| sdake_ | isnt' include header and footer for proxies? | 21:29 |
| kklimonda | no | 21:29 |
| inc0 | you were one sayng "customize everything" | 21:29 |
| kklimonda | I mean, there is also http_proxy and https_proxy env variables login in build.py files | 21:29 |
| sean-k-mooney | include header and fuooter does work for proxies | 21:29 |
| kklimonda | logic* | 21:29 |
| sdake_ | inc0 we are not amking progress - I will write spec to unblock us | 21:29 |
| inc0 | please do | 21:30 |
| inc0 | and make sure to specify problems you're trying to solve | 21:30 |
| inc0 | because I struggle with this part | 21:30 |
| sdake_ | of course | 21:30 |
| sdake_ | what I reallyt need right now is cusstom repositories | 21:30 |
| inc0 | that's one is doable | 21:30 |
| sdake_ | what I really want is for the pain in build..py to go away | 21:31 |
| sdake_ | the structure of build.py is not ideal | 21:31 |
| inc0 | that's one is called refactoring | 21:31 |
| sdake_ | it does dep management, variable replacemnet, now macro insertions, etc | 21:31 |
| sdake_ | and it communicates with docker as well ;-) | 21:31 |
| inc0 | therefore we should split it into modules | 21:31 |
| inc0 | for jinja2 parsing, docker communication | 21:32 |
| inc0 | refactoring in a word | 21:32 |
| inc0 | refactoring != write new prettier wheel | 21:32 |
| sdake_ | i am pretty sure I know what refactoring is | 21:32 |
| *** ravig has quit IRC | 21:32 | |
| sdake_ | i want a bunch of stuff customizable, and a bunch of stuff overrideable | 21:33 |
| sdake_ | (the same things) | 21:33 |
| sdake_ | these are | 21:33 |
| sdake_ | binaries installed, pip installed, repos installed | 21:33 |
| sdake_ | i would like to keep c&p to a minimum | 21:34 |
| sdake_ | run the eparse code | 21:34 |
| sdake_ | look at the output | 21:34 |
| sdake_ | it wlil take ou 5 minutes | 21:34 |
| sean-k-mooney | one thing that has been raised in the past is the ablity to build non openstack project from source to allow custimisablity | 21:34 |
| sean-k-mooney | e.g. openvswitch to allow dpdk or ovn support | 21:35 |
| inc0 | sean-k-mooney, that's a different discussion to have.. | 21:35 |
| sdake_ | sean-k-mooney actually my wife had suggested I take this dsl thing to a wider audience but I'd like to limit to kolla for the moment | 21:35 |
| dmsimard | I haven't been very involved in kolla but creating and maintaining a new DSL and parser exclusive to Kolla docker files sounds nuts, this sounds like a problem that other people should have had and solved before Kolla | 21:35 |
| sean-k-mooney | inc0: my point was that makeing the docker files more overriable would allow for that without requireing the code to be in tree | 21:35 |
| wirehead_ | Well, I just showed one of the Helm developers Kolla-Kubernetes today. :D | 21:35 |
| inc0 | sean-k-mooney, yeah, I have idea how to do it;) | 21:36 |
| inc0 | dmsimard, +2 to that | 21:36 |
| inc0 | sean-k-mooney, https://review.openstack.org/#/c/323589/ | 21:36 |
| patchbot | inc0: patch 323589 - kolla - Refactor of base Dockerfile to allow custom repo | 21:36 |
| inc0 | idea is - we'll add bunch of {% block %} in our code | 21:36 |
| sean-k-mooney | dmsimard: this is one example https://github.com/avirshup/DockerMake | 21:36 |
| inc0 | and you'll be able to put custom stuff in each | 21:37 |
| inc0 | including building and such | 21:37 |
| sdake_ | inc0 what about custom stuff in the neutron container? | 21:37 |
| wirehead_ | I think there are some big concerns that are fairly specific to Kolla. c.f. the majority of software bundlers are OK with picking one distro, but picking a single distro to run OpenStack on would be a disaster. Thus, we’re always going to need some artisinal handmade software. | 21:37 |
| inc0 | we can add {% block neutron_server_post_install %} | 21:37 |
| inc0 | and then you have custom dockerfile that injects stuff there | 21:38 |
| inc0 | easy and clean | 21:38 |
| sean-k-mooney | are people familar with devstack plugin model? | 21:38 |
