| *** ssurana has joined #openstack-kolla | 00:00 | |
| *** fragatina has quit IRC | 00:01 | |
| *** matrohon has joined #openstack-kolla | 00:02 | |
| *** fragatin_ has quit IRC | 00:02 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla: Config files having underscores breaks Kolla-Kubernetes. https://review.openstack.org/326181 | 00:03 |
|---|---|---|
| *** salv-orlando has quit IRC | 00:04 | |
| *** ssurana has quit IRC | 00:04 | |
| *** mbound has quit IRC | 00:04 | |
| *** salv-orlando has joined #openstack-kolla | 00:05 | |
| *** fragatina has joined #openstack-kolla | 00:05 | |
| *** sdake_ has quit IRC | 00:05 | |
| *** fragatina has quit IRC | 00:07 | |
| *** fragatin_ has joined #openstack-kolla | 00:07 | |
| *** fragatin_ has quit IRC | 00:08 | |
| *** godleon has joined #openstack-kolla | 00:15 | |
| *** asalkeld has joined #openstack-kolla | 00:16 | |
| *** rahuls has quit IRC | 00:18 | |
| *** ssurana has joined #openstack-kolla | 00:27 | |
| *** asalkeld has quit IRC | 00:37 | |
| *** sdake has joined #openstack-kolla | 00:38 | |
| *** zhiwei has joined #openstack-kolla | 00:44 | |
| *** salv-orlando has quit IRC | 00:53 | |
| *** salv-orlando has joined #openstack-kolla | 00:54 | |
| *** sacharya has joined #openstack-kolla | 00:59 | |
| *** sacharya has quit IRC | 01:04 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Remove the deprecated kolla-build section https://review.openstack.org/325709 | 01:05 |
| *** harlowja has joined #openstack-kolla | 01:08 | |
| *** ssurana has quit IRC | 01:40 | |
| *** sdake has quit IRC | 01:50 | |
| *** sacharya has joined #openstack-kolla | 01:59 | |
| *** sdake has joined #openstack-kolla | 02:11 | |
| *** sdake_ has joined #openstack-kolla | 02:18 | |
| *** sdake has quit IRC | 02:20 | |
| *** mark-casey has joined #openstack-kolla | 02:46 | |
| *** mbound has joined #openstack-kolla | 02:47 | |
| *** phuongnh has joined #openstack-kolla | 02:49 | |
| *** yuanying_ has joined #openstack-kolla | 02:51 | |
| *** yuanying has quit IRC | 02:51 | |
| *** mbound has quit IRC | 02:51 | |
| *** matrohon has quit IRC | 02:51 | |
| *** klint has joined #openstack-kolla | 02:52 | |
| *** fragatina has joined #openstack-kolla | 03:10 | |
| *** fragatin_ has joined #openstack-kolla | 03:11 | |
| *** phuongnh has quit IRC | 03:13 | |
| *** fragatina has quit IRC | 03:15 | |
| coolsvap | sdake_, evening | 03:17 |
| *** sdake_ has quit IRC | 03:38 | |
| *** mark-casey has quit IRC | 03:45 | |
| *** daneyon_ has joined #openstack-kolla | 03:51 | |
| *** daneyon has quit IRC | 03:54 | |
| *** daneyon has joined #openstack-kolla | 04:00 | |
| *** daneyon_ has quit IRC | 04:03 | |
| *** yuanying_ has quit IRC | 04:18 | |
| *** yuanying has joined #openstack-kolla | 04:21 | |
| *** salv-orl_ has joined #openstack-kolla | 04:25 | |
| *** salv-orlando has quit IRC | 04:27 | |
| *** dave-mccowan has quit IRC | 04:35 | |
| *** sacharya has quit IRC | 04:36 | |
| *** sacharya has joined #openstack-kolla | 04:37 | |
| *** sacharya has quit IRC | 04:42 | |
| *** sacharya has joined #openstack-kolla | 05:38 | |
| *** salv-orl_ has quit IRC | 05:40 | |
| *** salv-orlando has joined #openstack-kolla | 05:40 | |
| *** sacharya has quit IRC | 05:43 | |
| *** tfukushima has joined #openstack-kolla | 06:01 | |
| *** fragatin_ has quit IRC | 06:34 | |
| *** Mr_Broken has joined #openstack-kolla | 06:43 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Bump the ansible version to >2 in the kolla-toolbox https://review.openstack.org/311625 | 07:01 |
| *** salv-orlando has quit IRC | 07:04 | |
| *** salv-orlando has joined #openstack-kolla | 07:05 | |
| *** Serlex has joined #openstack-kolla | 07:07 | |
| *** daneyon_ has joined #openstack-kolla | 07:16 | |
| *** shardy has joined #openstack-kolla | 07:21 | |
| *** daneyon_ has quit IRC | 07:21 | |
| *** matrohon has joined #openstack-kolla | 07:23 | |
| *** coolsvap is now known as coolsvap|lunch | 07:25 | |
| *** coolsvap|lunch is now known as coolsvap | 07:26 | |
| *** athomas has joined #openstack-kolla | 07:26 | |
| *** mikelk has joined #openstack-kolla | 07:29 | |
| *** salv-orlando has quit IRC | 07:56 | |
| *** salv-orlando has joined #openstack-kolla | 07:56 | |
| *** SiRiuS has joined #openstack-kolla | 07:57 | |
| *** godleon has quit IRC | 07:59 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Fix the neutron reconfigure issue https://review.openstack.org/326280 | 08:01 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Run more kolla-ansible task in the gate https://review.openstack.org/326281 | 08:05 |
| *** iceyao has joined #openstack-kolla | 08:21 | |
| *** tyrola has joined #openstack-kolla | 08:24 | |
| *** wmiller has left #openstack-kolla | 08:42 | |
| *** Jeffrey4l has joined #openstack-kolla | 08:57 | |
| *** iceyao has left #openstack-kolla | 08:57 | |
| *** iceyao has joined #openstack-kolla | 08:58 | |
| *** tfukushima has quit IRC | 09:02 | |
| *** tfukushima has joined #openstack-kolla | 09:02 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: DO NOT MERGE: TEST UBUNTU MULTIGATE https://review.openstack.org/326307 | 09:04 |
| *** dmk0202 has joined #openstack-kolla | 09:05 | |
| *** tfukushima has quit IRC | 09:07 | |
| *** papacz has quit IRC | 09:14 | |
| *** wmiller has joined #openstack-kolla | 09:21 | |
| *** jmccarthy has left #openstack-kolla | 09:22 | |
| *** jmccarthy has quit IRC | 09:22 | |
| coolsvap | Jeffrey4l, around? | 09:22 |
| Jeffrey4l | coolsvap, yup. | 09:23 |
| coolsvap | jeblair, regarding your patch 326281 | 09:23 |
| patchbot | coolsvap: https://review.openstack.org/#/c/326281/ - kolla - Run more kolla-ansible task in the gate | 09:23 |
| coolsvap | did you had a chance to look at https://etherpad.openstack.org/p/kolla-newton-summit-kolla-functional-gating | 09:23 |
| coolsvap | jeblair, unping | 09:24 |
| coolsvap | Jeffrey4l, ^^ | 09:24 |
| Jeffrey4l | coolsvap, checked. | 09:24 |
| Jeffrey4l | so? | 09:24 |
| Jeffrey4l | what's your opinion about this? | 09:24 |
| coolsvap | we need to split the ansible tasks in different gates rather than keeping all | 09:25 |
| coolsvap | so that it gives better coverage of whats working and whats not | 09:25 |
| *** jmccarthy has joined #openstack-kolla | 09:25 | |
| Jeffrey4l | i think put it to the same gate should be OK. | 09:25 |
| Jeffrey4l | at least, the precheck, reconfigure here is Ok. | 09:26 |
| *** tzn has joined #openstack-kolla | 09:26 | |
| *** mgoddard_ has quit IRC | 09:26 | |
| coolsvap | yes, but we need to look at a long list of configuration values that can be used to configure kolla | 09:27 |
| coolsvap | precheck can be here but we need the deploy_aio configurable | 09:27 |
| Jeffrey4l | ` but we need to look at a long list of configuration values that can be used to configure kolla` what means by this? | 09:28 |
| coolsvap | if you look at the above etherpad | 09:28 |
| Jeffrey4l | est Reconfigure with copy_always easy | 09:29 |
| Jeffrey4l | Test Reconfigure with copy once easy ? this? | 09:29 |
| Jeffrey4l | so, we will have a gate to test the copy_always, and another gate to test the copy_once? | 09:30 |
| Jeffrey4l | right | 09:30 |
| coolsvap | reconfigure has two ways when we do it | 09:30 |
| Jeffrey4l | ? | 09:30 |
| coolsvap | yes | 09:30 |
| Jeffrey4l | in that way, the number of the gate will grow rapidly. | 09:32 |
| Jeffrey4l | (btw, iirc, we talked the copy_once should be removed long time ago) | 09:32 |
| coolsvap | Jeffrey4l, yes and sdake said number of gates should not be a concern as long as they do the job in 30-45 mins | 09:32 |
| coolsvap | which most of our gates do now, if the gates go beyond the 90min check too often then we need to assess the reasons behind it | 09:34 |
| Jeffrey4l | in this way, kolla will build the image for several times for one PS. | 09:35 |
| coolsvap | there was one opinion to setup a registry in infra, but i think we need to revisit its status with sdake | 09:36 |
| coolsvap | do you know anything related to it? | 09:36 |
| Jeffrey4l | no. but the registry is a good idea. | 09:37 |
| Jeffrey4l | Or deploy multi times in the gate. | 09:37 |
| Jeffrey4l | 1. build the required image 2. deploy with copy_always, and test then clean 3. deploy with copy_once and test then clean | 09:38 |
| coolsvap | yes and this can be done for most of the combinations | 09:38 |
| *** mbound has joined #openstack-kolla | 09:39 | |
| coolsvap | and thats why we need deploy script configurable | 09:39 |
| Jeffrey4l | kolla have many combination, we can not set each gate for every combination. | 09:39 |
| Jeffrey4l | yes. | 09:39 |
| Jeffrey4l | maybe we can talk this in the meeting. | 09:39 |
| coolsvap | yes | 09:39 |
| Jeffrey4l | ok. cool | 09:39 |
| *** sacharya has joined #openstack-kolla | 09:40 | |
| Jeffrey4l | Or, whatever the conclusion is in the meeting. I expect to merge my PS ( may be add a TODO is better). we need test them. | 09:41 |
| *** Serlex has quit IRC | 09:42 | |
| Jeffrey4l | Whatever the conclusion is in the meeting. I expect to merge my PS right now( may be add a TODO is better). we need test them. we can make it better in the future. | 09:42 |
| Jeffrey4l | coolsvap, ^ | 09:42 |
| coolsvap | Jeffrey4l, yeah add TODO but before merge also it should pass the gate | 09:43 |
