*** Haomeng has joined #kolla | 00:03 | |
*** achanda_ has quit IRC | 00:03 | |
*** achanda has quit IRC | 00:05 | |
*** Haomeng|2 has quit IRC | 00:05 | |
*** diogogmt has quit IRC | 00:07 | |
*** gfidente has quit IRC | 00:14 | |
*** zhiwei has joined #kolla | 00:20 | |
*** weiyu has quit IRC | 00:20 | |
*** shadower has quit IRC | 00:23 | |
*** banix has joined #kolla | 00:26 | |
*** Haomeng|2 has joined #kolla | 00:37 | |
*** Haomeng has quit IRC | 00:40 | |
*** dolpher has joined #kolla | 00:46 | |
*** dims_ has quit IRC | 00:53 | |
*** weiyu_ has joined #kolla | 01:06 | |
*** dims_ has joined #kolla | 01:09 | |
jpeeler | is there supposed to exist a kollaglue/centos-rdo-kolla-ansible image? | 01:11 |
---|---|---|
*** dims__ has joined #kolla | 01:15 | |
*** dims__ has quit IRC | 01:16 | |
*** dims_ has quit IRC | 01:17 | |
*** dims_ has joined #kolla | 01:18 | |
*** mfalatic has quit IRC | 01:21 | |
*** mfalatic_ has joined #kolla | 01:21 | |
*** erkules has joined #kolla | 01:21 | |
*** erkules_ has quit IRC | 01:23 | |
*** jasonsb has joined #kolla | 01:24 | |
*** dolpher has quit IRC | 01:39 | |
*** dolpher has joined #kolla | 01:40 | |
*** dims__ has joined #kolla | 01:42 | |
*** dims_ has quit IRC | 01:44 | |
*** dolpher1 has joined #kolla | 01:44 | |
jpeeler | i just pushed it. | 01:45 |
*** dolpher has quit IRC | 01:47 | |
*** dims_ has joined #kolla | 01:54 | |
*** unicell has quit IRC | 01:56 | |
*** dims__ has quit IRC | 01:56 | |
SamYaple | jpeeler: no | 02:09 |
SamYaple | there shouldnt be any images with the latest tag | 02:09 |
jpeeler | SamYaple: huh? i know i've been out of the loop for a while | 02:13 |
jpeeler | https://github.com/stackforge/kolla/commit/a5fa276c3a8f30a25e009d69ee93a80635337caa | 02:13 |
jpeeler | but that makes it seem like they all should be using latest | 02:13 |
SamYaple | yes, master uses the latest tag | 02:13 |
SamYaple | we dont push latest images | 02:13 |
SamYaple | basically if you use the latest tag you must be building your own images | 02:13 |
jpeeler | should we remove all the latest tags present then? | 02:14 |
SamYaple | yea im working on it at the moment | 02:14 |
SamYaple | it just leads us to outdated images | 02:14 |
jpeeler | yeah i understand | 02:14 |
SamYaple | the only thing we can gaurantee to work is someone checkouts the liberty2 tag and has libert2 images to pull without building | 02:14 |
jpeeler | right | 02:14 |
SamYaple | we cant expect the random tags to all work with master | 02:14 |
jpeeler | SamYaple: you may be able to fix this in a jiffy - https://bugs.launchpad.net/kolla/+bug/1482406 | 02:20 |
openstack | Launchpad bug 1482406 in kolla "ansible deploy yielding undefined variable " [Undecided,New] | 02:20 |
SamYaple | looking | 02:20 |
jpeeler | i haven't looked at it honestly | 02:20 |
SamYaple | did you setup the 'network_interface' in your globals.yml correctly? | 02:21 |
SamYaple | also it looks like thats calling multinode inventory, is this a multinode deploy? | 02:21 |
jpeeler | it is multinode yes, and i honestly can't say for sure if eth0 is correct, but it definitely exists | 02:23 |
SamYaple | are your hosts reachable via the address "network01" ? | 02:23 |
jpeeler | yes | 02:23 |
SamYaple | i mean that error to mea is pretty straight forward, you dont have an eth0 interface with an ip address | 02:23 |
SamYaple | can i reproduce this somehow? | 02:24 |
jpeeler | it is present with an ip | 02:24 |
*** achanda has joined #kolla | 02:24 | |
jpeeler | SamYaple: i think harmw reproduced it also. we're using vagrant, so perhaps that environment is not quite right | 02:25 |
SamYaple | i know nothing about vagrant | 02:25 |
jpeeler | is the docker in docker stuff close to finished? | 02:25 |
SamYaple | but from an ansible perspective it may not be gatehering facts correctly | 02:25 |
SamYaple | jpeeler: the containerized deps just merged, but i have to switch the modules we use first | 02:26 |
SamYaple | then i can finish the dind stuff | 02:26 |
SamYaple | i think ansible uses netifaces to pull network info, makesure it is installed | 02:26 |
jpeeler | it isn't actually | 02:27 |
SamYaple | hhmm | 02:27 |
jpeeler | nevermind, pip installed | 02:27 |
jpeeler | checked rpm first | 02:27 |
SamYaple | ah | 02:27 |
SamYaple | if you run `ansible -i inventory/multinode -m setup` | 02:28 |
SamYaple | do you see the interface eth0 returned? | 02:28 |
SamYaple | if its not returning wit hthe ip info, thats the issue | 02:28 |
jpeeler | is that exactly the correct command? | 02:29 |
jpeeler | giving me usage output | 02:30 |
SamYaple | yea thats running the setup module | 02:31 |
SamYaple | it returns everythign about the system it can return | 02:31 |
SamYaple | from bios version to ip addresses | 02:31 |
SamYaple | whats your ansible version? | 02:31 |
jpeeler | 1.9.2 | 02:31 |
SamYaple | yup latest | 02:31 |
SamYaple | man we really need to clean up our dockerhub | 02:33 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Update Ansible docs to remove requirements https://review.openstack.org/209716 | 02:37 |
SamYaple | hey jpeeler | 02:42 |
SamYaple | there is lots to do on the build.py script | 02:42 |
jpeeler | i'm sure | 02:42 |
SamYaple | i dont know of anything in progress for that except your patchsets | 02:42 |
SamYaple | so you make it what you want too | 02:42 |
jpeeler | yeah | 02:42 |
SamYaple | i will be quick to review it for you | 02:42 |
jpeeler | cool thanks | 02:43 |
openstackgerrit | Merged stackforge/kolla: Glance api and Glance registry shouldn't share tags https://review.openstack.org/210109 | 02:46 |
*** ccrouch has joined #kolla | 03:28 | |
*** diogogmt has joined #kolla | 03:40 | |
*** achanda has quit IRC | 03:56 | |
*** dims_ has quit IRC | 04:04 | |
*** loth has joined #kolla | 04:24 | |
*** weiyu_ has quit IRC | 04:29 | |
*** jruano has joined #kolla | 04:39 | |
*** unicell has joined #kolla | 04:49 | |
*** jruano has quit IRC | 04:58 | |
*** diogogmt has quit IRC | 05:00 | |
*** weiyu_ has joined #kolla | 05:14 | |
*** unicell has quit IRC | 05:15 | |
*** banix has quit IRC | 05:18 | |
*** unicell has joined #kolla | 05:19 | |
*** achanda has joined #kolla | 05:25 | |
*** sdake has joined #kolla | 05:29 | |
sdake | evening folks | 05:29 |
sdake | samyaple you around | 05:29 |
*** echoingumesh has joined #kolla | 05:31 | |
*** echoingu_ has joined #kolla | 05:34 | |
*** echoingumesh has quit IRC | 05:37 | |
coolsvap | sdake, evening :) morning for me | 06:09 |
ccrouch | SamYaple: another ping :-) | 06:34 |
ccrouch | sdake: ut? | 06:34 |
ccrouch | going to bed shortly, but if anyone had any feedback on | 06:37 |
ccrouch | https://etherpad.openstack.org/p/8ZACSjaDCH | 06:37 |
ccrouch | its an attempt to describe the high level, and sometimes low level steps, that kolla use to get the openstack services started | 06:38 |
ccrouch | when using config-external-copy-once and ansible obviously | 06:38 |
ccrouch | cheers! | 06:40 |
*** sdake has quit IRC | 06:58 | |
*** inc0 has joined #kolla | 07:02 | |
inc0 | good morning | 07:03 |
coolsvap | morning inc0 | 07:10 |
*** achanda has quit IRC | 07:15 | |
*** mandre has quit IRC | 07:19 | |
*** pbourke has quit IRC | 07:33 | |
*** pbourke has joined #kolla | 07:34 | |
*** jmccarthy has quit IRC | 07:36 | |
*** jmccarthy has joined #kolla | 07:37 | |
*** gfidente has joined #kolla | 07:50 | |
*** macjack has left #kolla | 07:54 | |
*** inc0_ has joined #kolla | 08:13 | |
*** inc0 has quit IRC | 08:16 | |
*** inc0_ has quit IRC | 08:27 | |
*** inc0 has joined #kolla | 08:27 | |
*** erkules has quit IRC | 08:29 | |
*** erkules has joined #kolla | 08:29 | |
*** dwalsh has joined #kolla | 08:30 | |
*** weiyu_ has quit IRC | 08:46 | |
*** weiyu_ has joined #kolla | 08:46 | |
*** tfukushima has joined #kolla | 08:50 | |
coolsvap | pbourke, around? | 08:50 |
vbel | good morning/evening folks | 08:58 |
*** vbel has quit IRC | 08:58 | |
*** vbel has joined #kolla | 08:59 | |
*** dolpher1 has quit IRC | 09:00 | |
pbourke | coolsvap: morning | 09:02 |
coolsvap | pbourke, morning i built environment for vagrant + virtualbox till i get it running with kvm, i have a couple of queries | 09:03 |
coolsvap | do you have time? | 09:03 |
*** dwalsh has quit IRC | 09:11 | |
pbourke | shoot | 09:15 |
coolsvap | when i run ansible-playbook the setup stops at haproxy itself, i checked if its downloading any images or something but there's nothing in background | 09:19 |
pbourke | i haven't tried haproxy myself yet | 09:22 |
pbourke | try -vvv and see what command the playbook is hanging at? | 09:22 |
inc0 | pbourke, did you manage to set up neutron kilo? | 09:23 |
pbourke | inc0: again, haven't really tried. i tend to just test small parts in isolation, its a lot of overhead to test the whole system | 09:24 |
pbourke | i probably should but right now just haven't had the need | 09:24 |
inc0 | ok, I'm trying to get this working, but so far I hit problems | 09:25 |
inc0 | I'll let you know if there is anything to be done | 09:25 |
pbourke | actually, I did do neutron a while back. but I only tested the easy parts, didn't actually try booting an instance, attaching an ip, etc | 09:26 |
*** dwalsh has joined #kolla | 09:29 | |
*** weiyu_ has quit IRC | 09:34 | |
*** dims_ has joined #kolla | 09:54 | |
*** inc0 has quit IRC | 09:55 | |
openstackgerrit | Alicja Kwasniewska proposed stackforge/kolla: Partially-implements: bp dockerfile-template https://review.openstack.org/210288 | 09:56 |
*** echoingu_ has quit IRC | 09:56 | |
*** echoingumesh has joined #kolla | 09:57 | |
*** inc0 has joined #kolla | 09:58 | |
*** weiyu has joined #kolla | 10:00 | |
*** echoingumesh has quit IRC | 10:01 | |
*** inc0 has quit IRC | 10:05 | |
*** inc0_ has joined #kolla | 10:05 | |
SamYaple | ccrouch: ping | 10:12 |
SamYaple | ccrouch: responded | 10:16 |
vbel | SamYaple: hi, any thoughts on how to proceed with run_once bootstrap? I suggest as workaround copy of non-pid block from start.yml into bootstrap.yml. It works. Do we ever need to bootstrap anything with pid=host? | 10:17 |
SamYaple | vbel: https://review.openstack.org/#/c/210041/ | 10:18 |
SamYaple | thats my thoughts on proceeding | 10:18 |
*** bradjones has quit IRC | 10:20 | |
vbel | SamYaple: +2 :) | 10:22 |
vbel | will be easier to trace/debug things | 10:24 |
*** walters has quit IRC | 10:25 | |
*** walters has joined #kolla | 10:27 | |
*** sdake has joined #kolla | 10:28 | |
sdake | morning | 10:29 |
sdake | hey ccrouch up late | 10:30 |
*** bradjones_ has joined #kolla | 10:30 | |
sdake | samyaple ping | 10:30 |
*** bradjones_ is now known as bradjones | 10:30 | |
sdake | ccrouch when you wake up that workflow looks good - should make an md for it to explain it to others! | 10:33 |
sdake | pbourke around | 10:39 |
pbourke | sdake: hi | 10:39 |
sdake | hey bro can you review my blueprint | 10:39 |
*** dwalsh has quit IRC | 10:39 | |
pbourke | sure, throw it out | 10:39 |
sdake | https://blueprints.launchpad.net/kolla/+spec/deploy-ha | 10:40 |
*** inc0_ has quit IRC | 10:40 | |
sdake | the idea is to have a two stage ansible deploy actually | 10:41 |
sdake | so you do an ansible step1 which builds your deployment nodes | 10:42 |