| dmsimard | Surely there has to be a solution outside of Kolla, I wish I was more familiar with both docker and kolla but I'm not. You're the experts and I'm sure you've looked at what was out there like docker-compose and whatever. It just strikes me as odd that this would be a problem just for Kolla. | 21:38 |
| sean-k-mooney | they have a number of phase that allow externel plugins to execute at different stages in the install. | 21:38 |
| *** dmk0202 has quit IRC | 21:38 | |
| inc0 | dmsimard, imho we already have an answer - jinja2 | 21:38 |
| dmsimard | surely it has it's flaws, but it's certainly more approachable than a custom dsl that would higher the barrier to entry imo | 21:39 |
| sean-k-mooney | inc0: if you add entry points a block and define different phases such as (pre-repo-setup, post-repos-setup ...) it would allow costimisation as you suggest | 21:40 |
| inc0 | sean-k-mooney, jinja2 allows that in a better way imho | 21:40 |
| dmsimard | sean-k-mooney: so docker make is essentially a dsl :p | 21:40 |
| sdake_ | inc0 in closing read this: https://review.openstack.org/#/c/322345/2/doc/elemental-guide.rst | 21:40 |
| patchbot | sdake_: patch 322345 - kolla - Documentation for elemental DSL | 21:40 |
| inc0 | in a way that people will be able to create whole projects consuming our jinja2 dockerfiles | 21:40 |
| inc0 | dmsimard, yeah, dockerfile is a dsl, jinja2 is a dsl...and when we create dsl for dsl I become suspicious | 21:41 |
| inc0 | and we're thinking about dsl for dsl for dsl | 21:41 |
| sean-k-mooney | dmsimard: yes my point there was you are right there are DSLs for this already so rather then invent an new one for kolla we could extend an existing one. | 21:41 |
| sean-k-mooney | dmsimard: if that was a better path then jinja2 | 21:41 |
| dmsimard | sean-k-mooney: ah, yes, agreed. Creating a new one is probably not the right approach. | 21:41 |
| wirehead_ | I repeat: Most of the world of software doesn’t need multi-distro. | 21:42 |
| inc0 | sean-k-mooney, we already have jinja2 so we don't include new stuff | 21:42 |
| inc0 | wirehead_, but there is not a single distro they need | 21:42 |
| sdake_ | wirehead_ if your tlaking about openstack, you do need multidistro | 21:42 |
| sdake_ | wirehead_ that topic isn't open for debate ;) | 21:42 |
| wirehead_ | Yep. | 21:42 |
| inc0 | but I agree, there is no value in having multidistro coexist in same deployment | 21:42 |
| inc0 | we won't support that | 21:43 |
| wirehead_ | OpenStack needs multi-distro. The rest of the world, largely, doesn’t. With Docker, doubly so. | 21:43 |
| sdake_ | docker doesns't solve world hunger ;) | 21:43 |
| sdake_ | dmsimard on your point about more complexity, tell me how this is complex: | 21:43 |
| sean-k-mooney | inc0: oh im not opposed to just just more of jinja2 i previously proposed using the macros for this https://review.openstack.org/#/c/275932/1/docker/openvswitch/openvswitch-base/Dockerfile.j2 | 21:43 |
| patchbot | sean-k-mooney: patch 275932 - kolla - [WIP] Support installing openvswitch from source (ABANDONED) | 21:43 |
| sdake_ | https://review.openstack.org/#/c/321468/ | 21:44 |
| patchbot | sdake_: patch 321468 - kolla - Implements elemental.yml DSL for base | 21:44 |
| wirehead_ | Ergo, since we need multi-distro and the majority of other container infras aren’t needing it, we’re kinda stuck in Kolla-land writing our own tools. | 21:44 |
| inc0 | sean-k-mooney, yeah, that'd fit nicely to my idea | 21:44 |
| wirehead_ | I’m wondering if it’s eventually going to be simpler to not write a DSL that builds Dockerfiles but a new DSL that generates container images. | 21:45 |
| sdake_ | dmsimard that link was for you ^^ | 21:45 |
| sdake_ | wirehead_ sounds like an optimization | 21:45 |
| sdake_ | but my take would be no its not going to be simpler ;) | 21:46 |
| *** ravig has joined #openstack-kolla | 21:46 | |
| wirehead_ | Heh. | 21:46 |
| sean-k-mooney | inc0: i keep coming back to devstack but if we started to use more of jinja2 we could have "functions_common" with common macros | 21:47 |