| *** Serlex has joined #openstack-kolla | 09:43 | |
| Jeffrey4l | coolsvap, the reconfigure crashed the gate. another PS fixing it. | 09:44 |
| coolsvap | Jeffrey4l, sure no worries i am 100% in on functional gating to start whichever way we can always make it better | 09:44 |
| Jeffrey4l | coolsvap, about the sanity check. what about to use the tempest? | 09:45 |
| *** sacharya has quit IRC | 09:45 | |
| coolsvap | Jeffrey4l, tempest or rally both could work, rally internally uses tempest for verification, so unless we are doing some performance tests tempest should be sufficient i believe | 09:46 |
| Jeffrey4l | rather than write some sanity check script in the ansible | 09:46 |
| Jeffrey4l | agree | 09:49 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Run more kolla-ansible task in the gate https://review.openstack.org/326281 | 09:54 |
| *** daneyon_ has joined #openstack-kolla | 09:59 | |
| openstackgerrit | Dave Walker proposed openstack/kolla: setup_{RedHat,Debian}.sh, make more idempotent https://review.openstack.org/326347 | 09:59 |
| *** wmiller has quit IRC | 10:02 | |
| *** daneyon_ has quit IRC | 10:03 | |
| openstackgerrit | Dave Walker proposed openstack/kolla: setup_{RedHat,Debian}.sh, make more idempotent https://review.openstack.org/326347 | 10:07 |
| openstackgerrit | Dave Walker proposed openstack/kolla: setup_{RedHat,Debian}.sh, make more idempotent https://review.openstack.org/326347 | 10:10 |
| *** pbourke has quit IRC | 10:12 | |
| *** pbourke has joined #openstack-kolla | 10:12 | |
| *** papacz has joined #openstack-kolla | 10:21 | |
| *** athomas has quit IRC | 10:24 | |
| *** salv-orl_ has joined #openstack-kolla | 10:25 | |
| *** tzn has quit IRC | 10:26 | |
| *** mdnadeem_ has joined #openstack-kolla | 10:28 | |
| *** salv-orlando has quit IRC | 10:28 | |
| *** mdnadeem has quit IRC | 10:31 | |
| *** dave-mccowan has joined #openstack-kolla | 10:35 | |
| *** rhallisey has joined #openstack-kolla | 10:37 | |
| *** rhallisey has quit IRC | 10:38 | |
| *** rhallisey has joined #openstack-kolla | 10:38 | |
| *** rhallisey has quit IRC | 10:57 | |
| *** salv-orl_ has quit IRC | 11:04 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Run more kolla-ansible task in the gate https://review.openstack.org/326281 | 11:04 |
| *** salv-orlando has joined #openstack-kolla | 11:05 | |
| *** salv-orl_ has joined #openstack-kolla | 11:18 | |
| *** salv-orlando has quit IRC | 11:23 | |
| *** iceyao has quit IRC | 11:23 | |
| *** iceyao has joined #openstack-kolla | 11:24 | |
| *** mliima has joined #openstack-kolla | 11:27 | |
| *** mgoddard has joined #openstack-kolla | 11:28 | |
| *** tzn has joined #openstack-kolla | 11:32 | |
| *** jmccarthy has quit IRC | 11:37 | |
| *** rhallisey has joined #openstack-kolla | 11:37 | |
| *** jmccarthy has joined #openstack-kolla | 11:38 | |
| mliima | morning all | 11:41 |
| *** sacharya has joined #openstack-kolla | 11:41 | |
| *** coolsvap_ has joined #openstack-kolla | 11:44 | |
| *** coolsvap has quit IRC | 11:45 | |
| *** sacharya has quit IRC | 11:46 | |
| *** coolsvap_ has quit IRC | 11:46 | |
| *** coolsvap has joined #openstack-kolla | 11:46 | |
| *** dwalsh has joined #openstack-kolla | 11:47 | |
| *** salv-orl_ has quit IRC | 11:51 | |
| *** salv-orlando has joined #openstack-kolla | 11:52 | |
| *** dwalsh has quit IRC | 11:55 | |
| *** salv-orlando has quit IRC | 11:58 | |
| *** salv-orlando has joined #openstack-kolla | 11:58 | |
| *** Jeffrey4l has quit IRC | 11:58 | |
| *** Jeffrey4l has joined #openstack-kolla | 11:59 | |
| openstackgerrit | Merged openstack/kolla: Remove the deprecated kolla-build section https://review.openstack.org/325709 | 12:04 |
| *** coolsvap has quit IRC | 12:07 | |
| *** diogogmt has quit IRC | 12:14 | |
| *** diogogmt has joined #openstack-kolla | 12:16 | |
| *** Serlex has left #openstack-kolla | 12:18 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Implement Rally docker file https://review.openstack.org/326414 | 12:18 |
| *** diogogmt has quit IRC | 12:24 | |
| *** tzn has quit IRC | 12:37 | |
| *** diogogmt has joined #openstack-kolla | 12:37 | |
| *** daneyon_ has joined #openstack-kolla | 12:41 | |
| *** salv-orlando has quit IRC | 12:42 | |
| *** salv-orlando has joined #openstack-kolla | 12:43 | |
| *** iceyao has quit IRC | 12:45 | |
| *** daneyon_ has quit IRC | 12:46 | |
| *** Mr_Broken has quit IRC | 12:53 | |
| *** Mr_Broken has joined #openstack-kolla | 12:54 | |
| *** tzn has joined #openstack-kolla | 12:55 | |
| *** ppowell has joined #openstack-kolla | 12:56 | |
| *** Jeffrey4l has quit IRC | 12:56 | |
| *** Mr_Broken has quit IRC | 12:58 | |
| *** klint has quit IRC | 13:04 | |
| *** Mr_Broken has joined #openstack-kolla | 13:12 | |
| *** Serlex has joined #openstack-kolla | 13:12 | |
| *** inc0 has joined #openstack-kolla | 13:13 | |
| *** jtriley has joined #openstack-kolla | 13:13 | |
| inc0 | good morning | 13:13 |
| *** dwalsh has joined #openstack-kolla | 13:22 | |
| *** sdake has joined #openstack-kolla | 13:27 | |
| *** sdake_ has joined #openstack-kolla | 13:34 | |
| *** sdake has quit IRC | 13:36 | |
| *** iceyao has joined #openstack-kolla | 13:41 | |
| *** tfukushima has joined #openstack-kolla | 13:43 | |
| *** tfukushima has quit IRC | 13:43 | |
| *** mgoddard_ has joined #openstack-kolla | 13:45 | |
| *** mgoddard has quit IRC | 13:48 | |
| wirehead_ | inc0: not sure if there's a better way than blindly renaming things. If I remove wsrep_notify.sh, then mariadb.json complains about the lack of a file. Create two configs, that means we're really starting to diverge containers with kube-only containers and ansible-only containers. | 13:59 |
| inc0 | ok, fair enough, | 14:00 |
| inc0 | wirehead_, mind adding this explanation to review | 14:01 |
| inc0 | ? | 14:01 |
| inc0 | for future sake | 14:01 |
| inc0 | also changed -1 to +2 | 14:01 |
| wirehead_ | Done. | 14:02 |
| wirehead_ | I did a quick check and the majority of config files are dashes, not underscores. | 14:03 |
| inc0 | I guess we should catch that in review of this patch | 14:03 |
| wirehead_ | I hate dealing with the OpenStack gate, but I also hate not having an OpenStack gate. :D | 14:11 |
| rhallisey | wirehead_, there are others. I know liyi has a patch up for some other _ issues | 14:21 |
| wirehead_ | not sure if there's a better way than blindly renaming things. If I remove wsrep_notify.sh, then mariadb.json complains about the lack of a file. Create two configs, that means we're really starting to diverge containers with kube-only containers and ansible-only containers. | 14:25 |
| wirehead_ | Err. | 14:25 |
| wirehead_ | https://review.openstack.org/#/c/320159/ | 14:25 |
| patchbot | wirehead_: patch 320159 - kolla - Replace underscore sign with dash in neutron confi... | 14:25 |
| mag009_ | morning!! | 14:25 |
| wirehead_ | Well, I can yank some kube-chains later on today, but I think we're kinda stuck on some level because I'm not sure if they'd make such a change in a 1.2.x build regardless. | 14:28 |
| *** daneyon_ has joined #openstack-kolla | 14:30 | |
| *** mgoddard_ has quit IRC | 14:33 | |
| *** daneyon_ has quit IRC | 14:35 | |
| *** sdake_ has quit IRC | 14:36 | |
| *** mgoddard has joined #openstack-kolla | 14:43 | |
| *** Serlex has quit IRC | 14:46 | |
| *** tzn has quit IRC | 14:49 | |
| *** tzn has joined #openstack-kolla | 14:54 | |
| mag009_ | guys | 14:55 |
| mag009_ | can someone help me out here | 14:55 |
| mag009_ | even with lastest images i'm having issue with the inventory / conditional check | 14:56 |
| *** vhosakot has joined #openstack-kolla | 15:02 | |
| *** absubram has quit IRC | 15:02 | |
| *** mliima_ has joined #openstack-kolla | 15:08 | |
| *** inc0 has quit IRC | 15:09 | |
| *** inc0 has joined #openstack-kolla | 15:09 | |
| *** thumpba has joined #openstack-kolla | 15:10 | |
| *** mliima has quit IRC | 15:10 | |
| *** zhiwei has quit IRC | 15:15 | |
| *** coolsvap has joined #openstack-kolla | 15:15 | |
| *** Jeffrey4l has joined #openstack-kolla | 15:16 | |
| *** vinkman has joined #openstack-kolla | 15:16 | |
| *** mliima_ is now known as mliima | 15:16 | |
| *** thumpba has quit IRC | 15:24 | |
| *** coolsvap has quit IRC | 15:26 | |
| *** sdake has joined #openstack-kolla | 15:28 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Implement Rally docker file https://review.openstack.org/326414 | 15:29 |
| *** coolsvap has joined #openstack-kolla | 15:30 | |
| *** mikelk has quit IRC | 15:33 | |
| *** thumpba has joined #openstack-kolla | 15:33 | |
| *** tzn has quit IRC | 15:35 | |
| *** lifeless has quit IRC | 15:37 | |
| *** vhosakot has quit IRC | 15:38 | |
| *** Mr_Broken has quit IRC | 15:39 | |
| *** Mr_Broken has joined #openstack-kolla | 15:40 | |
| *** ppowell has quit IRC | 15:41 | |
| mag009_ | finally found my problem... | 15:41 |
| *** sacharya has joined #openstack-kolla | 15:42 | |
| inc0 | mag009_, do tell | 15:44 |
| *** Mr_Broken has quit IRC | 15:44 | |
| mag009_ | my inventory was missing a group lbaas | 15:45 |
| mag009_ | but the error message in ansible give no clue about it | 15:45 |
| mag009_ | i did copied the inventory from the latest well I thought | 15:45 |
| mag009_ | guess i didnt | 15:46 |
| mag009_ | -_- | 15:46 |