sdake | then you customize ansible config on one of those 2-3 nodes | 10:42 |
sdake | and deploy openstack next | 10:42 |
pbourke | kevsi: you might be interested in a read of this also ^ | 10:42 |
pbourke | sdake: has bmace read this also? | 10:43 |
sdake | yes bmace helped come up with it sort of | 10:43 |
sdake | he asked me 2-3 months ago "the deploy node needs to be ha" | 10:43 |
sdake | and I just sorted out how t od o it :) | 10:43 |
pbourke | yeah its a feature I know we originally were looking to do using his deploy tool | 10:44 |
pbourke | it makes sense though to handle most of this with ansible I think | 10:44 |
sdake | i gave a schedule of 1 week which seems doable | 10:44 |
*** bradjones has quit IRC | 10:45 | |
*** ccrouch has quit IRC | 10:45 | |
*** daneyon has quit IRC | 10:45 | |
*** openstackgerrit has quit IRC | 10:45 | |
*** smeisner has quit IRC | 10:45 | |
*** thomasem has quit IRC | 10:45 | |
sdake | atleast the etcd backend | 10:45 |
sdake | what i told yo uabout with the deploy after deploy is not documented anywhere | 10:45 |
sdake | i'm not sure if my ansible assertion is correct | 10:46 |
pbourke | yeah thats one question I have after reading | 10:46 |
pbourke | is what form will the backend take | 10:46 |
pbourke | is it etcd? | 10:46 |
sdake | nce Ansible gets all state of the current system during the runtime, it doesn't need any more storage then a) whether a deployment operation is in progress and b) the configuration of the deployment. | 10:47 |
sdake | configurable pbourke | 10:47 |
sdake | but etcd is what i'm gonig to implement first | 10:47 |
*** bradjones has joined #kolla | 10:49 | |
*** ccrouch has joined #kolla | 10:49 | |
*** daneyon has joined #kolla | 10:49 | |
*** openstackgerrit has joined #kolla | 10:49 | |
*** smeisner has joined #kolla | 10:49 | |
*** thomasem has joined #kolla | 10:49 | |
pbourke | it sounds good, definitely something we need | 10:49 |
pbourke | so im +1 :) | 10:49 |
pbourke | can you explain the deployment_active flag a little more? | 10:49 |
*** bradjones has quit IRC | 10:49 | |
*** bradjones has joined #kolla | 10:49 | |
*** bradjones has quit IRC | 10:49 | |
*** bradjones has joined #kolla | 10:49 | |
*** pbourke has quit IRC | 10:50 | |
sdake | when you go to deploy it gets ste to true or osmething | 10:51 |
sdake | when deploy finishes it gets unset | 10:51 |
sdake | if deployment starts and it sets, deployment aborts | 10:51 |
sdake | the only thing that would keep the flag set while a ansible deployment wasn't going on is a data center lights out or b) ansible crash | 10:51 |
sdake | in eithre of those situations you would need a way to blow away the lock | 10:52 |
*** pdb has joined #kolla | 10:53 | |
sdake | pdb | 10:53 |
*** pdb is now known as pbourke | 10:54 | |
sdake | 03:51:05] <sdake>when you go to deploy it gets ste to true or osmething | 10:54 |
sdake | [03:51:11] <sdake>when deploy finishes it gets unset | 10:54 |
sdake | [03:51:19] <sdake>if deployment starts and it sets, deployment aborts | 10:54 |
sdake | [03:51:54] <sdake>the only thing that would keep the flag set while a ansible deployment wasn't going on is a data center lights out or b) ansible crash | 10:54 |
sdake | [03:52:11] <sdake>in eithre of those situations you would need a way to blow away the lock | 10:54 |
pbourke | is it possible to run more than one ansible at a time though? | 10:54 |
sdake | pleae add comment to the blueprint - that is what discussion phase is for and I'd like to get cracking soon | 10:55 |
sdake | no if you run a second ansible it errors out and says "op in progress" :) | 10:55 |
sdake | unless yu know better, and can wipe the flag | 10:55 |
sdake | (in reality there is no op in progress) | 10:55 |
sdake | think of it like a pid file in /var/lock | 10:56 |
*** echoingumesh has joined #kolla | 10:57 | |
*** bmace has quit IRC | 10:58 | |
*** echoingumesh has quit IRC | 11:02 | |
vbel | https://review.openstack.org/#/c/208451/ containerizing host deps - was it easier than to write ansible host provisioning task? | 11:03 |
*** akwasnie has joined #kolla | 11:03 | |
*** inc0 has joined #kolla | 11:03 | |
*** bmace has joined #kolla | 11:04 | |
vbel | ... including installation of docker itself | 11:04 |
*** inc0 has quit IRC | 11:07 | |
*** inc0_ has joined #kolla | 11:07 | |
pbourke | vbel: not sure about easier but I think its better | 11:11 |
pbourke | vbel: we get the benefits of docker and not having to taint the host | 11:12 |
pbourke | vbel: we could still maybe do a provisoning task though for docker itself (and docker-py) | 11:12 |
*** zhiwei has quit IRC | 11:14 | |
vbel | docker-py is not included in docker rpms? | 11:15 |
vbel | pbourke: yes, having docker 'agent' to manage the rest is good idea. It has been merged as I understand? no more host dependencies? | 11:17 |
sdake | vbel in that container is only shade and its 50+ dependencies | 11:17 |
sdake | vbel trust me you dont want those on each host | 11:17 |
sdake | results in unreliable operation imo | 11:17 |
vbel | sdake: thanks, sure :) | 11:17 |
sdake | i haven't actually looked at how sam implemented that | 11:18 |
sdake | let me look | 11:18 |
pbourke | yup seems merged | 11:18 |
vbel | so from now on I will not provision hosts with anything but docker itself (and docker-py ) | 11:19 |
sdake | holy christ i don't even know how that implemetnation works | 11:19 |
sdake | but I assume it does ;-) | 11:19 |
vbel | :)))) That's a big change | 11:19 |
sdake | sadface container lsymlinking into ansible directory | 11:21 |
sdake | https://blueprints.launchpad.net/kolla/+spec/deploy-ha | 11:22 |
sdake | pbourke^^ add something to whiteboard saying you review and approve plz :) | 11:22 |
sdake | dont approve it | 11:22 |
sdake | just say you approve of the idea | 11:22 |
sdake | or if you dont say that | 11:22 |
sdake | pbourke do you know how the haproxy container works atm | 11:32 |
*** tfukushima has quit IRC | 11:34 | |
*** dwalsh has joined #kolla | 11:37 | |
*** echoingumesh has joined #kolla | 11:58 | |
vbel | haproxy question - do we do net.ipv4.ip_nonlocal_bind=1 anywhere to get it bind to VIP that is not on current node | 12:01 |
vbel | ? | 12:01 |
vbel | I see, we do | 12:01 |
sdake | pbourke vbel here is part two https://blueprints.launchpad.net/kolla/+spec/deploy-deploynodes | 12:02 |
sdake | comments welcome | 12:02 |
vbel | ./roles/haproxy/tasks/config.yml: sysctl: name="net.ipv4.ip_nonlocal_bind" value=1 sysctl_set=yes <- good stuff | 12:02 |
sdake | please use whiteboard - i'd like to finish the discussion so we can either proceed or say we are not doing it ;) | 12:03 |
*** echoingumesh has quit IRC | 12:07 | |
*** banix has joined #kolla | 12:11 | |
*** banix has quit IRC | 12:11 | |
sdake | also this might be m material but i'd like to see if i can get it in for liberty in a tech previewish sort of way | 12:14 |
*** sdake_ has joined #kolla | 12:16 | |
*** sdake_ has quit IRC | 12:16 | |
*** sdake_ has joined #kolla | 12:16 | |
*** sdake__ has joined #kolla | 12:17 | |
*** sdake has quit IRC | 12:18 | |
pbourke | sdake_: is there one part in particular you're wondering about? | 12:18 |
*** sdake_ has quit IRC | 12:21 | |
sdake__ | just want cks or nacks | 12:22 |
sdake__ | what verification the idea looks ok | 12:22 |
sdake__ | can you do me a solid also pbourke and try pip install openstack-crux | 12:22 |
sdake__ | I am loki of asguard | 12:23 |
sdake__ | and I am burdened with glorious purpose | 12:23 |
sdake__ | we have no quarrel with your people | 12:23 |
sdake__ | an ant has no quarrel with a boot ;-) | 12:23 |
sdake__ | this move rocks | 12:23 |
sdake__ | 10k reviews and like 4.75 stars | 12:23 |
sdake__ | definately the best marvel movie | 12:23 |
*** dwalsh has quit IRC | 12:24 | |
pbourke | antman? | 12:24 |
pbourke | sdake__: http://paste.openstack.org/show/412174/ | 12:26 |
sdake__ | groan I get that too | 12:27 |
sdake__ | well atleast I got it uploaded ;-) | 12:27 |
sdake__ | any tips on how to fix that python error? | 12:28 |
sdake__ | i just took it out in the rpm version and hard specified the dependencies | 12:28 |
sdake__ | but that wont obviously work for from source installs | 12:28 |
sdake__ | it would be nice if we can just deprecate config-internal - then crux goes byebye | 12:29 |
*** britthouser has quit IRC | 12:31 | |
pbourke | ive not seen any engagement from tripleo | 12:32 |
pbourke | going by their latest email they're only starting to look at kolla | 12:32 |
sdake__ | they have made a pretty big commitment | 12:33 |
pbourke | in talk or code | 12:33 |
sdake__ | dan prince spending 1-2 months working on this project | 12:33 |
sdake__ | he has got the code into an operational state for puppet and tripleo iirc | 12:33 |
*** dims_ has quit IRC | 12:33 | |
sdake__ | iiuc | 12:33 |
sdake__ | plus rhallisey/sloewr are interested in the tripleo case | 12:33 |
sdake__ | this is good for us, it makes our containers mo better | 12:33 |
*** dims_ has joined #kolla | 12:34 | |
pbourke | are they using config-internal? | 12:34 |
sdake__ | offers mor ecustomer choice with the same container infrastructure | 12:34 |
sdake__ | the proposal on the maling list was to use config-external | 12:34 |
sdake__ | initiailly they were using config-internal | 12:34 |
sdake__ | I would have deprecated it in the ansible-multi spec | 12:34 |
sdake__ | but tripleo was going down thatroute first | 12:34 |
*** tfukushima has joined #kolla | 12:35 | |
sdake__ | but I think they figured out as we did, config-internal is not ideal | 12:35 |
sdake__ | the reason I didn't like config external initially was there was not immutability | 12:35 |
sdake__ | but kevin fox sort of soled that with my help :) | 12:35 |
pbourke | anywho | 12:36 |
pbourke | for crux, it seems like you just didn't package requirements.txt? | 12:36 |
sdake__ | its definately there | 12:37 |
sdake__ | i am converting to pbr | 12:37 |
sdake__ | i'll have a new version in 15 mins for you to test if your available | 12:37 |
*** rhallisey has joined #kolla | 12:37 | |
pbourke | sounds good | 12:38 |
sdake__ | rhallisey can you get the tripleo folks to make a deicsion on config/internal vs config/external | 12:38 |
sdake__ | sounds like they want config external | 12:38 |
sdake__ | if we can deprecate config-internal before liberty I'd really like to do so | 12:38 |
rhallisey | config external | 12:38 |
sdake__ | so we aren't stuck with it forever | 12:38 |
rhallisey | the config internal deprecation will be quite the process | 12:38 |
rhallisey | since ansible will have to be to the default | 12:39 |
*** Haomeng has joined #kolla | 12:39 | |
rhallisey | a lot of the dos wills change | 12:39 |
sdake__ | here is the deal, I made a serious commitment in the ansible multi spec to tripleo specifically to maintain config-internal | 12:39 |
rhallisey | ya we've change | 12:39 |
rhallisey | we can deprecate | 12:39 |
sdake__ | good, so can we get a couple tripleo cores to sign off on the mailing list to that effect | 12:40 |
*** tfukushima has quit IRC | 12:40 | |
sdake__ | so someone doesn't bitch later that I ninja changed the entire architecture | 12:40 |
rhallisey | sure | 12:40 |