| *** ravig has quit IRC | 21:48 | |
| inc0 | yup | 21:48 |
| inc0 | and easily add new ones | 21:48 |
| inc0 | customized | 21:48 |
| sean-k-mooney | inc0: e.g. install_package enable_repo ... | 21:48 |
| inc0 | in fact there is install package macro in my review;) | 21:48 |
| sean-k-mooney | yep i see that :) | 21:49 |
| inc0 | this is only todays work, so more will come | 21:49 |
| sean-k-mooney | out of interest and for contex why was the decision to only use if statements in jinja2 originally made? | 21:49 |
| sean-k-mooney | keep docker files as close to raw docker as posible? | 21:50 |
| *** fragatina has joined #openstack-kolla | 21:52 | |
| *** ravig has joined #openstack-kolla | 21:52 | |
| inc0 | sean-k-mooney, issue was "it's new language, operators doesn't know it" | 21:58 |
| inc0 | let's keep it simple | 21:59 |
| mark-casey | wirehead_: what was the feedback from the Helm dev? | 21:59 |
| sean-k-mooney | inc0: makes sense but im also sure docker was failly new to them too :) | 21:59 |
| inc0 | sean-k-mooney, yeah, now we can move forward with advanced jinja | 22:00 |
| sean-k-mooney | inc0: jinja2 is certenly a good options in my view. the only factor that i can see as appealing to a dsl is if we decople it for docker files entirely | 22:01 |
| sean-k-mooney | a dsl would allow building rkt containers for example if a parser for the dsl was written | 22:02 |
| inc0 | yeah, but dockerfiles, simple as they are, doesn't limit you | 22:02 |
| inc0 | dsl might if we fail to include something | 22:02 |
| inc0 | which I'm sure we will | 22:03 |
| sean-k-mooney | yeah, im not sure kolla is at a point where extending its contaiers to other container enginges is a need that must be addressed now | 22:03 |
| sean-k-mooney | espically since you can execute docker images in rkt | 22:04 |
| inc0 | open container format will solve that | 22:05 |
| sean-k-mooney | yes that is true | 22:05 |
| sean-k-mooney | in any case i look forward to seeing a spec on this topic | 22:06 |
| *** ravig has quit IRC | 22:07 | |
| *** nilk has quit IRC | 22:09 | |
| *** sdake has joined #openstack-kolla | 22:10 | |
| *** sean-k-mooney has quit IRC | 22:12 | |
| *** sdake_ has quit IRC | 22:12 | |
| wirehead_ | mark-casey: I didn’t get any yet. | 22:16 |
| wirehead_ | Like, with me as a proxy for Helm… Helm doesn’t need to deal with building containers, but it does need to deal with orchestrating the deployment thereof, but only in the Kubernetes context. | 22:17 |
| *** JoseMello has quit IRC | 22:19 | |
| wirehead_ | They had the same discussions that Kolla did with respect to startup order and went strictly towards the direction that software that requires complex bootstrapping is not going to fit into the world of Helm and that processes should detect if they are able to continue starting up and either wait-loop or crash-with-docker-restarting-it. | 22:19 |
| wirehead_ | Helm does things on a more grandular level. c.f. Nova would be a Chart and would be self-contained. | 22:20 |
| *** pbourke-home has quit IRC | 22:21 | |
| *** vhosakot has quit IRC | 22:22 | |
| *** vhosakot has joined #openstack-kolla | 22:23 | |
| wirehead_ | The Kolla way of templating the Dockerfile is weird, but I honestly can’t see many alternatives given the unique set of constraints placed on Kolla. | 22:23 |
| *** vhosakot has quit IRC | 22:29 | |
| *** ravig has joined #openstack-kolla | 22:33 | |
| *** ravig has quit IRC | 22:36 | |
| openstackgerrit | Steven Dake proposed openstack/kolla: Introduction of a new DSL: Elemental https://review.openstack.org/323612 | 22:37 |
| sdake | inc0 there ya go | 22:37 |
| inc0 | kk will read through it and we can discuss it on meeting tomorrow | 22:38 |
| sdake | instead of using the meeting time for that lets use thee mailing list pls | 22:44 |
| sdake | that way there is a permanent record and people can think about what they are thinking :) | 22:44 |
| sdake | thinking/writing | 22:44 |