| sdake | mag009_ fwiw I agree with you ansible diagnostics are pretty poor | 15:46 |
| sdake | mag009_ i'm not sure what we can do about that in kolla land | 15:46 |
| sdake | other then offer some type of linting | 15:46 |
| mag009_ | but I can deal with that :) | 15:47 |
| sdake | mag009_ in the big scheme of things, a linting project is low priority imo :) | 15:47 |
| mag009_ | i'd rather debug ansible than puppet... | 15:47 |
| *** sacharya has quit IRC | 15:47 | |
| mag009_ | the only bug left is the delegate_host for the ceph | 15:48 |
| mag009_ | it's not a bug but a miss understanding of how delegate work from my end | 15:48 |
| *** lifeless has joined #openstack-kolla | 15:49 | |
| sdake | hey lifeless | 15:49 |
| sdake | hows life less? ;) | 15:50 |
| coolsvap | sdake, good morning | 15:50 |
| sdake | hey coolsvap | 15:50 |
| *** vhosakot has joined #openstack-kolla | 15:52 | |
| mag009_ | funny thing tho to know which group I was missing I used inventory_hostname with union of all the group and it pointed me to the error if I leave the Or it give me a random error | 15:53 |
| *** absubram has joined #openstack-kolla | 15:55 | |
| mag009_ | btw anyone had this behavior before the cleanup sometime leave some docker volume ceph_mon_config for example sometime doesn't get cleanup | 15:58 |
| *** sacharya has joined #openstack-kolla | 16:03 | |
| *** inc0 has quit IRC | 16:06 | |
| *** inc0 has joined #openstack-kolla | 16:06 | |
| Mech422 | Morning | 16:06 |
| mag009_ | it's noon here | 16:08 |
| mag009_ | morning! | 16:09 |
| *** mgoddard has quit IRC | 16:10 | |
| openstackgerrit | Steve Hindle proposed openstack/kolla: Read partition tables directly, bypassing stale kernel data for ceph. The Ceph osd bootstrap/startup logic depends upon reading/writing partition names. Unfortunately, the /sys file system doesn't get updated when partition names change. To work aroun https://review.openstack.org/326609 | 16:12 |
| Mech422 | mag009_: heh - noon actually sounds like a good time to get up :-) | 16:13 |
| mag009_ | Mech422: where you at ? west coast? | 16:14 |
| *** matrohon has quit IRC | 16:14 | |
| wirehead_ | naw, us westcoasters are just getting to work. | 16:14 |
| *** daneyon_ has joined #openstack-kolla | 16:18 | |
| Mech422 | mag009_: yeah - PHX,AZ | 16:19 |
| Mech422 | mag009_: but I'm a night owl - so sleeping till noon sounds good in general :-) | 16:20 |
| *** daneyon_ has quit IRC | 16:23 | |
| *** salv-orl_ has joined #openstack-kolla | 16:25 | |
| *** ssurana has joined #openstack-kolla | 16:25 | |
| *** salv-orlando has quit IRC | 16:27 | |
| ccesario | hey guys | 16:28 |
| Mech422 | Mornin | 16:28 |
| ccesario | morning Mech422 | 16:28 |
| *** thumpba has quit IRC | 16:29 | |
| *** dmk0202 has quit IRC | 16:30 | |
| sdake | pbourke around? | 16:30 |
| pbourke | sdake: hi | 16:31 |
| ccesario | guys, according https://github.com/openstack/kolla/blob/master/doc/cinder-guide.rst I 'm getting this error : http://paste.openstack.org/show/508716/ | 16:31 |
| ccesario | But running the commands manually on the "host" it seems work... http://paste.openstack.org/show/508717/ | 16:31 |
| ccesario | any tip !? | 16:31 |
| *** coolsvap has quit IRC | 16:33 | |
| *** fragatina has joined #openstack-kolla | 16:34 | |
| *** coolsvap has joined #openstack-kolla | 16:34 | |
| *** fragatina has quit IRC | 16:35 | |
| *** fragatina has joined #openstack-kolla | 16:35 | |
| *** vinkman has left #openstack-kolla | 16:36 | |
| Mech422 | ccesario: all the basics checkout? your ansible user has sudo and all that fun stuff ? | 16:43 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Make build.py importable lib https://review.openstack.org/326108 | 16:43 |
| ccesario | Mech422, yep.... double checking now | 16:44 |
| inc0 | guys I'd appreciate eyes on this https://review.openstack.org/#/c/326108/3 | 16:44 |
| patchbot | inc0: patch 326108 - kolla - Make build.py importable lib | 16:44 |
| inc0 | every change in build.py is hell to rebase | 16:44 |
| *** iceyao has quit IRC | 16:51 | |
| Mech422 | inc0: I only have time for a quick look - but it seems fine. Project_root etc shouldn't need adjusting cuz the code moved 'sideways' to the new directory | 16:55 |
| inc0 | yeah, I just want to get this merged sooner rather than later;) | 16:56 |
| inc0 | will have more patches moving stuff around soon | 16:56 |
| Mech422 | inc0: the relative paths are always what get me in projects like that... | 16:57 |
| Mech422 | inc0: always _something_ you forget to adjust :-P | 16:57 |
| inc0 | pbourke, rhallisey mandre can anyone please remind my why are we copying whole docker directory to tmp during build? | 16:58 |
| cinerama | sean-k-mooney, you around? we had a handful of minor fixes on your bifrost patch, but also we'd like to help you rebase it if you want | 16:59 |
| cinerama | sean-k-mooney: really looking forward to getting this landed :) | 17:00 |
| mandre | inc0: two reasons | 17:01 |
| mandre | 1. we need to render the jinja2 templates | 17:01 |
| mandre | 2. it allows you to change git branch during build | 17:01 |
| inc0 | ok fair enough, I'll keep it there for now but I think we should revisit it, maybe there is some way to make docker build from string | 17:04 |
| *** thumpba has joined #openstack-kolla | 17:04 | |
| inc0 | then we can render jinja and put it into StringIO | 17:04 |
| *** Lyncos has joined #openstack-kolla | 17:09 | |
| Mech422 | inc0: is include_footer the proper place to put extra drivers for neutron? I want to get started on midonet and calico testing ? | 17:14 |
| inc0 | Mech422, so we want to completely change way customizations work | 17:14 |
| inc0 | hold on | 17:15 |
| inc0 | https://review.openstack.org/#/c/324665/ | 17:15 |
| patchbot | inc0: patch 324665 - kolla - Mechanism for Dockerfile customization | 17:15 |
| inc0 | I think if this lands we'll deprecate include_footer | 17:15 |
| inc0 | sdake, ^ any progress on this one? | 17:16 |
| inc0 | can I rebase it, remove sample blocks and get this merged? and solve your missing spots in later patches | 17:16 |
| Mech422 | inc0: will that get rolled into stable/mitaka ? Hmm - should I be targeting stable mitaka for new deploys? | 17:16 |
| inc0 | Mech422, no, this will be Neutron | 17:16 |
| inc0 | for stable mitaka you need to make it work with includes | 17:17 |
| inc0 | newton | 17:17 |
| inc0 | omg | 17:17 |
| Mech422 | inc0: yeah - just occured to me, maybe I should target N, or even kolla-kube? | 17:17 |
| Mech422 | inc0: prolly a couple of months before this goes live... | 17:17 |
| inc0 | depends what you need really | 17:17 |
| Mech422 | inc0: kolla-kube sounds pretty slick... but probably more then a few months out | 17:18 |
| inc0 | as for kolla-kube, sure, just make sure you know what you're getting yourself into;) | 17:18 |
| inc0 | it might be very slick, but it's a 2 months old project | 17:18 |
| inc0 | so it's RnD still | 17:18 |
| Mech422 | inc0: yeah - thats what I'm wondering - got a status meeting at the top of the hour - might try to pin a timeline down more | 17:18 |
| Mech422 | master targets N now, right ? | 17:19 |
| inc0 | yeah | 17:19 |
| inc0 | well | 17:19 |
| inc0 | master is master | 17:19 |
| inc0 | next release will be N | 17:19 |
| Mech422 | ohh - and I could ditch ansible 1.9! :-) | 17:19 |
| inc0 | yup | 17:19 |
| *** tyrola has quit IRC | 17:19 | |
| inc0 | but if you really want kolla for mitaka with midokura | 17:20 |
| inc0 | we can work on that too | 17:20 |
| Mech422 | any rough timelines for a stable/newton release ? couple of months ? 6 months ? | 17:20 |
| inc0 | with include footer you can do lot of stuff | 17:20 |
| inc0 | http://releases.openstack.org/newton/schedule.html Oct 03-07 | 17:20 |
| Mech422 | inc0: midokura is targeting mitaka right now - but I'm sure they'll support newton as well - they want to become the 'default' network stack for openstack... | 17:21 |
| Mech422 | inc0: Ahh thanks.... | 17:21 |
| inc0 | Mech422, so again, you really can do stuff as for now | 17:21 |
| inc0 | it will be hacky but might work | 17:21 |
| Mech422 | inc0: I don't mind the hacky bit for home - but I'd really prefer not to have to re-invent everything in a couple of months... | 17:22 |
| inc0 | and when proper customization mechanism lands, you can rewrite it in a proper way | 17:22 |
| Mech422 | inc0: I'll see who faints if I propose and end of Oct. release | 17:22 |
| inc0 | no, not reinvent | 17:22 |
| Mech422 | s/and/an/ | 17:22 |
| inc0 | general idea will be exactly the same | 17:23 |
| inc0 | you'll get just a lot of new tools at your disposal and you'll be able to refactor your customizations | 17:23 |
| inc0 | so for example for midokura you'll need to uninstall ovs I think | 17:23 |
| inc0 | and build everything and then rebuild only neutron client with customization | 17:24 |
| Mech422 | inc0: yeah - you uninstall openvswitch-switchd, leave the kernel module, and Mido drives the kernel module via flows | 17:24 |
| *** tyrola has joined #openstack-kolla | 17:24 | |
| inc0 | so with include footer you'd need someting like RUN apt-get uninstall openvswitch-vswitchd... | 17:24 |
| inc0 | and you'll be able to ditch that in N | 17:25 |
| inc0 | so it won't be that horrible | 17:25 |