sdake__ | dprince & slagle would wfm :) | 12:40 |
sdake__ | I REALLY want ot get rid of config-internal | 12:40 |
sdake__ | i'll do all the commits to do the job | 12:40 |
sdake__ | the reason its there is for tripleo | 12:41 |
sdake__ | because we dont abandon our partner projects | 12:41 |
sdake__ | you guys had a config-internal implementation and if we just took that out, it might have sent the wrong message | 12:41 |
*** jruano has joined #kolla | 12:41 | |
vbel | config-internal is not to be supported by us either :) | 12:41 |
*** Haomeng|2 has quit IRC | 12:41 | |
sdake__ | if the tripleo community really is satisified with dprince's config strategy (which looks totally sound) | 12:41 |
rhallisey | dprince is cool with it. The purpose of that email was to get opinions | 12:42 |
sdake__ | lets get it on the record so we can nuke config-internal | 12:42 |
sdake__ | we only have one month to do the job | 12:42 |
rhallisey | kk | 12:42 |
sdake__ | pbourke try now please pip install crux in container | 12:43 |
sdake__ | make sure deps get installed | 12:43 |
vbel | config-internal is promise to please everyone (story of https://en.wikipedia.org/wiki/Vasa_%28ship%29 ) | 12:44 |
sdake__ | However, Vasa was dangerously unstable due to too much weight in the upper structure of the hull. | 12:45 |
pbourke | SamYaple: same error unfortunately | 12:45 |
sdake__ | sdake actually :) | 12:45 |
*** sdake__ is now known as sdake | 12:45 | |
sdake | damn don't coufse me with sam plz!! | 12:46 |
pbourke | ah wrong autocomplete | 12:46 |
sdake | let me try deleting | 12:46 |
pbourke | SamYaple: please ignore last comment! | 12:46 |
vbel | sdake: they wanted to combine everything architecture-wise :) | 12:47 |
sdake | pbourke pull the trigger again | 12:52 |
pbourke | sdake: oo so close | 12:53 |
sdake | got a paste? | 12:53 |
pbourke | sdake: installs successfuly but 'crux -v' complains it's missing the debtcollector module | 12:53 |
sdake | did it install python-keystoneclient? | 12:54 |
pbourke | pip install debtcollector fixes it | 12:54 |
pbourke | so I reckon add that to requirements.txt ? | 12:54 |
sdake | debtcollector is not used in crux | 12:54 |
sdake | it must be a dependency of something else | 12:54 |
sdake | got a backtrace? | 12:54 |
sdake | that implies requirements.txt isn't being installed | 12:54 |
pbourke | one sec | 12:54 |
pbourke | http://paste.openstack.org/show/412179/ | 12:55 |
sdake | are you doing this in a container? | 12:55 |
pbourke | no | 12:55 |
pbourke | I'll do that now too | 12:55 |
sdake | yes please do i want to see the full container build in a log pls | 12:56 |
pbourke | works in the container | 12:56 |
sdake | Requirement already satisfied (use --upgrade to upgrade): pbr>=0.11,<2.0 in /usr/local/lib/python2.7/dist-packages (from openstack-crux) | 12:56 |
sdake | Requirement already satisfied (use --upgrade to upgrade): python-keystoneclient>=0.9.0 in /usr/local/lib/python2.7/dist-packages (from openstack-crux) | 12:56 |
sdake | Requirement already satisfied (use --upgrade to upgrade): cliff>=1.4.3 in /usr/local/lib/python2.7/dist-packages (from openstack-crux) | 12:56 |
sdake | so it is installing the requirements | 12:56 |
sdake | i think you just have a dependency problem on your host machine | 12:57 |
sdake | if it works int he container | 12:57 |
sdake | does crux work inthe container? | 12:58 |
pbourke | yeah it works in the container so I think we're good | 12:58 |
sdake | cool i'll submit a pr to larsks repo | 12:59 |
sdake | hopefully he will take a pbr implementation | 12:59 |
sdake | toerhwise I dont know how to fix it | 12:59 |
sdake | pbourke do you have the original log for the pip failure we were experiencing | 13:00 |
pbourke | I dont think so | 13:00 |
pbourke | sorry | 13:00 |
sdake | damn | 13:01 |
sdake | i lost it too | 13:01 |
sdake | would have been nice for larsks to understand why :) | 13:01 |
sdake | let me delete the update and reprouce the log | 13:02 |
sdake | nah that wont work | 13:02 |
*** echoingumesh has joined #kolla | 13:05 | |
sdake | larsks https://github.com/larsks/crux/pull/3 | 13:07 |
sdake | larsks if you know another way to fix this problem i'm open to it :) | 13:08 |
larsks | sdake: taking a look! :) | 13:08 |
sdake | larsks if you could tag as version 8 I'd appreciate that as well | 13:08 |
larsks | sdake: actually, would you mind deleting your pypi repository for this? I'd like to maintain the pypi packaging. | 13:09 |
sdake | i have no problem with you maintnaining the pypi | 13:09 |
larsks | I'll merge and push a new pypi today. | 13:09 |
sdake | can I just transfer ownership to you? | 13:09 |
larsks | I don't know, is that a thing? | 13:09 |
sdake | I think so | 13:09 |
sdake | do you mind openstack-crux? | 13:09 |
sdake | because crux is taken | 13:09 |
larsks | Fine by me. Does keystone-crux make more sense? I'm happy either way. | 13:10 |
sdake | let me see how the transfer works - hang tight | 13:10 |
sdake | keystone-crux is good | 13:10 |
sdake | that is probably better | 13:10 |
*** echoingumesh has quit IRC | 13:10 | |
sdake | i'll see if I can delete openstack-crux - not sure if its possible to delete repos or not on pypi | 13:10 |
sdake | can you register it now? | 13:10 |
sdake | use python setup.py register | 13:11 |
sdake | have to change openstack-crux in my example | 13:11 |
sdake | err my pulll requet | 13:11 |
sdake | request | 13:11 |
sdake | larsks are you good with the pbr changes tho? | 13:11 |
larsks | I suspect I know why you were getting problems with pypi. | 13:12 |
sdake | larsks I personally dont care either way as long as the pypi install works properly for from-source builds | 13:12 |
larsks | I may look at that first. | 13:12 |
sdake | it said requirements.txt was not found | 13:12 |
sdake | if you could get it done soon it would be appreciated, because atm we are a bit blocked on this | 13:12 |
larsks | No problem. | 13:12 |
sdake | how about this, I'll leae openstack-crux up until you get done | 13:13 |
sdake | because I know you ahvea big plate fulll of shit | 13:13 |
sdake | and when you have a working pypi install we will roll with that | 13:13 |
sdake | tbh I'd like to deprecate config-internal | 13:13 |
ccrouch | thanks for looking at the etherpad SamYaple | 13:19 |
sdake | ccrouch we can turn that into a md in the docs dir | 13:20 |
sdake | probably design.md or something | 13:20 |
sdake | larsks does that approach sound amenaible to you? | 13:22 |
ccrouch | cool | 13:24 |
larsks | sdake: you can now 'pip install keystone-crux' | 13:26 |
sdake | larsks that was fast:) | 13:26 |
sdake | pbourke can you do me another solid and test that out in the container | 13:26 |
sdake | larsks did you use crux or some other solution and if so mind pasting a diff | 13:26 |
sdake | (just so I can learn) | 13:26 |
sdake | rather pbr | 13:26 |
larsks | I...what? | 13:26 |
larsks | Oh, no. | 13:26 |
pbourke | sdake: wasn't watching irc, has openstack-crux been superceded/ | 13:26 |
pbourke | ? | 13:27 |
sdake | pbourke right keystone-crux | 13:27 |
larsks | The problem was simply that 'requirements.txt' wasn't included in the generated sdist. | 13:27 |
sdake | larsks will maintain | 13:27 |
larsks | For files that are not included by default you need to create a MANIFEST.in file. | 13:27 |
sdake | so it needed to go in MANIFEST.in? | 13:27 |
larsks | Yup. | 13:27 |
pbourke | works fine | 13:27 |
sdake | ya i should have tried that | 13:27 |
pbourke | I'll update my original commit | 13:27 |
pbourke | thanks all | 13:27 |
sdake | ok i'll see if I can delete openstack-crux | 13:27 |
sdake | few minutes - in a conversation | 13:27 |
larsks | There is a big "DO NOT USE THIS BUTTON" thing at the bottom of the project details page for deleting a project. | 13:28 |
*** tfukushima has joined #kolla | 13:35 | |
openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Use the Glance Data container for ansible https://review.openstack.org/210417 | 13:37 |
sdake | working on it- in a conversation :) | 13:37 |
sdake | larsks ok pulled trigger on that | 13:38 |
sdake | thanks for your quick help :) | 13:39 |
*** jtriley has joined #kolla | 13:39 | |
openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Use the Glance Data container for ansible https://review.openstack.org/210417 | 13:39 |
larsks | Sure. And you know, if people have feature requests or prs, send them my way. | 13:40 |
sdake | larsks we are thinking of deprecating config-internal | 13:42 |
sdake | so we may drop keystone-crux entirely | 13:42 |
sdake | hard to say at this point | 13:42 |
sdake | depends on what tripleo wants to do | 13:42 |
sdake | its nothing wrong with crux just with crudini model | 13:42 |
larsks | Sure. | 13:42 |
openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Use the Glance Data container for ansible https://review.openstack.org/210417 | 13:42 |
sdake | yu already know problems with crudini model i don't think i have to explain it :) | 13:43 |
*** inc0_ has quit IRC | 13:45 | |
*** athomas has quit IRC | 13:55 | |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 13:55 |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 13:58 |
*** echoingumesh has joined #kolla | 14:06 | |
*** athomas has joined #kolla | 14:07 | |
*** diogogmt has joined #kolla | 14:10 | |
*** echoingumesh has quit IRC | 14:11 | |
*** diogogmt has quit IRC | 14:11 | |
*** diogogmt has joined #kolla | 14:13 | |
*** tfukushima has quit IRC | 14:17 | |
*** rhallisey has quit IRC | 14:17 | |
*** rhallisey has joined #kolla | 14:17 | |
*** tfukushima has joined #kolla | 14:18 | |
*** ccrouch has quit IRC | 14:19 | |
*** sdake_ has joined #kolla | 14:21 | |
sdake_ | samyaple you around? | 14:24 |
openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Use Nova Data container for ansible https://review.openstack.org/210463 | 14:24 |
*** sdake has quit IRC | 14:25 | |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Install correct crux in centos-source-base https://review.openstack.org/204977 | 14:27 |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Fix missing config in horizon source image https://review.openstack.org/209415 | 14:27 |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Add horizon source for oraclelinux https://review.openstack.org/209430 | 14:27 |
*** diogogmt has quit IRC | 14:28 | |
rhallisey | sdake_, I do sleep the containers | 14:28 |
rhallisey | I just use the existing data containers | 14:29 |
sdake_ | ya I think sam was anti that for some reason | 14:29 |
sdake_ | just -1'ed it so the conversation can happen before it ninja merges | 14:29 |
rhallisey | kk | 14:29 |
*** tfukushima has quit IRC | 14:32 | |
*** tfukushima has joined #kolla | 14:33 | |
*** diogogmt has joined #kolla | 14:33 | |
sdake_ | pbourke nice job on murano (and mick!) | 14:41 |
pbourke | sdake_: thanks, it sucks about horizon though | 14:41 |
sdake_ | which part - i missed that in the review | 14:41 |
pbourke | sdake_: its throwing some sort of angularjs error that I can't get to the bottom of. which means we can't really test murano | 14:42 |
sdake_ | fantastic! | 14:42 |
sdake_ | so we agreed to put this stuff in extras - that is one of the reasons | 14:42 |