| sdake | inc0 also i dind't make up the reuqirements out of thin air, they are legitimately what I htink we need as far as customization goes | 22:47 |
| inc0 | sdake, ok, wfm | 22:48 |
| inc0 | in fact right now I'm working on specifying your own sources list | 22:48 |
| sdake | my own sources list? | 22:48 |
| sdake | dont grok | 22:49 |
| inc0 | apt-get uses sources.list file | 22:49 |
| inc0 | for repos | 22:49 |
| inc0 | and that won't be trivial with dsl | 22:49 |
| inc0 | as it's well...a file | 22:49 |
| inc0 | https://github.com/openstack/kolla/blob/master/docker/base/sources.list | 22:49 |
| sdake | it is trivial with a dsl | 22:49 |
| sdake | i implemented repo files and uris for centos | 22:49 |
| sdake | if it can be done for centos it can be done for ubuntu | 22:49 |
| inc0 | well, it's even more trivial with jinja;) | 22:50 |
| sdake | the challenge for people's attention ;) | 22:50 |
| sdake | you can make uris and repo files work in one include in jinja2? | 22:51 |
| inc0 | yup | 22:51 |
| sdake | i think if we take the approach you recommend to its extreme conclusion at present we would end up with 316 files on teh disk | 22:51 |
| sdake | at minimum | 22:51 |
| sdake | to customize everything | 22:51 |
| inc0 | nope | 22:51 |
| inc0 | in fact | 22:51 |
| sdake | expalin that to me because that is where i am at big odds with jinja2 | 22:52 |
| inc0 | it will take at most as much files as we have containers | 22:52 |
| inc0 | because we don't only do includes but also block syntax | 22:52 |
| sdake | 119 | 22:52 |
| inc0 | well if you want to have one yaml per multiple containers, that will be ugly as fuck | 22:53 |
| inc0 | I mean, still doable with jinja | 22:53 |
| sdake | i had it all in one file | 22:53 |
| inc0 | but ugly as fuck | 22:53 |
| sdake | but sbezverk wanted it per container | 22:53 |
| inc0 | and he's right | 22:53 |
| sdake | so I split the different and made it per container set | 22:53 |
| sdake | which is 39 | 22:53 |
| inc0 | bottom line, we can do whatever | 22:53 |
| sdake | but my technique still allows full customization :) | 22:53 |
| sdake | of all 119 containers | 22:54 |
| sdake | so anyway back on topic | 22:54 |
| sdake | how do you have a repo file for neutron and then for say neutron_lbaas_agent? | 22:54 |
| inc0 | you want different base images for both? | 22:55 |
| sdake | different repo file syes | 22:55 |
| inc0 | since repo file is included in base you want different base image | 22:56 |
| sdake | right | 22:56 |
| sdake | i want to override the base image | 22:56 |
| inc0 | or change repo in fly which is ugly as...you know | 22:56 |
| sdake | and throw out its repo files :) | 22:56 |
| inc0 | just for one neutron container? | 22:56 |
| sdake | in my case, for the whole thing | 22:56 |
| sdake | i was using neutron as an example | 22:57 |
| inc0 | ok, so you want custom sources.list right? | 22:57 |
| sdake | for all 119 dockerfiles, I need custom repositories | 22:57 |
| sdake | that is one of 12 things i want :) | 22:57 |
| inc0 | hold on, example coming right up | 22:57 |
| *** ravig has joined #openstack-kolla | 22:57 | |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Refactor of base Dockerfile to allow custom repo https://review.openstack.org/323589 | 22:59 |
| sdake | inc0 just to be clear ,for all of the 119 dockerfiles, I want *different* custom repositories | 22:59 |
| inc0 | which means you want custom repo for base image | 22:59 |
| sdake | i want to build the world | 22:59 |
| sdake | not build nova | 22:59 |
| sdake | build neutron | 22:59 |
| sdake | etc | 22:59 |
| sdake | i dont want to rebuild the base each itme | 23:00 |
| sdake | too pokey | 23:00 |
| inc0 | because thats where we setup repos | 23:00 |
| sdake | i recognize that | 23:00 |
| sdake | lets just assume we didn't do that | 23:00 |
| sdake | or we did it, and what came later overrode what was already there | 23:00 |
| inc0 | look at my example, this is how you can change repos for everything | 23:00 |