| Mech422 | inc0: KK - thanks...I'll start digging :-) | 17:25 |
| inc0 | sure, let me know if I can be of any help, I'd really like to have success story of "kolla running midonet/calico";) | 17:25 |
| Mech422 | inc0: me too :-) | 17:26 |
| Mech422 | inc0: I need ceilo at some point too | 17:26 |
| Mech422 | inc0: though thats not as big a deal as the networking stuff | 17:27 |
| inc0 | ceilo patches are proposed already | 17:27 |
| Lyncos | ceilo patches ? | 17:27 |
| Lyncos | you didn't tell me about that yesterday | 17:27 |
| *** thumpba has quit IRC | 17:27 | |
| inc0 | https://review.openstack.org/#/c/300574/ | 17:28 |
| patchbot | inc0: patch 300574 - kolla - Implement ceilometer ansible role | 17:28 |
| inc0 | I did tell you, they're in review | 17:28 |
| inc0 | well...a patch, singular | 17:28 |
| Lyncos | :-) | 17:29 |
| Lyncos | ah ok | 17:29 |
| Lyncos | that one | 17:29 |
| Lyncos | sorry | 17:29 |
| Lyncos | I took a look at Talligent | 17:31 |
| Lyncos | it look awesome | 17:31 |
| inc0 | I'm glad | 17:32 |
| inc0 | let me know if that meets your needs | 17:32 |
| Lyncos | it would meet our needs if we had the budget tu buy it | 17:32 |
| inc0 | I never tested that, but I know people who do and they like it | 17:32 |
| Lyncos | I'm speaking about the enterprise version | 17:33 |
| *** ravig has joined #openstack-kolla | 17:39 | |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Mechanism for Dockerfile customization https://review.openstack.org/324665 | 17:40 |
| *** salv-orl_ has quit IRC | 17:41 | |
| *** shardy has quit IRC | 17:41 | |
| *** salv-orlando has joined #openstack-kolla | 17:41 | |
| *** signed8bit has joined #openstack-kolla | 17:45 | |
| mag009_ | neutron_plugin_agent technically i can force a specific plugin in my inventory right ? | 17:46 |
| mag009_ | I replaced it with calico but it still load openvswitch | 17:47 |
| klindgren | sdake, you were talking about scaleability of neutron. We did run into one issue.... | 17:51 |
| klindgren | we needed to bump the files ulimit otherwise neutron api eventlet work would just exit after a week or so | 17:52 |
| klindgren | worker* | 17:52 |
| *** fragatina has quit IRC | 17:56 | |
| *** ravig has quit IRC | 18:04 | |
| *** tyrola has quit IRC | 18:08 | |
| wirehead_ | sdake / rhallisey : So, dcwangmit01 and I were chatting about uber-dev focused stuff that’s not useful for end-users and is in fact the wrong way to actually run things… but would help us scale-up future Kolla-Kubernetes committers more quickly. Put that in the hacking.rst or contributing.rst? Create a doc page that’s even more archane than quickstart? Put it in the wiki for the time being because it’s too fast-moving r | 18:09 |
| wirehead_ | now? | 18:09 |
| *** diogogmt has quit IRC | 18:11 | |
| *** diogogmt has joined #openstack-kolla | 18:13 | |
| mag009_ | any objection that change the neutron_plugin_agent variable to look at the inventory | 18:14 |
| mag009_ | i realize that it's a general variable but I'd like to have the ability to say X compute use calico and Y compute use openvswitch | 18:15 |
| *** diogogmt has quit IRC | 18:15 | |
| *** thumpba has joined #openstack-kolla | 18:16 | |
| mag009_ | unless someone has another proposal I'll make the changes... | 18:16 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Add customizations for apt keys and packages https://review.openstack.org/326688 | 18:26 |
| inc0 | mag009_, soo...you can do that | 18:28 |
| inc0 | with config overrides | 18:28 |
| inc0 | also if it works with inventory, and it should, that also works | 18:28 |
| inc0 | inventory is your own file | 18:28 |
| mag009_ | yes but it does not | 18:28 |
| inc0 | ahh | 18:29 |
| mag009_ | because it doesn't lookup to the hostvars | 18:29 |
| rhallisey | wirehead_, contributing is fine | 18:29 |
| inc0 | ok I know why | 18:29 |
| mag009_ | I have to tell the start.yml to use the hostvar | 18:29 |
| inc0 | so when we call our plays | 18:29 |
| inc0 | its ansible-playbook -e @globals.yml | 18:29 |
| mag009_ | i do that | 18:29 |
| inc0 | and I think value in globals takes priority | 18:29 |
| inc0 | do you have this var in globals? | 18:29 |
| mag009_ | but I dont want to set a global var | 18:29 |
| inc0 | defined? | 18:29 |
| mag009_ | I want to be able to say host A = calico , host B = openvswitch | 18:30 |
| inc0 | I understand | 18:30 |
| inc0 | hold on, let me refresh my ansible | 18:30 |
| mag009_ | if I define let say neutron_plugin_agent = "openvswitch" in global | 18:30 |
| inc0 | it's about which variable takes priority | 18:30 |
| mag009_ | and set the var in my inventory to calico | 18:31 |
| inc0 | no, what I mean, don't define it there at all | 18:31 |
| inc0 | and see if it works | 18:31 |
| mag009_ | it take the global one | 18:31 |
| inc0 | if you define it in inventory only | 18:31 |
| mag009_ | ah ok i see what you mean | 18:31 |
| mag009_ | hold on | 18:31 |
| mag009_ | I'm almost done with the calico integration | 18:33 |
| mag009_ | awesome it worked | 18:34 |
| mag009_ | lol | 18:34 |
| mag009_ | thx inc0 | 18:34 |
| inc0 | cool, glad to hear | 18:41 |
| *** coolsvap has quit IRC | 18:41 | |
| inc0 | sdake, ping | 18:41 |
| *** ppowell has joined #openstack-kolla | 18:44 | |
| *** tzn has joined #openstack-kolla | 18:48 | |
| *** mbound has quit IRC | 18:51 | |
| *** dwalsh has quit IRC | 18:53 | |
| *** tzn has quit IRC | 18:54 | |
| sdake | inc0 anything in globals.yml takes priority and overrides eveerything else | 18:57 |
| sdake | inc0 wound me | 18:57 |
| *** fragatina has joined #openstack-kolla | 18:58 | |
| inc0 | https://review.openstack.org/#/c/326688/1 adding new packages is even easier than I thought | 18:58 |
| patchbot | inc0: patch 326688 - kolla - Add customizations for apt keys and packages | 18:58 |
| inc0 | all we need to do is to refactor our current dockerfiles to use variables instead of raw paste | 18:59 |
| inc0 | like that https://github.com/openstack/kolla/blob/master/docker/base/Dockerfile.j2#L208 | 19:00 |
| inc0 | then in customization file you just set new value, you can use var = var + whatever | 19:00 |
| inc0 | or just var = whatever | 19:00 |
| *** dwalsh has joined #openstack-kolla | 19:03 | |
| sdake | so that allows easier override of packages installed | 19:04 |
| sdake | either override or addition | 19:04 |
| sdake | how is that extended to other areas | 19:04 |
| *** rahuls has joined #openstack-kolla | 19:04 | |
| sdake | there is a case where we want to run raw docker commands as well | 19:04 |
| inc0 | then you use original block idea | 19:05 |
| *** SiRiuS has quit IRC | 19:09 | |
| sdake | inc0 we need bsae_all_apt_packages | 19:11 |
| sdake | base_bin_apt_packages | 19:11 |
| sdake | and base_src_apt_packages | 19:11 |
| inc0 | I'd rather have if binary then packages = [binary] elif source then packages=[source] | 19:13 |
| inc0 | one var, value changes depending on build mode | 19:13 |
| sdake | the bin and src packages sometimes differ | 19:14 |
| inc0 | they do, and that will be reflected in value | 19:14 |
| inc0 | also having all the packages for every possible combination in same dockerfile really hurt readibility | 19:15 |
| inc0 | we need to decouple it somehow | 19:15 |
| *** Kennan has quit IRC | 19:15 | |
| Mech422 | mag009_: If you get a chance - I'd love to see your calico stuff... | 19:16 |
| *** Kennan has joined #openstack-kolla | 19:16 | |
| sdake | inc0 i find the conditionals the hardeest part of our dockerfiles to read/write | 19:16 |
| inc0 | for the record, I was against it from the beggining | 19:18 |
| inc0 | I wanted to do it right way back in Santa Clara | 19:18 |
| sdake | can't change the past | 19:18 |
| sdake | focus on the future ;) | 19:18 |
| sdake | at that point in time the decision we made was the best for the project | 19:18 |
| sdake | doing it "right"sometimes means failing, which at that point was not really an option | 19:19 |
| inc0 | well, my point is, now it's time to pay off this tech debp | 19:19 |
| inc0 | debt | 19:19 |
| sdake | inc0 just to be clear, kyouwere agaist which in sc | 19:20 |
| inc0 | limting ourselves to only "if" | 19:20 |
| sdake | well if we go through a massive refactor, I htink we can agree we want it done right this time around | 19:21 |
| sdake | in a way that developers can understand clearly | 19:21 |
| sdake | so it can be implemented in distributed fashion without having to be inside your brain :) | 19:21 |
| sdake | inc0 taking the if source pkg = if binary pkg = to its natural concllusion for consistency sake, this applies to all components we assemble? | 19:22 |
| inc0 | well, most of them | 19:23 |
| inc0 | same goes for keys in apt | 19:23 |
| inc0 | might go for repos in yum too | 19:23 |
| sdake | as well as repos atleast in yum | 19:23 |
| inc0 | pretty much anything that is a list | 19:23 |
| sdake | what owuld be good is a standard "template" | 19:23 |
| sdake | that every dockerfile follows | 19:24 |
| inc0 | so I don't want to write strict rules from start and let them evolve as we go | 19:24 |
| inc0 | but I think it's safe to say that any var/block should start by container name | 19:24 |
| inc0 | and then some descriptive suffix that makes sens | 19:25 |
| inc0 | e | 19:25 |
| sdake | consistency is key | 19:25 |