pbourke | see note in the commit msg (https://review.openstack.org/#/c/209905/) | 14:42 |
sdake_ | but that depends on templating being finished first | 14:42 |
pbourke | there was some uncertainty about the tagging yesterday | 14:43 |
pbourke | are we still good with liberty2 in compose files? | 14:44 |
*** ccrouch has joined #kolla | 14:44 | |
sdake_ | for the moment yes pls | 14:45 |
sdake_ | people should build with --tag=liberty2 | 14:46 |
sdake_ | although i'm not sure if that is the correct answer | 14:46 |
sdake_ | or a better answer is to commit a change that changes master to "latest" again | 14:46 |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 14:46 |
sdake_ | pbourke can you file a bug to move liberty2 to latest and submit a patch | 14:49 |
sdake_ | it may end up being a waste of time - i just dont know | 14:49 |
sdake_ | i dont know what the best solution to that problem is, but if we get a review out the CR team can debate it in the review | 14:49 |
sdake_ | rhallisey can you review that murano patch pls | 14:49 |
rhallisey | ya just a minute | 14:50 |
pbourke | wait hang on | 14:51 |
pbourke | sdake_ and me are behind the times :P | 14:51 |
pbourke | https://review.openstack.org/#/c/210027/ | 14:51 |
pbourke | so murano should in fact be latest right | 14:52 |
*** dims_ is now known as dimsum__ | 14:53 | |
*** blahRus has joined #kolla | 14:53 | |
vbel | anyone can answer mariadb bootstrap/register in new model (after kolla-ansible proxy container had been merged) | 14:53 |
sdake_ | pbourke ya revert that liberty2 thing i told you to fix in murano and i'll reapprove | 14:53 |
sdake_ | sorry for missing that was traveling | 14:53 |
vbel | why do we run bootstrapping container twice - in bootstrap and in register.yml | 14:53 |
pbourke | na I missed it too | 14:53 |
pbourke | well I saw it fly by yesterday but forgot about it | 14:53 |
sdake_ | I do try to review every patch even once they have hit the repo but we have serious veloicyty nad its hard | 14:53 |
openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 14:54 |
sdake_ | the good news is it looks like the cr team is happy with the revert idea ;) | 14:54 |
sdake_ | rhallisey can you review that now ^^ :) | 14:55 |
rhallisey | y | 14:55 |
sdake_ | thanks | 14:55 |
sdake_ | would be nice if rdo could figure out ho wto package the big tent :) | 14:56 |
sdake_ | I think we need to have a blueprint for the extras idea | 14:56 |
sdake_ | we agreed that was the path forward at the midcycle as well | 14:56 |
sdake_ | so I'll file that | 14:56 |
sdake_ | vbel no idea how that works - sorry can't answer | 14:56 |
sdake_ | vbel samyaple is your man, he is usually in in a couple hours | 14:57 |
sdake_ | but he was not on lsat night so maybe he is taking a breather | 14:57 |
vbel | ok, will wait, thanks | 14:57 |
pbourke | vbel: will it not work the same way as keystone? | 14:59 |
vbel | I wonder why the bootstrap block was added via https://review.openstack.org/#/c/208451/13 | 15:00 |
pbourke | sorry you need to be more specific | 15:01 |
vbel | I wonder why the bootstrap block was added via https://review.openstack.org/#/c/208451/13 to https://review.openstack.org/#/c/208451/13/ansible/roles/mariadb/tasks/register.yml | 15:02 |
sdake_ | dummy question, does ha proxy depend on mysql? | 15:03 |
sdake_ | rhallisey you up for making a docker directory structure for rhel | 15:04 |
rhallisey | sure | 15:04 |
pbourke | vbel: good point | 15:04 |
pbourke | :/ | 15:04 |
sdake_ | cool file a blueprint and i'll approve | 15:04 |
vbel | sdake_: no | 15:05 |
sdake_ | rhallisey ^^ | 15:05 |
pbourke | sdake_: I dont think so, other than the fact it may have an entry for mysql | 15:05 |
sdake_ | vbel thanks | 15:05 |
sdake_ | i want to use haproxy+keepalived for the etcd work too | 15:05 |
sdake_ | i'm not quite sure how to do it | 15:05 |
sdake_ | but it neesd to be done | 15:06 |
pbourke | just checked, it does | 15:06 |
sdake_ | pbourke got a link to code | 15:06 |
pbourke | sdake_: I imagine you can add them after the fact | 15:06 |
pbourke | https://github.com/stackforge/kolla/blob/master/ansible/roles/haproxy/templates/haproxy.cfg.j2#L17-L27 | 15:06 |
sdake_ | the problem is keepalived and haproxy need to be deployed prior to the openstack deployment | 15:07 |
vbel | it does not require mysql/mariadb to be running, just config | 15:07 |
pbourke | it also seems like a soft dependency, as in its not going to fail to startup if mysql is missing | 15:07 |
pbourke | what vbel said :) | 15:07 |
sdake_ | so we could have a new role haproxy_deploy | 15:07 |
*** echoingumesh has joined #kolla | 15:07 | |
sdake_ | without that stuff in it | 15:07 |
vbel | it listens to mysql port and sends that to real mysql if the latter is present | 15:07 |
sdake_ | and only use that for the deploy host | 15:07 |
sdake_ | right | 15:08 |
vbel | yes | 15:08 |
pbourke | sdake_: may also help: https://github.com/stackforge/kolla/blob/master/etc/kolla/defaults.yml#L39-L45 | 15:08 |
sdake_ | pbourke what I'm thinking is we have a keepalived_deploy and haproxy_deploy and just restart them with new configs once the actual deploy occurs | 15:09 |
sdake_ | does oracle linux have packaging for etcd? | 15:10 |
sdake_ | I could probably use the centos version | 15:10 |
vbel | we will make etcd if necessary | 15:11 |
sdake_ | rhallisey please review and comment in the whiteboardon https://blueprints.launchpad.net/kolla/+spec/deploy-ha | 15:11 |
sdake_ | and its dependency | 15:11 |
*** echoingumesh has quit IRC | 15:12 | |
*** unicell1 has joined #kolla | 15:14 | |
*** unicell has quit IRC | 15:16 | |
sdake_ | ok guys we are deprecating config-internal - dprince acked it and hes the one doing the heavy lifting on integrating kolla and puppet (tripleo's CM tool) | 15:19 |
rhallisey | ya he helped a lot with the puppet work | 15:20 |
*** sdake_ is now known as sdake | 15:20 | |
sdake | wtb sam yaple | 15:23 |
sdake | samyaple come alive | 15:23 |
rhallisey | lol you keep saying his name it pings his phone | 15:23 |
rhallisey | he will never sleep! | 15:23 |
sdake | oh didn't know that | 15:32 |
sdake | usually he is awake right now | 15:32 |
sdake | i want him to look over my deploy ha bp | 15:32 |
sdake | and see if he has any better ideas | 15:33 |
sdake | config internal removal going to be a big job | 15:35 |
sdake | yay for 20 patch patchsets :) | 15:35 |
sdake | https://blueprints.launchpad.net/kolla/+spec/remove-config-internal | 15:37 |
*** echoingumesh has joined #kolla | 16:08 | |
*** echoingumesh has quit IRC | 16:13 | |
*** unicell1 has quit IRC | 16:17 | |
*** mstachow has quit IRC | 16:19 | |
*** echoingumesh has joined #kolla | 16:20 | |
*** echoingumesh has quit IRC | 16:20 | |
*** jasonsb has quit IRC | 16:20 | |
*** jasonsb has joined #kolla | 16:21 | |
*** sdake_ has joined #kolla | 16:22 | |
*** jasonsb has quit IRC | 16:26 | |
*** sdake has quit IRC | 16:26 | |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from Barbican https://review.openstack.org/210519 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from ceilometer https://review.openstack.org/210520 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from cinder https://review.openstack.org/210521 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from designate https://review.openstack.org/210522 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove docker-compose container https://review.openstack.org/210523 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from galera https://review.openstack.org/210524 | 16:26 |
openstackgerrit | Steven Dake proposed stackforge/kolla: Remove config-internal from glance https://review.openstack.org/210525 | 16:26 |
sdake_ | have a look | 16:27 |
sdake_ | I have to head to the dentist | 16:27 |
sdake_ | leave comments in irc ;) | 16:27 |
*** daneyon_ has joined #kolla | 16:39 | |
*** unicell has joined #kolla | 16:40 | |
*** daneyon has quit IRC | 16:42 | |
pbourke | if someone has a few mins to spare and could test if they experience this bug it would be good help https://bugs.launchpad.net/kolla/+bug/1482196 | 16:54 |
openstack | Launchpad bug 1482196 in kolla "Horizon dashboard not loading properly, can't click on anything" [Undecided,New] | 16:54 |
pbourke | sdake_: the config-internal removals look ok but I won't +2 such a big change on a friday evening ;) | 16:55 |
pbourke | will let it sit for the weekend, have another look on monday to be sure nothing else comes up | 16:55 |
pbourke | have a good one all | 16:56 |
vbel | anyone can run ansible with latest code? I see that we expect 'localhost | SUCCESS >>' from kolla-ansible container but ansible actually returns 'localhost | success >>' | 16:57 |
vbel | and therefore all such remote ansible things fail | 16:57 |
vbel | if you run yours ansible with -vvvvv you can see what is returned | 16:58 |
vbel | ansible]$ grep -rl ' SUCCESS =>' . | 17:03 |
vbel | ./roles/glance/tasks/register.yml | 17:03 |
vbel | ./roles/mariadb/tasks/register.yml | 17:03 |
vbel | ./roles/neutron/tasks/register.yml | 17:03 |
vbel | ./roles/nova/tasks/register.yml | 17:03 |
vbel | ./roles/bootstrap.yml | 17:03 |
vbel | wonder how it is supposed to work | 17:04 |
*** bmace has quit IRC | 17:08 | |
*** ccrouch has quit IRC | 17:09 | |
*** jasonsb has joined #kolla | 17:12 | |
*** mfalatic_ has quit IRC | 17:16 | |
*** bmace has joined #kolla | 17:17 | |
SamYaple | sdake_: ping | 17:17 |
SamYaple | vbel: rebuild your ansible container | 17:18 |
SamYaple | vbel: you are running with an old ansible container most likely | 17:18 |
SamYaple | the different between the two is the difference between 1.x and 2.x ansible return codes | 17:19 |
*** tfukushima has quit IRC | 17:22 | |
*** akwasnie has quit IRC | 17:23 | |
*** tfukushima has joined #kolla | 17:23 | |
*** tfukushima has quit IRC | 17:25 | |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for zaqar https://review.openstack.org/210546 | 17:26 |
*** daneyon_ has quit IRC | 17:26 | |
*** daneyon has joined #kolla | 17:26 | |
*** ccrouch has joined #kolla | 17:27 | |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for swift https://review.openstack.org/210548 | 17:31 |
*** athomas has quit IRC | 17:34 | |
*** achanda has joined #kolla | 17:44 | |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for rabbitmq https://review.openstack.org/210552 | 17:45 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for swift https://review.openstack.org/210548 | 17:49 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for zaqar https://review.openstack.org/210546 | 17:49 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for openvswitch https://review.openstack.org/210554 | 17:53 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for nova https://review.openstack.org/210561 | 18:07 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for neutron https://review.openstack.org/210566 | 18:21 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for memcached https://review.openstack.org/210567 | 18:24 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for neutron https://review.openstack.org/210566 | 18:24 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for keystone https://review.openstack.org/210571 | 18:33 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for Magnum https://review.openstack.org/210572 | 18:34 |
openstackgerrit | Merged stackforge/kolla: Add check_script for haproxy in keepalived https://review.openstack.org/209463 | 18:38 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for keepalived https://review.openstack.org/210574 | 18:38 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for Horizon https://review.openstack.org/210577 | 18:42 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for haproxy https://review.openstack.org/210578 | 18:44 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for gnocchi https://review.openstack.org/210580 | 18:49 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for Heat https://review.openstack.org/210584 | 18:54 |