| sdake | inc0 yes but I need it per dockerfile | 23:01 |
| sdake | not global | 23:01 |
| inc0 | ehh...you have neutron-base and nova-base with same repos or each with different? | 23:01 |
| sdake | inc0 different repos | 23:02 |
| sdake | example - one is juno one is kilo | 23:02 |
| sdake | we have real world use cases around this scenario and we are not alone | 23:03 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Refactor of base Dockerfile to allow custom repo https://review.openstack.org/323589 | 23:03 |
| sdake | now take that example out to the other stuff you want to customize | 23:03 |
| inc0 | but in same deployment 2 containers will have different stuff in it? | 23:03 |
| sdake | the binary packages installed | 23:03 |
| sdake | the source packages installed | 23:03 |
| sdake | inc0 neutron will have kilo nova will have liberty | 23:04 |
| sdake | inc0 but I want to run build.py once | 23:04 |
| inc0 | ok...you do that by tags... | 23:04 |
| sdake | clearly I could hack the base dockerfile and rebuild each one indivdually | 23:04 |
| sdake | your solution is more complex then mine is ;) | 23:04 |
| inc0 | no, you should have tags for mitaka and liberty and override ansible deploy | 23:05 |
| sdake | explain tags please | 23:05 |
| inc0 | that's how you should do it | 23:05 |
| inc0 | we do have vars in ansible for that | 23:05 |
| sdake | yes i understand | 23:05 |
| sdake | thatis one approach | 23:05 |
| sdake | listen, I can't give you the goddamn fucking examle that I want | 23:05 |
| sdake | because i am not permitted | 23:05 |
| inc0 | that's only approach because we don't support mitaka code deploying liberty repos | 23:05 |
| *** ravig has quit IRC | 23:05 | |
| sdake | what I can do is give you examples that are "LIKE" the example I have | 23:06 |
| inc0 | ok... | 23:06 |
| sdake | but in essence, I need repo files per container | 23:06 |
| sdake | pick a reason out of thin air | 23:06 |
| sdake | I did :) | 23:06 |
| inc0 | still doable | 23:06 |
| inc0 | you add block neutron-lbaas-pre-install | 23:06 |
| sdake | so you will have 119 custom files for repos? | 23:07 |
| inc0 | and then in your custom-dockerfile you change stuff there | 23:07 |
| inc0 | no, just these you want to edit | 23:07 |
| inc0 | and yes, we can scale it down to even one file | 23:07 |
| inc0 | with lots of blocks in it | 23:07 |
| inc0 | but I feel that would be horrible and we should have customization file per dockerfile | 23:07 |
| sdake | 119 containers | 23:07 |
| inc0 | either way works | 23:07 |
| sdake | do the math :) | 23:07 |
| inc0 | yes, I can do all the customization in single file if you want | 23:08 |
| sdake | i think when you get done with your dockerfile it will look more like a jinj2 file and not a dockerfile | 23:08 |
| sdake | and it will still be DSL->DSL | 23:08 |
| sdake | but not general purpose | 23:08 |
| inc0 | with block neutron-lbaas-pre-install -> <<change repos here>> and then nova-compute-pre-install -> << change repo there >> | 23:08 |
| sdake | rather general purpose and not domain specific | 23:08 |
| inc0 | well, your dsl will look like completely nothing around | 23:09 |
| inc0 | I'd rather have it look like jinja2 based dockerfile | 23:09 |
| sdake | oh well the spec process is started since once is submitted | 23:09 |
| sdake | community can decide | 23:10 |
| inc0 | true | 23:10 |
| inc0 | that's how we should handle that | 23:10 |
| inc0 | have a proper discussion and then make a call as community | 23:10 |
| inc0 | we'll make it work sdake so don't worry, your day job won't get hurt | 23:11 |
| sdake | duh | 23:11 |
| sdake | what I have already works for producing dockerfiles | 23:11 |
| sdake | my day job is not in jeaporady | 23:11 |
| inc0 | I'm saying we can do all of that without rearchitecturing whole thing | 23:11 |
| sdake | like i said before, I'd strongly have preferred to have done a DSL upfront | 23:11 |
| sdake | but I didn't think we could pull it off | 23:11 |