| sdake | that lowers learning curve | 19:25 |
| sdake | which is hwy i suggested the template idea | 19:25 |
| inc0 | sure, but we will develop standards after first few prototypes | 19:25 |
| sdake | our j2 srt of follows a templat enow | 19:25 |
| inc0 | all I'm saying is make first few and then take a look at them and use them as standard | 19:26 |
| inc0 | or template as you say | 19:26 |
| sdake | is there code to build this j2 file you published? | 19:27 |
| inc0 | just check it out and it works | 19:27 |
| inc0 | --template-override=override-file | 19:27 |
| sdake | i like hearing that | 19:27 |
| inc0 | it;s follow up patch from one I submitted last week | 19:27 |
| inc0 | https://review.openstack.org/#/c/324665/ I cleaned it up and left only code for mechanism | 19:28 |
| patchbot | inc0: patch 324665 - kolla - Mechanism for Dockerfile customization | 19:28 |
| inc0 | so blocks keep working | 19:28 |
| sdake | inc0 hat is the purpose of line 231 | 19:28 |
| sdake | {% block base_ubuntu_package_installation %} | 19:28 |
| inc0 | in case you'd like to completely override whole block | 19:29 |
| inc0 | and really I could remove pre and post block | 19:29 |
| inc0 | because you can acieve both with {{ super }} | 19:29 |
| inc0 | like {% block base_ubutnu_package_installation %}RUN my-pretty-custom-command {{super}} {% endblock %} | 19:30 |
| sdake | shuoudln't line 233 hae a block around it using same logic? | 19:31 |
| sdake | what is super | 19:31 |
| inc0 | 233 is inside this block | 19:31 |
| inc0 | {{ super }} => it will fill it with original content | 19:32 |
| sdake | yes should it be base_ubuntu_repo_installation | 19:32 |
| sdake | in block format | 19:32 |
| inc0 | well it installs keys and packages, not repos;)_ | 19:32 |
| sdake | base_ubuntu_key_installatino | 19:32 |
| sdake | then is w hat i emant ^^ :) | 19:32 |
| sdake | in elemental, keys, rpeos, and package are all separate things | 19:33 |
| inc0 | not sure, if you want to manage keys, you do this with var | 19:33 |
| sdake | the fourth type of thing is raw commands | 19:33 |
| inc0 | but regardless, you can put block inside a block | 19:33 |
| inc0 | so we can do both | 19:33 |
| inc0 | here, if you want custom keys you override variable base_apt_keys to your liking | 19:34 |
| inc0 | including stuff like base_apt_keys = "prettykey" + base_apt_keys | 19:34 |
| sdake | what i'm getting at is we either don't need line 231 | 19:34 |
| sdake | or we need a line 233 with a block to be consistent | 19:34 |
| sdake | i dont see a whole lot of value in teh line 231 block | 19:35 |
| inc0 | well, question tho, what's the use case of overriding line 233? | 19:35 |
| sdake | same question for line 231 ;) | 19:35 |
| inc0 | we can remove both | 19:35 |
| sdake | cool | 19:35 |
| sdake | lets make it minimal but fit the use cases outlined in the elemental spec | 19:35 |
| inc0 | thing is, having too many blocks is better than having to few | 19:35 |
| inc0 | then pretty much all you need is to turn all install/repo/keys into list-based;) | 19:36 |
| inc0 | and it will just work | 19:36 |
| sdake | next up | 19:36 |
| sdake | 230, | 19:36 |
| sdake | shouldn't that be base_deb | 19:36 |
| inc0 | we don't really have naming convention yet | 19:37 |
| sdake | we are making them now :) | 19:37 |
| inc0 | so instead of 2 block | 19:37 |
| inc0 | pre and post | 19:37 |
| sdake | so lets think about it before we go an drewrite 115 dockerfiles | 19:37 |
| inc0 | I'd just leave base_ubuntu_package_installation | 19:37 |
| inc0 | we can base on os family as debian/redhat | 19:38 |
| sdake | i dont ssee any value in overriding the installation mechanism | 19:38 |
| sdake | only what is installed | 19:38 |
| inc0 | then we can replace all if base_distro in [ubutnu, debian] with if base_distro_faimly==debian | 19:38 |
| sdake | ya that wont work | 19:39 |
| sdake | different keys | 19:39 |
| inc0 | sdake, but do you see value in adding somethign before or after? | 19:39 |
| sdake | with fedora, different ttools | 19:39 |
| sdake | inc0 let me switch to a cooler part of my hosue moment | 19:39 |
| sdake | https://review.openstack.org/#/c/323612/5/specs/elemental.rst | 19:41 |
| patchbot | sdake: patch 323612 - kolla - Introduction of a new DSL: Elemental | 19:41 |
| sdake | 36-55 | 19:41 |
| sdake | sometimes repo comes from a file, sometimes repo comes from a uri | 19:41 |
| sdake | atleast in rpm environment | 19:42 |
| inc0 | not in apt world | 19:42 |
| inc0 | anyway, easy | 19:42 |
| inc0 | macro | 19:42 |
| sdake | which not in apt world - you mean from uri? | 19:43 |
| *** SiRiuS__ has joined #openstack-kolla | 19:43 | |
| inc0 | http://jinja.pocoo.org/docs/dev/templates/#macros write a macro that will understand repo and install it | 19:43 |
| inc0 | sdake, in apt world you have sources.list | 19:43 |
| inc0 | just a file | 19:43 |
| sdake | gpg keys can also come from file or uri | 19:44 |
| sdake | does this macro go in one master file somewhere and get included in? | 19:44 |
| inc0 | macro is a function | 19:45 |
| *** mliima_ has joined #openstack-kolla | 19:45 | |
| inc0 | you can call it {{ install_repo(file_or_url) }} | 19:45 |
| sdake | i know its a function | 19:45 |
| inc0 | and depending on value of file_or_url will eirther install file or url | 19:46 |
| sdake | my quetion is do we write teh function once | 19:46 |
| sdake | or do we copy it to every 115 dockerfiles | 19:46 |
| inc0 | or rather print RUN install file or RUN install url | 19:46 |
| inc0 | we can have file with common macros and include it in every dockerfile | 19:46 |
| inc0 | or have master dockerfule every other dockerfile will extend | 19:46 |
| sdake | ok, that was my original question :) | 19:46 |
| inc0 | I think latter is better | 19:46 |
| sdake | i dont see immediate value of extension - inheritence is ah ard concept for alot of people to grasp | 19:47 |
| sdake | whereas includes are nobrainers | 19:47 |
| sdake | coud lyou explain how extension would offer a simpler setup? | 19:47 |
| *** mliima has quit IRC | 19:48 | |
| sdake | what would b ecool is if there was one master file that did all the work | 19:48 |
| inc0 | well, maybe include | 19:48 |
| sdake | and a lsit of variables in the children j2 files | 19:48 |
| inc0 | both will require single line in every Dockerfile | 19:48 |
| sdake | inc0 don't care about additional line in ever ydockerfile | 19:49 |
| inc0 | yup | 19:49 |
| sdake | care about a) is it simple for our devs to implement b) is it siimpl for operators to override | 19:49 |
| inc0 | {% include 'kolla_macros.j2' %} | 19:50 |
| sdake | i know include is simple | 19:50 |
| sdake | i was asking about extension | 19:50 |
| inc0 | I don't understand now. | 19:50 |
| sdake | extends() is the keywword iirc | 19:50 |
| inc0 | it is | 19:51 |
| sdake | wha twould be cool is to write one template that has all the raw work | 19:51 |
| sdake | and a file that extends it with variables | 19:51 |
| sdake | intuitively i'm not sure if that would work | 19:51 |
| inc0 | problem is there is little raw work | 19:51 |
| inc0 | our templats are fairly DRY | 19:51 |
| inc0 | unless we want to add macros and stuff, then true | 19:52 |
| *** mbound has joined #openstack-kolla | 19:52 | |
| *** mliima_ has left #openstack-kolla | 19:52 | |
| sdake | i get there is raw work | 19:52 |
| sdake | that could go in teh children dockerfile.j2 | 19:52 |
| sdake | but if we go copy that base.j2 modele around | 19:52 |
| sdake | we will have the same override blocks all ove the place in the same placess | 19:53 |
| sdake | seems like a waste ;) | 19:53 |
| sdake | so bacck to original question | 19:53 |
| inc0 | well, if we just include base dockerfiles then we don't use docker cache any more | 19:53 |
| sdake | would it be possible to psecify jsut the variables and raw stuff in the children j2 | 19:53 |
| sdake | and have the parent that we extend from use all that data | 19:53 |
| inc0 | yes...just not super valuable imho | 19:54 |
| inc0 | our Dockerfile, aside from base, are simple | 19:54 |
| inc0 | and we dont want to render 100 huge dockerfiles | 19:54 |
| sdake | i agree - we are talking about making them complex atm :) | 19:54 |
| inc0 | we want to render 1 with 90 lines and 99 with 10 lines | 19:54 |
| sdake | by making a list of variables in each one | 19:55 |
| sdake | and repos in each one | 19:55 |
| sdake | etc | 19:55 |
| inc0 | I don't understand what you want | 19:55 |
| sdake | i wish i knew jinja better - then i could sho wyou | 19:55 |
| inc0 | if you want image to be huge, you don't use docker image features any more | 19:55 |
| sdake | we want small imagess | 19:55 |
| sdake | as small as possible | 19:55 |
| inc0 | yeah | 19:55 |
| sdake | that should b e#1 on that list of erquirements ;) | 19:56 |
| inc0 | hence base image big, other small | 19:56 |
| sdake | let me show you example | 19:56 |
| sdake | moment | 19:56 |
| inc0 | if you want custom full build per image, you don't use base image any more | 19:56 |
| *** tzn has joined #openstack-kolla | 19:57 | |
| Mech422 | inc0: woot! got the go ahead to start working against HEAD - we'll switch to stable/newton once its released :-) | 19:57 |
| *** mbound has quit IRC | 19:57 | |