SamYaple | alright thats the majority of config-internal.sh removal | 18:58 |
SamYaple | one those patches start landing we can worry about cleaning up the remaing references | 18:59 |
*** jpeeler has quit IRC | 19:03 | |
*** jpeeler has joined #kolla | 19:05 | |
*** jpeeler has joined #kolla | 19:05 | |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for nova https://review.openstack.org/210561 | 19:08 |
*** banix has joined #kolla | 19:11 | |
SamYaple | rhallisey TripleO will still be using docker-compose right? | 19:12 |
jpeeler | it doesn't have to | 19:13 |
SamYaple | well i was asking morein reference to the requirements.txt | 19:14 |
SamYaple | im looking at the DinD gate thing | 19:14 |
SamYaple | if we dont gate with compose, i would prefer to not have it there since it has wierd restrictions on docker-py | 19:14 |
*** echoingumesh has joined #kolla | 19:19 | |
*** mfalatic has joined #kolla | 19:19 | |
*** daneyon_ has joined #kolla | 19:19 | |
*** daneyon has quit IRC | 19:22 | |
SamYaple | coolsvap: ping | 19:25 |
rhallisey | SamYaple, I'm not sure just yet | 19:36 |
*** banix has quit IRC | 19:41 | |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove config-internal for neutron https://review.openstack.org/210566 | 19:44 |
openstackgerrit | Sam Yaple proposed stackforge/kolla: Remove Ansible abstraction for Keystone https://review.openstack.org/210041 | 19:48 |
*** echoingumesh has quit IRC | 19:53 | |
*** echoingumesh has joined #kolla | 19:54 | |
*** echoingumesh has quit IRC | 19:58 | |
*** banix has joined #kolla | 20:00 | |
*** jtriley has quit IRC | 20:06 | |
*** jtriley has joined #kolla | 20:08 | |
*** achanda has quit IRC | 20:08 | |
*** echoingumesh has joined #kolla | 20:12 | |
sdake_ | samyaple you around? | 20:33 |
sdake_ | damn samyaple jacked all my commits ;-) | 20:34 |
sdake_ | samyaple wtb comments on ha deploy idea | 20:34 |
*** jruano has quit IRC | 20:45 | |
SamYaple | sdake_: ping | 20:45 |
SamYaple | sdake_: no ive just wanted to get rid of config-internal for a while | 20:45 |
SamYaple | i was highly motivated | 20:45 |
SamYaple | i am the highest commiter at RAX now | 20:46 |
SamYaple | so thats cool | 20:46 |
sdake_ | haha grats bro | 20:46 |
sdake_ | I would have finished the job but had 3 hour torture session at dentist | 20:46 |
sdake_ | so can you evaluate the ha deploy idea plz | 20:47 |
sdake_ | or have you already | 20:47 |
sdake_ | i'd like to see if you can come up with something better | 20:47 |
SamYaple | i have questions about it | 20:47 |
sdake_ | cool lets see if we can make it consistent :) | 20:47 |
sdake_ | shoot | 20:47 |
sdake_ | come up with something better->improve the idea | 20:47 |
SamYaple | so my idea was basic, since the passwords.yml can be encrypted in an ansible manner, why not tar and ship the /etc/kolla directory with date and timestamp each run? | 20:48 |
sdake_ | one thing that will complicate the process is etcd needs a haproxy in front of it | 20:48 |
SamYaple | that was my initial idea anyway | 20:48 |
SamYaple | yea i dont like having etcd | 20:48 |
SamYaple | at all | 20:48 |
SamYaple | i dont like external depends like that | 20:48 |
sdake_ | its totally optional | 20:48 |
SamYaple | but i still want an automated way to do this backup/versioning | 20:49 |
sdake_ | you can use backup/restore | 20:49 |
SamYaple | so do we have two in the repo? | 20:49 |
sdake_ | two which? methods of ha? | 20:49 |
sdake_ | the cool thing about etcd is we only need to store the data and retrieve it - so the ha is pretty straightforward | 20:49 |
SamYaple | im confused here. what is the end goal of that blueprint | 20:49 |
sdake_ | to provide a deploy host in case the disk goes byebye in a redundant fashion | 20:50 |
sdake_ | there are two blueprints | 20:50 |
sdake_ | it might help to read both if you haven't | 20:50 |
SamYaple | and a tar distributed around is not good enough? | 20:50 |
SamYaple | i only say the linked one | 20:50 |
sdake_ | the other is a dep in it | 20:51 |
sdake_ | if you look at the bottom of the blueprint | 20:51 |
sdake_ | it explains a workflow for deploying openstack involving ha deployment hosts | 20:51 |
sdake_ | i am not super hot on copying around tars to some place | 20:51 |
sdake_ | not sure why | 20:51 |
sdake_ | guess seems error prone | 20:52 |
sdake_ | but obviously less complex | 20:52 |
*** sdake_ is now known as sdake | 20:52 | |
SamYaple | but you want to introduce etcd? | 20:52 |
SamYaple | its not error prone at all | 20:52 |
sdake | or galera would be a differnent optoin | 20:52 |
sdake | or tar could be a third option via sshd | 20:52 |
sdake | basically pluggable protocol handlers | 20:52 |
SamYaple | im strongly opposed to using etcd. it is used no where else | 20:53 |
sdake | so tar work in this model too | 20:53 |
sdake | you mean in openstack not used elsewhere or in general? | 20:53 |
SamYaple | man i really really dont like this scope creep | 20:53 |
sdake | we dont have to do in liberty but people want ha deploy hosts vs one deploy host where bad things happen and the deployment is fubared | 20:53 |
SamYaple | well people can have that, that has nothing to do with kolla. why is kolla solving for that? | 20:54 |
SamYaple | i mean ceph-deploy doesnt solve for that issue, because its outside of scope | 20:54 |
sdake | it involves the plugin that does the merging of the configs | 20:54 |
sdake | ceph guys want to use etcd for configs too | 20:54 |
SamYaple | how so? | 20:54 |
SamYaple | but they dont | 20:54 |
SamYaple | because push back | 20:55 |
sdake | agree they dont but that is their stated intent | 20:55 |
SamYaple | why are we talking about modifying merge_configs? | 20:55 |
sdake | to read direclty from etcd rtaher then file:// | 20:55 |
sdake | ther would be etcd://ip/path | 20:55 |
SamYaple | ugh no thats horrible | 20:55 |
SamYaple | this is major scope creep in my opinion | 20:56 |
sdake | that way configs are basically stored in a database | 20:56 |
SamYaple | we are doing infrastructure for a product now | 20:56 |
sdake | but optional, file:// would be default | 20:56 |
sdake | you mean with that addition? | 20:56 |
SamYaple | this is a huge layer of complexity | 20:56 |
sdake | file or etcd or galera or whatever backend would go in the ansible config | 20:57 |
SamYaple | yea but now its on kolla to setup and maintain doing all that | 20:57 |
sdake | seems pretty simple - deploy an etcd container | 20:57 |
SamYaple | thats huge scope creep | 20:58 |
sdake | well its on us to figure out how to ha-ify deployment | 20:58 |
sdake | not the deployed openstackbut also the deployment process | 20:58 |
SamYaple | i disagree strongly. that is not on us | 20:58 |
SamYaple | but a tarball of the current configs is acceptable | 20:59 |
SamYaple | thats simple enough and secure with ansible vault | 20:59 |
sdake | would you agree people will want this sort of functionality? | 20:59 |
SamYaple | in a product they build, yes | 20:59 |
sdake | explain valut to me | 20:59 |
SamYaple | it just encrypts the passwords/global.yml | 20:59 |
SamYaple | so the tarball can be on all the host securely | 20:59 |
sdake | the downside of vault is it is propreitary | 20:59 |
SamYaple | with etcd you are moving raw plaintext passwords everywhere | 20:59 |
SamYaple | well this is ansible... | 21:00 |
sdake | you can secure with thttps | 21:00 |
SamYaple | merge_config is all ansible.. | 21:00 |
SamYaple | this doesnt solve for anyhing but ansible | 21:00 |
SamYaple | if someone wnats to do this for anythign else it wouldnt work | 21:00 |
SamYaple | and im against adding it to the ansible implementation | 21:00 |
SamYaple | this is suppose to be basic reference (but working) anyway | 21:00 |
SamYaple | we arent writting a product for a company, thats the companies job | 21:01 |
sdake | i want that product to be entirely open source ;-) | 21:01 |
SamYaple | you cant fix all business needs with open-source project | 21:01 |
SamYaple | our ansible implementation is lean and clean. i really want to keep it that way | 21:02 |
sdake | tripleo tackles ha of the overcloud - its mandatory | 21:02 |
SamYaple | start throwing in etcd does not make it clean and understandable | 21:02 |
sdake | I was actually thinking of a separate ansible directroy for the etcd stuff | 21:02 |
SamYaple | i will not ever be on board with this idea | 21:02 |
SamYaple | we have a clean small base right now | 21:03 |
SamYaple | i do not want scope creep | 21:03 |
SamYaple | the tarball shipping around prevents config loss | 21:03 |
sdake | ya thats backup/restore model | 21:03 |
sdake | not an ha model | 21:03 |
SamYaple | yea but we deploy openstack, we dont deploy deployment tools | 21:04 |
sdake | i agree backup restore is one approach | 21:04 |
sdake | i wasn't suggesting deploying deployment tools | 21:04 |
sdake | just infra to support the deployment tool we have | 21:05 |
SamYaple | here is the problem i have | 21:06 |
SamYaple | we just went through a bunch of effort to containerize dep | 21:07 |
SamYaple | deps* | 21:07 |
SamYaple | now you are talking about adding things to the host | 21:07 |
SamYaple | you _must_ have ansible installed on all nodes to have deploy ha | 21:07 |
SamYaple | i disagree with messing with the host | 21:07 |
SamYaple | we just removed all deps except docker and dokcer-py | 21:07 |
SamYaple | now you want ot add more | 21:07 |
sdake | in containers | 21:09 |
sdake | in the control nodes | 21:09 |
sdake | not on the deploy hosts | 21:09 |
SamYaple | so deploy hosts are a seperate thing now? | 21:09 |
*** achanda has joined #kolla | 21:10 | |
SamYaple | i mean we could have just had a dependacies node then | 21:10 |
SamYaple | the point was to not touch the host automatically | 21:10 |
SamYaple | the user has to install ansible and clone kolla to deploy | 21:10 |
SamYaple | now we are tlaking automatically adding that to nodes | 21:10 |
sdake | to the control nodes | 21:10 |
SamYaple | any nodes | 21:10 |
sdake | which we already ad a shit ton of stuff to anyway | 21:10 |
SamYaple | no we dont | 21:11 |
*** rhallisey has quit IRC | 21:11 | |
SamYaple | we dont add anything | 21:11 |
sdake | ya a whole lot of containers | 21:11 |
SamYaple | containers are different | 21:11 |
sdake | oh you missed i want etcd in container | 21:11 |
SamYaple | unless you are saying a kolla-deploy container with that container having ansible | 21:11 |