| sdake | dsl was discussed at our first midcycle | 23:12 |
| inc0 | and now we released stable | 23:12 |
| inc0 | and we need to follow deprecation | 23:12 |
| inc0 | we can't just throw away stuff now | 23:12 |
| sdake | i wasn't suggesting not following deprecation polices | 23:12 |
| sdake | but dockerfiles are not somethign that are part of the deprecation process | 23:12 |
| sdake | only cli and config options are | 23:13 |
| inc0 | well, that'd mean that your dsl would be our main way in O at its earliest | 23:13 |
| inc0 | and I can give you working solution before N | 23:13 |
| sdake | i was thinking more like newton | 23:13 |
| inc0 | no, | 23:13 |
| sdake | dockerfiles dont have any deprecation policy associated with them | 23:13 |
| inc0 | deprecation means whatever used to work will work for 1 release | 23:13 |
| sdake | that would be like saying "setup.py has a deprecation policy" | 23:13 |
| sdake | inc0 yes user visible changes work the same | 23:14 |
| inc0 | well, if you can make build.py with every possible option work, that would be sufficient for me | 23:14 |
| sdake | all the jinja2 stuff is black magic the operator never sees | 23:14 |
| inc0 | as for deprecation goes, I still don't like DSL | 23:14 |
| inc0 | well, not true | 23:14 |
| inc0 | people do change dockerfiles now | 23:14 |
| sdake | that needs to stop | 23:14 |
| sdake | or we are in a world of pain coming | 23:14 |
| *** mark-casey has left #openstack-kolla | 23:15 | |
| inc0 | nah, I like that | 23:15 |
| inc0 | if we allow them to make snowflakes without affecting trunk | 23:15 |
| inc0 | it's on them | 23:15 |
| inc0 | and they're happy with that | 23:15 |
| wirehead_ | It should at the very least be a big black block in the docs. | 23:15 |
| sdake | wirehead_ which should be | 23:15 |
| wirehead_ | People making snowflake dockerfiles. | 23:16 |
| inc0 | well, if we make jinja2 extensible, we can make super detailed howto make kolla behave however you like | 23:16 |
| sdake | i had not thought of this include mechanism, it works sort of like how our /etc/kollaconfig directory works, but I have serious concerns about being able to implement it and have it work correctly | 23:18 |
| *** rhallisey has joined #openstack-kolla | 23:18 | |
| sdake | as well as the raw number of custom files needed | 23:19 |
| *** diogogmt has quit IRC | 23:19 | |
| *** pbourke-home has joined #openstack-kolla | 23:20 | |
| inc0 | pbourke-home, https://review.openstack.org/#/c/323589/ <- demo I promised you this morning | 23:21 |
| patchbot | inc0: patch 323589 - kolla - Refactor of base Dockerfile to allow custom repo | 23:21 |
| *** pbourke-home has quit IRC | 23:22 | |
| *** vhosakot has joined #openstack-kolla | 23:37 | |
| *** vhosakot has quit IRC | 23:43 | |
| *** vhosakot has joined #openstack-kolla | 23:44 | |
| sdake | inc0 are you planning to have debain-dockerfile.j2 and ubuntu-dockerfile.j2 in every container directory? | 23:45 |
| inc0 | havent decided yet, playing around | 23:46 |
| *** vhosakot has quit IRC | 23:46 | |
| *** daneyon_ has joined #openstack-kolla | 23:46 | |
| prithiv | is it just me or anyone else facing issue with master branch | 23:47 |
| prithiv | ImportError: No module named plugins.action | 23:47 |
| prithiv | in task TASK: [ceph | Copying over ceph.conf] ***************************************** | 23:47 |
| prithiv | though i have enable_ceph: "no" | 23:47 |
| sbezverk | prithiv you need to update your ansible | 23:47 |
| prithiv | i have 1.9.4 | 23:48 |
| prithiv | recommended in quickstart guide | 23:48 |
| sbezverk | I had similar issue with ansiblew 1.9.4 but with new 2.0.2 it works fine | 23:48 |
| wirehead_ | prithiv: The quickstart guide hasn’t been updated. | 23:48 |
| *** daneyon_ has quit IRC | 23:51 | |
| *** ravig has joined #openstack-kolla | 23:52 | |
| *** berendt has quit IRC | 23:54 | |
| wirehead_ | rhallisey: I’m digging into mariadb bootstrapping. Is that just not ready yet? | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