| sdake | https://github.com/openstack/kolla/blob/master/docker/heat/heat-api/Dockerfile.j2 | 19:57 |
| sdake | Mech422 newton is 4+ months out | 19:57 |
| Mech422 | sdake: yeah - octoberish ? | 19:57 |
| sdake | Mech422 i'd highly recommend against not working with head | 19:57 |
| inc0 | Mech422, nice, you can help us with this customization stuff then:P | 19:57 |
| sdake | november | 19:57 |
| inc0 | sdake, Oct | 19:57 |
| inc0 | according to Newton release | 19:57 |
| sdake | ok | 19:57 |
| sdake | so link above | 19:58 |
| inc0 | and working == contributing not prod;) | 19:58 |
| Mech422 | yeah - I'm just building POC stuff that can be used to deploy to prod later... | 19:58 |
| sdake | there we have package we install in the binary case one for centos and one for ubuntu | 19:58 |
| *** dwalsh has quit IRC | 19:58 | |
| sdake | inc0 what would heat-api.j2 look like in a new world order ;) | 19:59 |
| sdake | inc0 assumign we threw kitchen sink at it | 19:59 |
| inc0 | so what we could do is something like that: << workign on paste >> | 19:59 |
| inc0 | http://paste.openstack.org/show/508759/ | 20:02 |
| inc0 | and install_package macro will understand if centos then yum elif ubuntu then apt | 20:02 |
| ccesario | hey guys ... doe somebody has tested cinder + lvm ? http://paste.openstack.org/show/508760/ | 20:03 |
| sbezverk_ | ccesario yes | 20:03 |
| sdake | inc0 missing repo and key install | 20:03 |
| sdake | inc0 lets keep iterating o n this | 20:03 |
| inc0 | just add macro install_repo and install_key | 20:04 |
| ccesario | sbezverk_, is it needed any magic for this ? | 20:04 |
| mag009_ | docker question i have this app where i want to run as user in a docker but it need to create a socket in /run/ how do I proceed in docker ? | 20:04 |
| mag009_ | i get a permission denied because the user cannot write in /run | 20:05 |
| sdake | inc0 macros can take null values and do the right thing? | 20:05 |
| sbezverk_ | ccesario nope, as long as you configure it correctly it works like a charm.. I am running 12 VMs in kolla aio each using 50GB cinder lvm volume | 20:06 |
| inc0 | http://jinja.pocoo.org/docs/dev/templates/#macros not sure about null, but default "" or [] should be ok | 20:06 |
| sdake | inc0 do macros have access to the global namespace - in other words, would it be able to operate on the evariable base_distro | 20:06 |
| inc0 | yup | 20:06 |
| sdake | the reason I ask is fedora uses dnf | 20:06 |
| sdake | insteadd of yum | 20:06 |
| inc0 | I think so, if not we can add base_distro to argument | 20:06 |
| sdake | so install_packages would need to be able to handle that | 20:06 |
| sbezverk_ | ccesario why do not you use just regular cinder create command for volumes? | 20:06 |
| inc0 | but 99% they have global namespace | 20:07 |
| sdake | inc0 can you find out now - I'd like to get this work rolling | 20:07 |
| sdake | and if we can come up with a basic template | 20:07 |
| sdake | i'll set the hounds loose | 20:07 |
| sbezverk_ | ccesario the way of creating volumes you pasted, really really weird | 20:07 |
| ccesario | sbezverk_, beucase I got error | 20:08 |
| ccesario | sbezverk_, http://paste.openstack.org/show/508714/ | 20:09 |
| sbezverk_ | ccesario can you get cinder.conf from your cinder_volume container? | 20:10 |
| sbezverk_ | ccesario you are on master right? | 20:11 |
| *** zhiwei has joined #openstack-kolla | 20:11 | |
| sdake | inc0 i really like the last approach you paste | 20:11 |
| sdake | d | 20:11 |
| sdake | inc0 assuming it would work (the difficulty would be in writing the macros) | 20:12 |
| inc0 | sdake, works with global | 20:12 |
| sdake | nice so no extra variable passing needed then? | 20:12 |
| inc0 | nope | 20:13 |
| ccesario | sbezverk_, well I did some changes to try identify a error on master code... http://paste.openstack.org/show/508666/ | 20:13 |
| ccesario | sbezverk_, already reported on dev-list | 20:13 |
| inc0 | so we can easily add install_packages(list_of_packages) macro | 20:13 |
| sdake | inc0 lets not focus on the base container at first | 20:13 |
| ccesario | my currently cinder.conf http://paste.openstack.org/show/508762/ | 20:13 |
| inc0 | that will understand which distro works | 20:13 |
| sdake | we need install_repos and install_keys too | 20:14 |
| inc0 | I'll start with install_packages | 20:15 |
| inc0 | you get me names to help | 20:15 |
| sdake | names to help which | 20:15 |
| sdake | the macro writing | 20:15 |
| ccesario | sbezverk_, https://www.mail-archive.com/openstack-dev%40lists.openstack.org/msg85315.html | 20:15 |
| sdake | or the distributed dev | 20:15 |
| inc0 | names of people that will help me | 20:15 |
| inc0 | both | 20:15 |
| sdake | with which part | 20:15 |
| mag009_ | anyone is working on integrating different auth for the keystone ? ldap for example | 20:15 |
| sdake | i dont think anyone in our community is a jinja2 rocket scientist | 20:15 |
| inc0 | I can write all, just will start with one | 20:15 |
| sdake | so the prototyping work is on you i think | 20:16 |
| *** zhiwei has quit IRC | 20:16 | |
| inc0 | but it would be good if I'm not only person in community who ever wrote a macro in jinja;) | 20:16 |
| sdake | but I will requeest folks in our community to do the implementation work after the macro file is written | 20:16 |
| sdake | and the prototype is written | 20:16 |
| inc0 | I'll get this patch upstream today or tomorrow | 20:16 |
| sdake | and then just so I am clear | 20:17 |
| inc0 | you get the word out and ack https://review.openstack.org/#/c/324665/ | 20:17 |
| patchbot | inc0: patch 324665 - kolla - Mechanism for Dockerfile customization | 20:17 |
| inc0 | kk? | 20:17 |
| inc0 | this one is requirement | 20:17 |
| sdake | to override heat_api_packages, I'd speciffy heat_api_packages = ['yang'] in a master file somewhere? | 20:18 |
| inc0 | in file you provide during build | 20:18 |
| sdake | ok | 20:18 |
| inc0 | --template-override | 20:18 |
| sdake | and that allows replacement or override? | 20:18 |
| inc0 | or append or prepend | 20:18 |
| sdake | nice | 20:18 |
| inc0 | heat_api_packages = heat_api_packages + ['yang'] | 20:19 |
| sdake | install order is repo, key, packages | 20:19 |
| inc0 | it's a list, it's ordered | 20:19 |
| sdake | i mean to help you out during dev | 20:19 |
| sdake | that is the order it has to go in as generated by the dockerfile otuput | 20:19 |
| inc0 | althout if you'd do apt-get install pkg1 pkg2 order shouldn't matter | 20:20 |
| sdake | package ordering doesn't matter | 20:20 |
| sdake | but repo and key order do matter in relationship to packages | 20:20 |
| inc0 | well, you can't really mix these two | 20:20 |
| inc0 | so you'd need to | 20:20 |
| sdake | repo then key then packages | 20:20 |
| inc0 | instal_repo install_package install_more_repos | 20:21 |
| sdake | can't mix which two | 20:21 |
| inc0 | nvm, I know what you mean | 20:21 |
| inc0 | and will tell you, not a problem | 20:21 |
| sdake | wfm | 20:21 |
| inc0 | we still write Dockerfiles at the end | 20:21 |
| sdake | so we have 3 of our 4 things completely customizable | 20:21 |
| sdake | the last thing we need customizable is the raw docker contents | 20:22 |
| ccesario | sbezverk_, have you any tip!?? | 20:22 |
| inc0 | but before I'm done, ack this plz | 20:22 |
| inc0 | https://review.openstack.org/#/c/326108/ | 20:22 |
| patchbot | inc0: patch 326108 - kolla - Make build.py importable lib | 20:22 |
| inc0 | so I won't have to rebase it more later | 20:22 |
| sdake | reviewing moment | 20:23 |
| inc0 | sdake, for raw docker content you have blocks | 20:23 |
| inc0 | just put them wherever you want | 20:23 |
| sbezverk_ | ccesario not sure, I have never hit any issues with the latest master.. but I would start with cinder.conf check | 20:25 |
| sdake | inc0 right | 20:25 |
| sdake | inc0 that was my thinking too | 20:25 |
| sbezverk_ | to make sure that it points NOT to VIP but to server's physical interface | 20:25 |
| sbezverk_ | and also vg is specified correctly | 20:26 |
| *** ppowell has quit IRC | 20:26 | |
| sbezverk_ | ccesario I am about to leave but if you want I am opened for a debugging session via webex tomorrow.. | 20:27 |
| sdake | inc0 ok well this idea seems workable and avoids all of the problems i have had in teh past with the jinja2 changes | 20:27 |
| ccesario | sbezverk_, currently after me change the cinder.conf ... it seems a permission problem happen | 20:28 |
| sdake | and especially avoids having a whole lotta blocksk in each dockerfile | 20:28 |
| inc0 | will make our files A LOT more readable as well | 20:28 |
| ccesario | sbezverk_, sure.... the server has access over ssh | 20:28 |
| ccesario | if you wants check... we can do it over ssh+screen :) | 20:28 |
| sdake | to turn off repos in base and only use the repos in heat-api, id do? | 20:29 |
| sdake | heat_api_repos = ['my_repos'] | 20:29 |
| sdake | base_repos = [''] | 20:29 |
| sdake | ? | 20:29 |
| sdake | inc0 ^^ | 20:30 |
| inc0 | sdake, when base is built, it's build | 20:31 |
| inc0 | built | 20:31 |
| sbezverk_ | ccesario what is your time zone? | 20:31 |
| inc0 | you can't change base from heat_api | 20:31 |
| sdake | yes, but it gets built as part of that set of configs | 20:31 |
| inc0 | no, it just get builds | 20:31 |
| inc0 | built | 20:31 |