sdake | so its just another thing that docker deploys | 21:11 |
sdake | not talking about adding dependencies in the root filesystem | 21:11 |
SamYaple | right, but that is not ha since youll need access to kolla and ansible somewhere | 21:12 |
SamYaple | same issue | 21:12 |
sdake | like via yum etc | 21:12 |
sdake | ya you could run ansible from any node point it at the etcd ip and your good to rock | 21:12 |
sdake | rather kolla | 21:12 |
SamYaple | but where would ansible get the initial configs to point to etcd? | 21:13 |
sdake | so there is a pull and push operation essentially pull downloads the config to your local etc dir push pushes to etcd | 21:14 |
sdake | you make changes, push and deploy | 21:14 |
sdake | covered in the other blueprint | 21:14 |
SamYaple | i know how it works, but you still need initial configuration | 21:14 |
SamYaple | so say your deploy node goes down | 21:14 |
SamYaple | then you have to run anisble somewhere else | 21:14 |
SamYaple | you still have to configure kolla to point to etcd | 21:14 |
*** jtriley has quit IRC | 21:15 | |
sdake | agree you do have to ocnfigure kolla to point to etcd | 21:15 |
SamYaple | still requires configuration, would prefer a restore from a tarball then i dont have any configuration | 21:15 |
sdake | i wanted to ask how we could do that | 21:15 |
sdake | possibly without a config file | 21:15 |
sdake | if thats even possible with ansible | 21:15 |
SamYaple | by having lots of feature creep? | 21:15 |
sdake | i am talking about a compeltely new "ansible_deploy" directory | 21:16 |
sdake | that has a completely new ansible deploy setup stuff in it | 21:16 |
sdake | but the regular ansible dir merge configs can read from etcd or whatever | 21:16 |
sdake | read the other blueprint real quick plz | 21:17 |
SamYaple | i alreayd have | 21:17 |
sdake | ok | 21:17 |
SamYaple | i am strongly against all of this. it is just feature creep and not a magic bullet | 21:18 |
sdake | basically the thing in kolla that is persistent is all in /etc so if we can put that anywhere (like a db for example) then we can run kolla from anywhere that can connect to the db (vs moving tarballsaround) | 21:18 |
SamYaple | the _only_ thing i might be ok with is a kolla_deploy container on all the host/group of host that has the configs | 21:18 |
SamYaple | youre trying to build infrastructure i feel has no place in kolla | 21:19 |
sdake | expand on that I dont grok (3 hr dental drilling sesion) | 21:19 |
sdake | say the config-internal, did you cover all the ones i didn't finish | 21:19 |
SamYaple | yea | 21:20 |
sdake | what would a kolla deploy container do exactly? | 21:20 |
SamYaple | the end goal here is to have the files in more than one location, correct? | 21:20 |
sdake | the end goal is to have the files in one location but replicated for resiliency | 21:20 |
SamYaple | how is that different than what i said? | 21:21 |
SamYaple | anyway | 21:21 |
sdake | you said more then one location I said one location ;) | 21:21 |
SamYaple | this still has a chicken and egg situation | 21:21 |
sdake | which is what | 21:22 |
SamYaple | how do you do that before you have the other containers? | 21:22 |
SamYaple | you are backing up configs about contaienrs that are running in running containers | 21:22 |
SamYaple | in the end youre still screwed if all your contaienrs go down | 21:22 |
sdake | you mean how do you deploy the first time? | 21:22 |
sdake | agree, its an ha solution not a backup solution | 21:23 |
sdake | backup solution is tar -> copy offsite | 21:23 |
sdake | in ha if alll your nodes go down, your down ;) | 21:23 |
SamYaple | so it sounds like we both understand each other. the difference is i do not agree with this massive feature creep | 21:23 |
SamYaple | you still need a backup because youll need to configure where etcd is | 21:24 |
SamYaple | this add complication without solving the issue its meant to solve | 21:24 |
sdake | its meant to solve an availabliity issue - wihch is when the deploy host goes down, you can no longer deploy | 21:24 |
SamYaple | but it doesnt | 21:24 |
sdake | if there is replicated store, avaiability goes up | 21:24 |
SamYaple | because then you have to configure a new node | 21:25 |
SamYaple | and know how to configure it | 21:25 |
sdake | or you could configure multiple deploy hosts simulatenously | 21:25 |
SamYaple | true, but you could do that with any number of methods | 21:25 |
sdake | this is more the model i was thinking people would use | 21:25 |
SamYaple | my point is they are all outside kolla scope | 21:26 |
*** banix has quit IRC | 21:26 | |
sdake | i hear you think they are outside of scope - i think they are inside scope but possibly further down the road | 21:27 |
sdake | i have had several people ask me how do we do ha of the deploy host | 21:27 |
sdake | pretty much everyone asks | 21:27 |
SamYaple | if you can make it fully solve the issue then sure | 21:27 |
sdake | and i go "we dont" | 21:27 |
SamYaple | but it doesnt | 21:27 |
SamYaple | we dont prevent it either | 21:28 |
SamYaple | we do lots of "we dont do it, but you can do it" | 21:28 |
SamYaple | the point is flexibility | 21:28 |
sdake | what could be more flexible then readin the config from any data store rather then the local filesystem only? | 21:28 |
SamYaple | i never said that wasnt flexible, i said thats not in scope *but* we dont prevent it | 21:29 |
SamYaple | i disagree with kolla having this functionality included | 21:29 |
sdake | we dont enable it either | 21:29 |
SamYaple | i also disagree with blocking it | 21:29 |
sdake | at the current mometn there is no way to deploy kolla in a ha way on the deployment host | 21:30 |
SamYaple | lies | 21:30 |
SamYaple | NFS share | 21:30 |
SamYaple | drbd | 21:30 |
SamYaple | two ways right there | 21:30 |
sdake | nfs share is not suitable - get real | 21:30 |
sdake | drdb , not familiar | 21:30 |
SamYaple | dont tell me that | 21:30 |
SamYaple | lots of people have NetAPP | 21:30 |
SamYaple | my point is this isnt our issue to solve | 21:30 |
SamYaple | both of those ways work | 21:31 |
SamYaple | there are plenty of other ways tat work | 21:31 |
sdake | i originally thoguth of nfs share | 21:31 |
SamYaple | not our issue to solve | 21:31 |
sdake | but the only way that works is if someone has a ha nfs system | 21:31 |
SamYaple | right, not our issue to solve | 21:31 |
sdake | ha nfs is a huge pita to setup | 21:31 |
SamYaple | i agree | 21:31 |
SamYaple | i dont recommend it | 21:31 |
SamYaple | but the fact is if a shop already has a NetAPP, this would be perfect for them | 21:32 |
SamYaple | if they are big uses for DRBD, that would also work | 21:32 |
SamYaple | not kollas issue to solve | 21:32 |
sdake | drdb isnt filesystem based is it? | 21:32 |
SamYaple | its that horrible distributed block device thing | 21:32 |
sdake | agree, its block storage so thats no good - it only allows one writer | 21:33 |
SamYaple | not true | 21:33 |
SamYaple | its drbd | 21:33 |
sdake | you can put a filesystem on it? | 21:33 |
SamYaple | yes | 21:33 |
SamYaple | it does distributed locking | 21:33 |
SamYaple | im not here to advocate for it | 21:33 |
sdake | and the filesystem can be writte nby both nodes? | 21:33 |
SamYaple | my point is right now there are several ways to solve it | 21:33 |
SamYaple | i dont want to include a way | 21:34 |
sdake | so essentiallly yo udont want to have a way to solve ha of the deployment nodes | 21:34 |
SamYaple | i dont want to include a way. i want to have many options and suggestions | 21:34 |
sdake | the only viable way today realistically is netapp nfs sharing | 21:35 |
sdake | or some other ha hardware nfs system | 21:35 |
SamYaple | wrong | 21:35 |
SamYaple | gluesterfs cephfs are both considered stable | 21:35 |
bmace | i don't think it would be terrible to have a default / tested way to do it, and allow for other people to use different technologies for it if they want. | 21:36 |
SamYaple | drbd does actually work | 21:36 |
sdake | i dont think drdb does filesystem sharing like yo uthink it does | 21:36 |
sdake | anyway drdb was a huge clusterfuck in teh cluster space :) | 21:36 |
SamYaple | i agree we can move on | 21:36 |
SamYaple | my point is there are solutions out there and its out of scope for kolla to solve infra problems | 21:37 |
sdake | so glusterfs cephfs are also viable options | 21:37 |
sdake | the proposal is optional in that if someone wanted to use a different way as bmace suggested it would be on them to sort out a way | 21:37 |
SamYaple | but by introducing a huge dependacies into ansible | 21:38 |
SamYaple | we would have to maintain a module that now deals with etcd | 21:38 |
sdake | which dependencies? | 21:38 |
sdake | its a simple curl operation | 21:38 |
SamYaple | which would be implemented how? | 21:38 |
sdake | it would add a http dependency of some sort | 21:38 |
SamYaple | ugh i dont want to go down this conversational path | 21:39 |
SamYaple | i dont need to agrue the semeantics there. there are options. i strongly feel this is out of scope for kolla | 21:39 |
SamYaple | it was brought up around the mission statement time and i stated it then too | 21:39 |
SamYaple | we are not here to solve infra problems | 21:39 |
sdake | at some point we may want to tackle bare metal deployment | 21:40 |
bmace | we deploy ha openstack.. we are directly fixing infra problems | 21:40 |
sdake | i imagine we will have the same argument ;) | 21:40 |
bmace | and the solution as a whole isn't really ha if the ansible node can go down and you are completely hosed. | 21:40 |
SamYaple | bmace: but thats my point, you would still be until you reconfigure a node | 21:40 |
sdake | you setup two nodes.. | 21:41 |
sdake | to start with | 21:41 |
sdake | no reconfigure | 21:41 |
SamYaple | which you can do now | 21:41 |
sdake | yes but the configs get out of sync! | 21:41 |
SamYaple | not with a shared backend | 21:41 |
SamYaple | already solved issue | 21:41 |
SamYaple | this is out of scope for kolla | 21:41 |
sdake | yes if someone deploys ceph or gluster or nfs in a ha fashion, and mounts /etc/kolla from it, its solved | 21:42 |
SamYaple | so what youre saying is, there are solutions? why are we trying to solve all issues with kolla | 21:42 |
SamYaple | this is the very definition of project scope creep | 21:42 |
sdake | scope expansion isn't a bad thing if it increases the appeal of the project to the userbase | 21:43 |
SamYaple | while i agree, i disagree this is something we should be expanding into | 21:43 |
sdake | if its optional, what is the harm? | 21:43 |
sdake | just like nfs/ceph/gluter are optional | 21:44 |
SamYaple | added code to maintain and you are tainting the merge_config module which is super tiny and easily maintainable at the moment | 21:44 |
sdake | tainting is a strong word there ;) | 21:44 |
SamYaple | fair point | 21:44 |
sdake | i agree it increases the complexity of merge_config but its really simple and adding other protocol handlers would be really simple as well | 21:45 |
SamYaple | but do we want to be in the position of maintaining ansible modules? we agreed we wouldn't | 21:45 |
sdake | i think its fine to maintain some ansible modules that are custom to our solution | 21:45 |
sdake | which that one fits that definition | 21:45 |
SamYaple | right if they are tiny, which they are | 21:45 |
SamYaple | they wont be | 21:45 |
SamYaple | i mean you dont really know whats involved. the playbooks wont work both ways | 21:46 |
sdake | the playbooks take an option "protocol_handler" or call it what you like | 21:46 |
SamYaple | we use the temlpating module at the moment and that has no etcd support | 21:46 |
sdake | file:// | 21:46 |
sdake | or etcd://ip | 21:46 |
SamYaple | we still have | 21:46 |
sdake | which tempalting module | 21:46 |
SamYaple | the module called template | 21:47 |
sdake | what does that do | 21:47 |
sdake | not familiar with that one | 21:47 |
SamYaple | takes templates and fills them out | 21:47 |
sdake | that creates /usr/share/kolla/templates or something? | 21:47 |
SamYaple | it basically all we use | 21:47 |
SamYaple | yea that sounds right | 21:47 |
sdake | i am not talking about modifying that | 21:47 |
sdake | only merge configs | 21:48 |
sdake | mege configs takea a bunch of inputs and writes a file to the filesystem | 21:48 |
sdake | i like that model | 21:48 |
sdake | what i prpose is reading th einputs from various sources other then file:// | 21:48 |
SamYaple | i dont hink you understand how this is going to work then | 21:48 |
sdake | once an ansible deploy is in progress, if it fails because the machine dies, you gotta start over ;) | 21:48 |
SamYaple | in your scenario how do files get into etcd | 21:49 |
sdake | your right I dont have any idea exactly how this would work | 21:49 |
SamYaple | those templat files would need to go _into_ etcd | 21:49 |
sdake | possibly a tool on the host which copies them to etcd | 21:49 |
sdake | no the template files stay on the local host - since they are only used during deployment | 21:49 |
SamYaple | do you not see how this is complicated and anti-kolla? | 21:49 |
sdake | on the local filesystem | 21:49 |
SamYaple | no they dont | 21:50 |
*** sdake_ has joined #kolla | 21:51 | |
sdake_ | got dced | 21:51 |
sdake_ | I see it adds minimal complication | 21:52 |
sdake_ | and is optional with filesystem based deefaults | 21:52 |
SamYaple | you are wrong | 21:52 |
SamYaple | it adds lots of complication | 21:52 |
SamYaple | and its not like ansible has good if then else logic | 21:52 |
sdake_ | why do we need conditional logic to do this? | 21:52 |
sdake_ | I agree btw ansible conditionals no bueno :) | 21:53 |
SamYaple | the flow would have to be that the templated files would need to come from etcd | 21:54 |
SamYaple | this isn't just merge_config modifictation | 21:54 |
sdake_ | no, just the etc files | 21:54 |
sdake_ | template files are created by etc files correct? | 21:54 |
SamYaple | what do you think templates them? | 21:54 |
SamYaple | ansible | 21:54 |
SamYaple | template module has not etcd support | 21:54 |
sdake_ | ok so explain flow to me - perhaps I dont understand well enough | 21:54 |
*** sdake has quit IRC | 21:55 | |
*** bmace has quit IRC | 21:55 | |
sdake_ | I thought it was ansible -> merge configs outputs to template files -> template files output to /opt/kolla on the nodes | 21:55 |
SamYaple | even if this was easy to do, i would still be against it since i view it as a problem kolla shouldnt be solving for | 21:55 |
SamYaple | ansible templates the stuff from /etc/kolla on the deploy host to the dest node | 21:55 |
SamYaple | then merge config runs | 21:55 |
*** bmace has joined #kolla | 21:55 | |
SamYaple | that saves it to the file location on the dest node | 21:55 |
SamYaple | you would have to fetch files from etcd to template them | 21:56 |
sdake_ | i see, so merge_configs isn't the correct integration point is what your saying | 21:56 |
sdake_ | i thought merge configs ran on the local node | 21:56 |
sdake_ | (on the deploy host) | 21:56 |
SamYaple | well it could be if we really expand it which is what im against | 21:56 |
SamYaple | no ansible code runs on hte deploy host (exception of local_action's) | 21:57 |
sdake_ | so i am also against having each deploy node connect to etcd | 21:57 |
sdake_ | rathe reach deployed node | 21:57 |
SamYaple | oh well then this will never work | 21:57 |
sdake_ | how do the config files get into the deployed host? | 21:57 |
sdake_ | from the deployment host | 21:57 |
SamYaple | before we continue | 21:57 |
SamYaple | can se standardize on names for 'deploy host' and 'destination host' | 21:58 |
SamYaple | what are we calling them? | 21:58 |
sdake_ | i could see why yo uwould be bent out of shape about each deployed node reading etcd | 21:58 |
sdake_ | i dont ike that either | 21:58 |
SamYaple | me either | 21:58 |
sdake_ | sounds good | 21:58 |
SamYaple | ok all ansible code will execute on the 'destination host' | 21:58 |
SamYaple | always | 21:58 |
SamYaple | thats kinda ansibles deal | 21:58 |
sdake_ | deploy host = wher ansible runs = destination host = where containers run in general | 21:58 |
SamYaple | yes | 21:58 |
SamYaple | ansible installed on deploy host | 21:58 |
SamYaple | it actually executtes code on 'desitination host' | 21:59 |
sdake_ | how does it get the etc configuration files? | 21:59 |
sdake_ | from the deploy host | 21:59 |
SamYaple | ansible copies the file from the deploy host to the destination host. before it writes out the file to the destniation host it passes it through the jinja2 templaing filter | 21:59 |
SamYaple | thats done for each file (6-7 files) | 22:00 |
sdake_ | ok | 22:00 |
sdake_ | how does that happen, ansible has some built in copy function? | 22:00 |
SamYaple | template | 22:00 |
SamYaple | https://github.com/stackforge/kolla/blob/master/ansible/roles/config.yml#L13 | 22:00 |
sdake_ | is that a built in module? | 22:00 |
SamYaple | yes | 22:00 |
sdake_ | hmm | 22:01 |
sdake_ | well that sucks | 22:01 |
SamYaple | well thats ansible | 22:01 |
sdake_ | i guess my idea wont work | 22:01 |
SamYaple | well thats my point | 22:01 |
SamYaple | it _can_ | 22:01 |
SamYaple | its just lots of scope creep | 22:01 |
sdake_ | template is some kind of ssh built in module to ansible? | 22:02 |
SamYaple | template is pretty core to ansible | 22:02 |
SamYaple | ssh im not so sure about | 22:02 |
SamYaple | what do you mean | 22:02 |
sdake_ | i am curious how template works | 22:02 |
SamYaple | same way all other ansible code works | 22:03 |
sdake_ | becuase your right template would hve to have th etcd integration for this to work | 22:03 |
SamYaple | which means really building template into merge_configs which i am strongly against | 22:03 |
SamYaple | i think thats way to much creep | 22:03 |
sdake_ | nah it just means having a template which can read from etcd, but I suspect that may be hard if its a built in module | 22:04 |
SamYaple | how do you modify the files in etcd? | 22:04 |
sdake_ | because somehow it copies from the local host | 22:04 |
sdake_ | python script loads new config from /etc | 22:04 |
sdake_ | like 5 lines of python | 22:04 |
sdake_ | really simple | 22:05 |
SamYaple | so external to ansible stuff :/ | 22:05 |
SamYaple | that we were also against | 22:05 |
ccrouch | so now we're in the big tent, when are we moving under the openstack namespace in github? | 22:05 |
SamYaple | ccrouch: there is a review for it | 22:05 |
sdake_ | ccrouch there is a review in progress for that | 22:05 |
ccrouch | ah ha, awesome | 22:05 |
sdake_ | but when - i dont know - depends on the next infra move process | 22:05 |
SamYaple | ccrouch: https://review.openstack.org/#/c/209343/ | 22:05 |
sdake_ | i haven't got an answer on that yet | 22:05 |
ccrouch | thx | 22:06 |
SamYaple | sdake_: youll need to fix up all your config-internal patches. i left reviews for errors, but youll need to run the min-env.sh script | 22:09 |
SamYaple | i figure once the docker container code is removed, we can be a bit rougher with the scripts | 22:09 |
SamYaple | but well play nice with them until that time | 22:09 |
sdake_ | ccrouch according to clarkb ther eis no proejct move scheduled | 22:10 |
sdake_ | but committers will still get free atc passes | 22:10 |
sdake_ | because the atc passes are based upon the governance repo -ag a in based upon clarkb | 22:11 |
SamYaple | wasn't the cutoff on that a while ago? | 22:11 |
sdake_ | typically l3 is the cutoff | 22:11 |
SamYaple | ah ok | 22:11 |
SamYaple | maybe they just sent me teh AC pass a while ago | 22:11 |
SamYaple | that might be what im thinking | 22:11 |
sdake_ | rather the 3rd milestone | 22:11 |
sdake_ | ya they send em out along the way | 22:11 |
sdake_ | but they get people in that last milestone typicallly in the past | 22:11 |
SamYaple | cool | 22:12 |
sdake_ | samyaple i'm likely to rebase a delete gen-min-env removal patch in front of my patches ;) | 22:13 |
SamYaple | you really should break those up for speed of merge reasons | 22:14 |
sdake_ | well bmace given the new understanding of how ansible works with the templat emodule, it may not be possible to do what i want | 22:14 |
sdake_ | is there a way to remove the deps? | 22:14 |
SamYaple | rebase against master? | 22:15 |
sdake_ | you mean each individual patch in a branchz? | 22:15 |
SamYaple | i mean you arent doing dependant changes | 22:15 |
SamYaple | ha yea | 22:15 |
SamYaple | i can hear the groan from here | 22:15 |
sdake_ | lol | 22:16 |
sdake_ | ya noot doing that | 22:16 |
sdake_ | too lazy :) | 22:16 |
SamYaple | eh i did it because i wanted them to not be dependant | 22:16 |
sdake_ | i was actually going to do a whole stream and then have a proper removal of compose at the end with modifications to the docs | 22:16 |
sdake_ | so it wa going to be dependent when i finished | 22:17 |
sdake_ | atleast thatwas my well laid plan until you ninjaed myw ork :) | 22:17 |
SamYaple | lies! i saw all your patches were bad and it was going to take forever | 22:17 |
SamYaple | but truthfully, I just wanted to do some mindless work for a while | 22:18 |
SamYaple | im tired of thinking | 22:18 |
*** dims_ has joined #kolla | 22:18 | |
sdake_ | ya me too | 22:18 |
sdake_ | at magnum midcycle constant brainstorming hard on the brain | 22:19 |
sdake_ | so samyaple heres a q, is it possible to run an ansible module on the deployment host? | 22:19 |
sdake_ | that template module does it | 22:19 |
sdake_ | is that only because its built in? | 22:20 |
sdake_ | or is template even a module | 22:20 |
SamYaple | that template module does what? | 22:20 |
SamYaple | it doesnt run on the deployment host | 22:20 |
sdake_ | the one you linked earlier that copies from the src to the deployment target | 22:20 |
SamYaple | oh yea thats a special module, but its still executing the code on the destination host | 22:21 |
sdake_ | somethign reads /etc/kolla/nova.conf | 22:21 |
SamYaple | it needs to be run on the destination host since it uses facts from there to file out the template | 22:21 |
sdake_ | so it calls bck into the source host how? | 22:22 |
bmace | sdake: pure file replication seems simple enough for the time being. suggest a clustered / shared file system unless you want to have issues with needing to sync files via some other mechanism. | 22:22 |