| sdake | right base is rebuilt | 20:31 |
| inc0 | and then heat api uses FROM | 20:32 |
| sdake | with the new base_repos of nothing | 20:32 |
| inc0 | then you will have 2 base images | 20:32 |
| inc0 | and if you build something else | 20:32 |
| inc0 | it will use newer, which is broken | 20:32 |
| ccesario | sbezverk_, UTC-03:00 | 20:32 |
| inc0 | so you can either change FROM in heat_api to some custom base image | 20:32 |
| inc0 | or manually uninstall stuff from heat_api and then install it | 20:33 |
| inc0 | it's how docker works.. | 20:33 |
| sdake | inc0 i'm talking about a global build of all containers | 20:33 |
| sdake | where each containe rhas its own repo file | 20:33 |
| sdake | and base has nothing | 20:33 |
| inc0 | then you don't really have base | 20:33 |
| *** dwalsh has joined #openstack-kolla | 20:33 | |
| inc0 | and this is horrible antipattern | 20:33 |
| sdake | i recognize base would be limited ;) | 20:33 |
| inc0 | you will lose benefit of layered builds | 20:34 |
| sdake | persoanlly I agree that one repo location makes more sense | 20:34 |
| sdake | but get alot of feedback sure would be nice if repos were cusotmizable per container | 20:34 |
| inc0 | if you build base image once, it stays this way | 20:34 |
| sbezverk_ | ccesario ok I am -4, I will ping you around 9am tomorrow, 10am your time, let's see what is broken.. meanwhile could you make sure that you have the latest master and freshly build images | 20:34 |
| inc0 | so for example | 20:35 |
| sdake | inc0 yes what my example above did is make sure bbaxe has no repos in it | 20:35 |
| inc0 | if you set repo1 in base, then apt-get install foo, it will get installed from repo1 | 20:35 |
| inc0 | if you then change it to repo2, foo will keep being installed in repo1 | 20:35 |
| mag009_ | inc0 | 20:35 |
| inc0 | from* | 20:35 |
| inc0 | wassup mag009_ | 20:36 |
| mag009_ | inc0: any objection to move the dnsmasq directly into dhcp-agent image | 20:36 |
| inc0 | and sdake yeah, if you do that you will do horrible antipatter, but will work | 20:36 |
| sdake | mag009_ no objection | 20:36 |
| mag009_ | in calico for example I need to replace with calico package | 20:36 |
| inc0 | mag009_, yeah, 1 pid per container | 20:36 |
| sdake | actuallly we hae multiple pids per container | 20:36 |
| inc0 | where sdake ? | 20:36 |
| sdake | most of our containers run multiple pids | 20:36 |
| sdake | but they do "one job" | 20:36 |
| inc0 | hold on, where? | 20:37 |
| sdake | that is the definition fo a thin container in my book ;) | 20:37 |
| sdake | heat-api for example spawns multiple api threads | 20:37 |
| sdake | all api containers behave in this way | 20:37 |
| sdake | i won't even get started on what libvirt spawns | 20:37 |
| ccesario | sbezverk_, sure... I 'm testing using AIO method | 20:37 |
| inc0 | well, but we don't have multiple services | 20:37 |
| inc0 | per container | 20:37 |
| sdake | right that is the "one job" | 20:38 |
| inc0 | so dhcp-agent and dnsmasq shouldn't coexist in same container | 20:38 |
| sdake | a thin container is one which does no job or purpose | 20:38 |
| sdake | no/one | 20:38 |
| sdake | isn't dnsmasq part of an implementation detail for dhcp? | 20:38 |
| inc0 | mag009_, how does calico work? it uses dnsmasq configs? | 20:38 |
| mag009_ | yes | 20:39 |
| inc0 | then just run separate calico container and mount this config | 20:39 |
| mag009_ | the way it work it replaces the neutron dhcp agent | 20:39 |
| inc0 | then just replace it;) | 20:39 |
| mag009_ | that agent launch a dnsmasq for each zone | 20:39 |
| inc0 | well if agent launches dnsmasq that's ok | 20:40 |
| mag009_ | i have no control over the dnsmasq the agent launch it | 20:40 |
| inc0 | it's like sub-threads | 20:40 |
| inc0 | or api threads | 20:40 |
| inc0 | in my book;) | 20:40 |
| sdake | it is a dependency | 20:40 |
| sdake | dependencies don't count for fat containers imho :) | 20:40 |
| mag009_ | sure but i'm using the neutron base and since it include dnsmasq when I try to reinstall with the calico version apt-get bitch about it | 20:41 |
| mag009_ | dependencies etc.. | 20:41 |
| mag009_ | its ugly that way.. | 20:41 |
| sdake | mag009_ where is dhcp used atm? | 20:41 |
| sdake | in the neutron-dhcp-agent? | 20:41 |
| openstackgerrit | Merged openstack/kolla: Make build.py importable lib https://review.openstack.org/326108 | 20:41 |
| mag009_ | each compute are a dhcp server | 20:41 |
| *** jtriley has quit IRC | 20:42 | |
| sdake | mag009_ bounce a review off the queue - and we will weigh in | 20:42 |
| mag009_ | what I did is create a new container/ Dockerfile called : calico-dhcp-agent | 20:42 |
| mag009_ | based on neutron-dhcp-agent | 20:42 |
| sdake | that sounds good | 20:43 |
| sdake | the issue is dnsmasq is instaleld in neutron-base? | 20:43 |
| mag009_ | yep | 20:43 |
| sdake | just dont install it | 20:43 |
| sdake | or uninstall it if it is | 20:43 |
| sdake | (if it gets pulled in as a transitive dep) | 20:43 |
| mag009_ | where does it run normally the dnsmasq in a normal installation ? | 20:43 |
| sdake | no clue | 20:44 |
| mag009_ | in a container or with the neutron dhcp agent | 20:44 |
| sdake | i asked you above | 20:44 |
| mag009_ | in calico I know | 20:44 |
| *** ayoung has quit IRC | 20:44 | |
| sdake | sbezverk_ do you knwo the answer to mag009_ 's question? | 20:45 |
| mag009_ | i think the dnsmasq run within the neutron-dhcp-agent | 20:45 |
| Lyncos | I think mag009 is right | 20:45 |
| sbezverk_ | sdake sorry which question? | 20:45 |
| mag009_ | if that is so why not move the package dnsmasq in Dockerfile of neutron-dhcp-agent | 20:45 |
| Lyncos | because when we setup the domain name returned by DHCP .. it is in dnsmasq | 20:45 |
| mag009_ | this is my question | 20:45 |
| mag009_ | its not a big deal I can work around it but i found it's ugly the way its going to be.. | 20:46 |
| mag009_ | :) | 20:46 |
| sdake | mag009_ i dont think there is any problem with that -if there is our gate will catch it | 20:46 |
| sdake | mag009_ only way to iknow is to try it :) | 20:46 |
| mag009_ | I'll submit a draft tomorrow anyway so we can look over it | 20:46 |
| sdake | inc0 is there any way i can help with the macro wizardry | 20:47 |
| sdake | inc0 or any other way | 20:47 |
| sbezverk_ | mag009_ would be good to try to move dnsmasq to dhcp agent container | 20:47 |
| sdake | do do not :) NO TRY | 20:47 |
| sbezverk_ | in this case we will not hit similar type of issues | 20:47 |
| inc0 | sdake, I'll make a prototype, you find people to change every single yum install to macro;) | 20:48 |
| sbezverk_ | when fancy neutron plugin wants to install their own "things" :-) | 20:48 |
| sdake | inc0 i can handle that part | 20:48 |
| *** rhallisey has quit IRC | 20:49 | |
| mag009_ | guys is there a milestone/roadmap for this project ? | 20:50 |
| sdake | mag009_ yes | 20:50 |
| mag009_ | like how do you determine the priorities etc.. | 20:50 |
| sdake | mag009_ people work on what they like | 20:50 |
| mag009_ | lol | 20:50 |
| mag009_ | ok | 20:50 |
| sdake | mag009_ I may set priorities on launchpad, but that doen't have much affect ;) | 20:50 |
| mag009_ | it'd be nice to have some that are respected tho | 20:51 |
| sdake | mag009_ could you be more specific | 20:51 |
| mag009_ | i dont know i just think that it's missing some of the core feature(s) like (ldap integration) as an examples | 20:52 |
| mag009_ | it's missing a lot of configurable settings in openstack | 20:52 |
| sdake | mag009_ moment | 20:52 |
| sdake | mag009_ have you read http://docs.openstack.org/developer/kolla/deployment-philosophy.html | 20:53 |
| sdake | mag009_ re ldap and other forms of extension, inc0 and I are working on that right now ;) | 20:53 |
| mag009_ | ok | 20:53 |
| Lyncos | we will need LDAP .. ASAP | 20:53 |
| mag009_ | its just an example | 20:53 |
| sdake | not ldap specifically | 20:53 |
| Lyncos | :-) | 20:53 |
| sdake | but extension in general | 20:53 |
| sdake | all the configurable settingss are totally available in kolla | 20:54 |
| *** kproskurin has joined #openstack-kolla | 20:54 | |
| sdake | see above philosphy document | 20:54 |
| mag009_ | yes I will read it | 20:54 |
| mag009_ | thx | 20:54 |
| sdake | what other core features you looking for | 20:54 |
| mag009_ | ldap was one so far.. and it's more about customizing stuff like the domain for dhcp etc.. | 20:55 |
| Lyncos | LDAP, new version of ceph | 20:55 |
| mag009_ | but I guess it says in the doc | 20:55 |
| mag009_ | how the customization work | 20:55 |
| sdake | you can customize any openstack specific variable | 20:55 |
| sdake | new version of ceph - you mean newer then hammer? | 20:56 |
| Lyncos | yeah | 20:56 |
| Lyncos | man They are releasing Jewel | 20:56 |
| Lyncos | :-) | 20:56 |
| sdake | we don't integrate pre-release software typically | 20:56 |
| Lyncos | ok | 20:56 |
| Lyncos | Infernalis then | 20:56 |
| sdake | once its in repos, I suspect someone will take it up | 20:56 |
| sdake | atm we are stuck with what centos storage sig gives us on centos | 20:57 |
| sdake | same story re ubuntu | 20:57 |
| mag009_ | how do I customize the qemu.conf for example ? | 20:57 |
| mag009_ | or dnsmasq.conf | 20:57 |
| sdake | mag009_ ini files can be overridden | 20:57 |