*** dimsum__ has quit IRC | 22:22 | |
bmace | it is what i was expecting for the time being anyway | 22:22 |
sdake_ | bmace you mean something like netapp nfs? | 22:22 |
SamYaple | sdake_: it doesnt ansible generates a python file with the content | 22:22 |
sdake_ | like some kind of json encode or something? | 22:23 |
SamYaple | i wish i oculd show you this | 22:23 |
SamYaple | you need to know how ansible works | 22:23 |
SamYaple | heres what you do | 22:23 |
bmace | i think there are other clustered fs options, zfs, etc. | 22:23 |
SamYaple | actually ill do a pastebin | 22:23 |
sdake_ | samyaple much appreciated | 22:24 |
SamYaple | zfs is a single host filesystem | 22:24 |
sdake_ | i need to learn how this stuff works ;) | 22:24 |
bmace | i believe there are some clustered zfs options | 22:24 |
SamYaple | nope | 22:24 |
SamYaple | zfs is only ever a single host filesystem | 22:24 |
SamYaple | its like there mission statement | 22:24 |
SamYaple | same with btrfs | 22:25 |
sdake_ | who would want to make a distributed disk based filesystem anyway | 22:25 |
* sdake_ groans | 22:25 | |
* sdake_ worked on gfs for some time | 22:25 | |
sdake_ | take my word for it - its just a bad bad idea :) | 22:25 |
SamYaple | ceph is amazing | 22:25 |
SamYaple | you need to not use gfs ever | 22:25 |
SamYaple | its bad | 22:25 |
sdake_ | its different then gfs | 22:25 |
SamYaple | its disk based | 22:26 |
sdake_ | yup | 22:26 |
SamYaple | its not different htough | 22:26 |
sdake_ | is it kernel based? ;-) | 22:26 |
SamYaple | i think you are refering to distributed POSIX based filesystem | 22:26 |
sdake_ | that is why gfs is full of fail | 22:26 |
sdake_ | i hear good things for ceph but haven't actually tried it | 22:27 |
sdake_ | i hear gluster has data corrutpion issues all the time but have tried it | 22:27 |
sdake_ | several years ago | 22:27 |
sdake_ | posix based filesystem is fine, implemented in kernel - bad idea - implemented with a dlm - bad idea | 22:28 |
*** gfidente has quit IRC | 22:28 | |
sdake_ | ^A distributed | 22:28 |
bmace | looking around the interwebs i see several mention of both active / active and active / passive mechanisms for HA ZFS.. | 22:29 |
sdake_ | samyaple pastebin or working on it? | 22:29 |
SamYaple | bmace: all using nfs | 22:29 |
SamYaple | sdake_: woking on it | 22:29 |
sdake_ | cool thanks sam | 22:29 |
SamYaple | bmace: thats not the same thing as a clustered filesystem (block level) | 22:29 |
sdake_ | i really want to understand how this stuff works | 22:30 |
SamYaple | and they suck really hard | 22:30 |
sdake_ | hard for me to be a technical leader if I dont ;) | 22:30 |
SamYaple | active/passive is alright | 22:30 |
sdake_ | the cats at hp use etcd in their ansible openstack deployment product | 22:31 |
SamYaple | and DVR | 22:31 |
sdake_ | too bad its proprietary | 22:31 |
sdake_ | bunch of fail proprietary development is | 22:31 |
SamYaple | http://paste.openstack.org/show/412237/ | 22:33 |
SamYaple | sdake_: ^ | 22:33 |
SamYaple | ill walk you through it | 22:33 |
SamYaple | i run the mariadb role with '--step' to as me which tasks to run | 22:34 |
SamYaple | i used the env variable ANSIBLE_KEEP_REMOTE_FILES=1 to keep around files after the task finishes | 22:34 |
SamYaple | the ~/.ansible/tmp/ directory has those files | 22:34 |
SamYaple | in this case it generated file ansible-tmp-1438986634.79-140852425034088/copy for that template task | 22:35 |
SamYaple | looking through that file youll see the source of the template is on the destination node | 22:35 |
SamYaple | src=/root/.ansible/tmp/ansible-tmp-1438986634.79-140852425034088/source | 22:35 |
bmace | btw, does the ansible deploy need to be done as root or sudoable user at this point? i think there was some sort of chmod or something there at some point that made me need to run as root | 22:36 |
SamYaple | bmace: if running AIO it needs to run as root | 22:37 |
SamYaple | it always will | 22:37 |
*** Haomeng has quit IRC | 22:37 | |
*** Haomeng|2 has joined #kolla | 22:37 | |
SamYaple | for multinode, ansible does not need to run as root on the deploy node | 22:37 |
SamYaple | it still needs to be logged in as root on the destnitation node (we dont have sudo support yet) | 22:37 |
bmace | that is a bit of a killer in a lot of enterprise IT environments | 22:38 |
SamYaple | feel free to write the sudo support, i havent had the cycles to be honest | 22:38 |
SamYaple | and unfortunately most deployment tools for openstack require root :/ | 22:38 |
bmace | i may end up doing that :) | 22:38 |
SamYaple | i mean it would still be passwordless sudo so its not like its that much more secure | 22:39 |
sdake_ | still secured by keys | 22:39 |
sdake_ | so its better then wide open root ;) | 22:39 |
bmace | but you think the user will always at least need sudo? just being in the docker group isn't enough? if we have some kolla user that can modify the files in /var/lib/kolla or whatever the containers mount in, and can do docker stuff, what else would they need sudo for? | 22:39 |
SamYaple | bmace: yes always | 22:40 |
bmace | for what commands? | 22:40 |
SamYaple | we do tweak sysctl and other things | 22:40 |
SamYaple | i mean, maybe with the right configuration and ahead of time preperation you wouldn't need to... | 22:40 |
SamYaple | but boy is that a bad idea | 22:40 |
SamYaple | since the containers run as root and --privileged | 22:41 |
SamYaple | the configs would have to be use modifable | 22:41 |
bmace | the only place i see any direct sysctl being done is in our gate, in the repo, unless it is hidden via being named something odd somewhere. | 22:41 |
sdake_ | could be group modifiable | 22:41 |
SamYaple | thats alot a big security hole | 22:41 |
SamYaple | bmace, sysctl has an ansible module | 22:41 |
SamYaple | we call that | 22:41 |
bmace | no bigger than letting someone have open sudo access to do anything they want | 22:41 |
SamYaple | also thanks for reminding me, cause we need to call that for other things too | 22:42 |
SamYaple | bmace: I would argue that it is | 22:42 |
SamYaple | the wrong permsissions and a user that doesnt have passwordless sudo could execute root code potentially | 22:42 |
bmace | that still isn't more access than someone that can, without a doubt, run anything as root via sudo | 22:43 |
bmace | it may be equal to, but not more than | 22:44 |
SamYaple | scenario here. user a -- privilieged has root trusted, user b -- no root access only | 22:44 |
SamYaple | user b, due to bad perms, can write to a file that root execs | 22:44 |
SamYaple | bam | 22:44 |
sdake_ | samyaple so in this case the template module is writing the file to /root/.ansible/tmp/ansible-tmp* of the contents of galera.cnf.j2 and the target host scps it from the server? | 22:44 |
sdake_ | deployment nod i mean? | 22:44 |
SamYaple | no the deployment node scps over the file, then then the template code processes it | 22:45 |
sdake_ | so one way to make this work woudl be to add etcd support to that template module | 22:45 |
sdake_ | or make something similar to it with protocol support besides file? | 22:46 |
sdake_ | or is template so hard wired into ansible that is knows to write teh file out locally? | 22:46 |
SamYaple | youre talking about rewriting a core ansible module. i have no idea how wired it in is, but i would still be against it | 22:47 |
SamYaple | even if it was possible | 22:47 |
*** bradjones has quit IRC | 22:49 | |
*** Ti-mo has quit IRC | 22:49 | |
sdake_ | tbh i am surprised ansible doesn't have etcd integration already | 22:49 |
sdake_ | its pretty hot tech | 22:49 |
SamYaple | ansible still wont support python 3 | 22:49 |
SamYaple | just saying | 22:49 |
sdake_ | and you use the template module to copy every file in /etc/kolla? | 22:50 |
sdake_ | like merge-configs uses it in some way? | 22:50 |
SamYaple | yes | 22:54 |
SamYaple | every file passes through the template filter | 22:54 |
*** bradjones has joined #kolla | 22:55 | |
*** bradjones has quit IRC | 22:55 | |
*** bradjones has joined #kolla | 22:55 | |
sdake_ | samyaple thanks for helping me understand - definately not as easy to implement as I thought it would be | 22:57 |
sdake_ | (and simple) | 22:57 |
*** Ti-mo has joined #kolla | 22:59 | |
sdake_ | someone should name their openstacck project metacortex imo ;) | 23:01 |
sdake_ | ok guys here is the deal on atc passes | 23:01 |
sdake_ | the foundation parses the governance repo, looks at the repos in there, and produces atc passes based upon commits for the last 6 months | 23:02 |
sdake_ | in our case, our governance repo lists openstack/kolla | 23:02 |
sdake_ | there is no openstack/kolla | 23:02 |
sdake_ | so no atc pass will be generated | 23:02 |
*** Ti-mo has quit IRC | 23:11 | |
*** Ti-mo has joined #kolla | 23:11 | |
SamYaple | quick, do a docs.openstack.org patch! | 23:12 |
SamYaple | if anyone would like help with that, I can point them to a few areas that could use some rework | 23:12 |
SamYaple | if you need an ATC that is | 23:12 |
SamYaple | sdake_: isn't the merge cycle to get us moved into github.com/openstack like in a few days? | 23:14 |
sdake_ | clarkb said they have no move planned | 23:14 |
sdake_ | i am pretty sure he is the infra ptl | 23:14 |
SamYaple | what does that mean? | 23:14 |
sdake_ | i am sending a mail to the mailing list now, hang tight | 23:14 |
sdake_ | and read that ok? | 23:14 |
SamYaple | good deal thanks | 23:15 |
sdake_ | (to inform people, because alot of people have their repos listed incorrectly to get atc passes). | 23:15 |
*** sdake_ is now known as sdake | 23:19 | |
*** Ti-mo has quit IRC | 23:22 | |
*** Ti-mo has joined #kolla | 23:29 | |
*** jruano has joined #kolla | 23:30 | |
sdake | samyaple https://review.openstack.org/#/c/210636/ | 23:36 |
sdake | and see post on the mailing list | 23:36 |
SamYaple | stackeforge doesnt get ATC you know.... | 23:37 |
sdake | according to fungi its all about hte gonvernance repo | 23:37 |
sdake | anything in the governance repo gets a pass | 23:37 |
sdake | if the repo is correct | 23:37 |
SamYaple | right but stackforge doesnt get ATC passes | 23:37 |
sdake | yes they do | 23:37 |
SamYaple | no they dont | 23:37 |
sdake | if they are listed in the governance repo | 23:37 |
sdake | fungi works at the foundation on atc pass generation | 23:38 |
sdake | i'll leave it up to the TC to sort it out ;) | 23:38 |
sdake | read his quote on the eavesdrop | 23:38 |
SamYaple | oh i see what youre saying | 23:39 |
SamYaple | yea stackforge projects dont get ATC | 23:39 |
SamYaple | but Kolla is no longer a stackforge project | 23:40 |
SamYaple | even though thats where teh code currently resides | 23:40 |
sdake | you should hang out on openstack-infra samyaple | 23:42 |
sdake | its an essential channel imo | 23:42 |
SamYaple | to many channels :( | 23:43 |
sdake | fungi said he is going to try to hack the script to ignore the identity namespace and just look at hte project name | 23:43 |
*** jasonsb has quit IRC | 23:45 | |
openstackgerrit | Merged stackforge/kolla: Start docker without -H to make chgrp work on socket https://review.openstack.org/210122 | 23:51 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!