| sdake | and merged | 20:57 |
| sdake | qmeu.conf can only be overridden | 20:57 |
| sdake | not merged | 20:57 |
| sdake | make your own qemu.conf in /etc/kolla/config/qemu.conf | 20:58 |
| mag009_ | ok so I have to run another playbook | 20:58 |
| sdake | I think - moment | 20:58 |
| sdake | mag009_ to apply your reconfigured settings you run kolla-ansible reconfigure | 20:58 |
| mag009_ | yes but for that I need to write my template with my own playbook | 20:59 |
| mag009_ | put it in kolla config folder and than run reconfigure | 20:59 |
| sdake | qemu.conf may not be customizable | 20:59 |
| sdake | (at this time) | 20:59 |
| mag009_ | well for qemu I've modified the template to support calico | 21:00 |
| mag009_ | I have some requirement for calico | 21:00 |
| sdake | ya we want policy files and other miscallenous files to be configurable via reconfigure | 21:00 |
| mag009_ | so I did a if condition if enable_calico than put the extra settings | 21:00 |
| sdake | its going to happen - essential priority for newton | 21:00 |
| sdake | atm only ini files are customizable | 21:00 |
| *** thumpba has quit IRC | 21:01 | |
| mag009_ | ok gotcha | 21:01 |
| sdake | other features? | 21:01 |
| mag009_ | i see where you guys are going with this make more sense now.. | 21:02 |
| Lyncos | we'll also have to modify the HA model | 21:02 |
| Lyncos | to stop using keepalived | 21:02 |
| Lyncos | and do ECMP | 21:02 |
| mag009_ | I've started the work on that | 21:02 |
| sdake | mag009_ you mean ater reading the philosophy document? | 21:02 |
| sdake | what is ecmp | 21:02 |
| mag009_ | yes | 21:02 |
| Lyncos | Equal cost multi path | 21:02 |
| *** tzn has quit IRC | 21:02 | |
| Lyncos | BGP load balancing | 21:02 |
| Lyncos | with health checks on haproxy .. (not sure if it's implemented yet) | 21:03 |
| mag009_ | are vip is mounted on each controller | 21:03 |
| mag009_ | all we need is haproxy for health check but i've already started doing the work to support ecmp | 21:03 |
| mag009_ | well it's basically excluding keepalived | 21:03 |
| mag009_ | in our case | 21:03 |
| sdake | what software provides the same functionality as keepalived that uses ecmp? | 21:04 |
| Lyncos | BHP | 21:04 |
| Lyncos | *BGP | 21:04 |
| mag009_ | nothing | 21:04 |
| *** ayoung has joined #openstack-kolla | 21:05 | |
| Lyncos | it's a network technique.. not a software | 21:05 |
| sdake | i see | 21:05 |
| sdake | well we generally dont say no to patches | 21:05 |
| sdake | but what you propose I don't know enough about to speak intelligently | 21:05 |
| Lyncos | it's quite easy | 21:05 |
| Lyncos | you add an IP to loopback interface of each controllers | 21:05 |
| Lyncos | and the network stack is sending equally the traffic to each adevertied host | 21:06 |
| Lyncos | if you want to stop .. you remove the IP from the host | 21:06 |
| sdake | which network tack - you mean the switch? | 21:06 |
| Lyncos | and BGP stop advertizing it | 21:06 |
| Lyncos | BGP | 21:06 |
| Lyncos | :-) | 21:06 |
| Lyncos | https://en.wikipedia.org/wiki/Equal-cost_multi-path_routing | 21:07 |
| sdake | m0ar reading | 21:07 |
| Lyncos | BGP will advertise multiple path to reach an IP that is hosted on multiple machine | 21:07 |
| Lyncos | The applet viewer thing in wikipedia show how it work very wall | 21:09 |
| sdake | devil's advocate here, how is that better then keepalived | 21:09 |
| Lyncos | it scale | 21:09 |
| sdake | i am familiar with bgp at a high level ;) | 21:09 |
| *** ravig has joined #openstack-kolla | 21:09 | |
| Lyncos | It scale better than keepalive.. and also is active-active strategy | 21:09 |
| Lyncos | instead of active-passive | 21:09 |
| Lyncos | and there is no single point of failure | 21:09 |
| Lyncos | no need for load balancers .. or moving ip around | 21:10 |
| sdake | does it wor with all gear or only specific gear? | 21:10 |
| Lyncos | it's pure BGP | 21:11 |
| Lyncos | so It should work with anything.. but you need to have a full L3 enabled network | 21:11 |
| Lyncos | With cisco you probably have to buy a module that is too expensive | 21:12 |
| sdake | i have a netgear switch (10gb) at my house | 21:12 |
| *** diogogmt has joined #openstack-kolla | 21:12 | |
| sdake | but I dont think it lists bgp as one of it sfeatures | 21:12 |
| Lyncos | :-) | 21:12 |
| Lyncos | We are using switches running cumulus linux | 21:12 |
| Lyncos | you can ask inc0 he want to buy one | 21:13 |
| Lyncos | ;) | 21:13 |
| inc0 | first I need to buy a house;) | 21:14 |
| sdake | any other features | 21:14 |
| inc0 | then lay down 10gig over whole house | 21:14 |
| sdake | inc0 cat7 ftw ;) | 21:14 |
| Lyncos | fibre is cheaper | 21:14 |
| Lyncos | the cable is cheaper | 21:14 |
| Lyncos | :) | 21:14 |
| sdake | my 1k spool of cat7 was about 200 bucks | 21:14 |
| *** ravig has quit IRC | 21:15 | |
| Lyncos | What you are using Cat7 for | 21:15 |
| sdake | 10gig copper switch | 21:15 |
| inc0 | Lyncos, 10gig over copper | 21:15 |
| Lyncos | you can run 10gig on Cat6 no ? | 21:15 |
| wirehead_ | That’s so metal. | 21:15 |
| inc0 | Lyncos, yeah, cat6 works too | 21:15 |
| sdake | Lyncos depends on the mhz rating of the cabling | 21:16 |
| Lyncos | ok | 21:16 |
| inc0 | but I think cat7 has longer length limit | 21:16 |
| sdake | Lyncos i couldn't actually find any cat6 at 500mhz or more | 21:16 |
| *** godleon has joined #openstack-kolla | 21:16 | |
| Lyncos | ok I see | 21:16 |
| Lyncos | anyway | 21:16 |
| Lyncos | Alerting would be a nice feature | 21:17 |
| Lyncos | Sensu | 21:17 |
| Lyncos | but I understand this is big | 21:17 |
| inc0 | Lyncos, we had session about that in summit | 21:17 |
| Lyncos | can be big | 21:17 |
| Lyncos | what was the conclusion ? | 21:17 |
| inc0 | we need a volounteer who will do it;) | 21:17 |
| Lyncos | ahah | 21:18 |
| inc0 | but we have few options as software stack goes | 21:18 |
| inc0 | we can use whisper/cassandra and grafana | 21:18 |
| inc0 | anyway, we spoke about that couple weeks ago | 21:18 |
| Lyncos | whisper/cassandra/grafana this is for Logging no ? | 21:18 |
| Lyncos | not monitoring | 21:18 |
| Lyncos | or I missed something | 21:19 |
| *** SiRiuS__ has quit IRC | 21:19 | |
| Lyncos | Cat7 is shielded | 21:19 |
| wirehead_ | Well, Prometheus would slot better into the Kubernetes side | 21:19 |
| inc0 | nto sure about whisper but grafana is monitoring | 21:19 |
| *** diogogmt has quit IRC | 21:19 | |
| inc0 | cassandra is database | 21:19 |
| inc0 | we already have logging | 21:19 |
| Lyncos | Grafana can send alerts ? | 21:19 |
| inc0 | it visualizes stuff | 21:20 |
| Lyncos | yeah | 21:20 |
| inc0 | not sure if can send as well | 21:20 |
| Lyncos | we are using grafana | 21:20 |
| inc0 | but I'm pretty sure there will be something that can | 21:20 |
| Lyncos | but not for alerting | 21:20 |
| wirehead_ | There’s basic support for collecting Prometheus metrics directly from the pods and using that to drive Auto Scaling in Kubernetes. | 21:20 |
| Lyncos | we use sensu for alerting and it connects to the same backend | 21:20 |
| inc0 | that makes sense | 21:20 |
| inc0 | anyway, whoever take it on will probably have freedom to take tech he/she likes most | 21:20 |
| Lyncos | We may have to do it | 21:21 |
| Lyncos | MAG is really good at it :-) | 21:21 |
| Lyncos | he will kill me | 21:22 |
| Lyncos | :-) | 21:22 |
| *** diogogmt has joined #openstack-kolla | 21:22 | |
| sdake | inc0 can you crank out those macros today | 21:23 |
| sdake | and then i'll set the hounds lose :) | 21:23 |
| inc0 | testing as we speak | 21:26 |
| inc0 | sdake, midcycle is 12-13 in Ralegh right? | 21:31 |
| inc0 | Jul | 21:31 |
| sdake | not in r raleigh | 21:31 |
| sdake | its in durhum | 21:31 |
| inc0 | ah | 21:31 |
| sdake | which is 1 hour away | 21:31 |
| inc0 | ok, well NC in any case | 21:31 |
| inc0 | Lyncos, ^ | 21:31 |
| sdake | rdu is the airport | 21:31 |
| sdake | we have max capacity for 30 people | 21:32 |
| Lyncos | Damn | 21:32 |
| Lyncos | I'm in vacation | 21:32 |
| mag009_ | i want in | 21:32 |
| sdake | mag009_ its ope nto anyone | 21:32 |
| sdake | i need to get eventbrite setup | 21:32 |
| Lyncos | I'll have to move my parentals :-) | 21:32 |
| mag009_ | so you were serious about doing it by the beach? | 21:33 |
| mag009_ | :P | 21:33 |
| *** ravig has joined #openstack-kolla | 21:33 | |
| inc0 | well Durhum is kinda far away from ocean:P | 21:33 |
| sdake | https://wiki.openstack.org/wiki/Sprints/KollaNewtonSprint | 21:33 |
| *** tzn has joined #openstack-kolla | 21:35 | |
| mag009_ | there's a lake near by | 21:35 |
| mag009_ | that will do | 21:35 |
| *** rhallisey has joined #openstack-kolla | 21:37 | |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Mechanism for Dockerfile customization https://review.openstack.org/324665 | 21:39 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Macro install packages https://review.openstack.org/326734 | 21:39 |
| inc0 | sdake, ^ | 21:39 |
| wirehead_ | It’s winter in the southern hemisphere. We want the *next* midcyle to be on a beach in Brazil. | 21:40 |
| wirehead_ | I’d prefer Sint Maarten. We could sit on the beach that’s astonishingly close to the airport runways. | 21:41 |
| *** tzn has quit IRC | 21:41 | |
| inc0 | wirehead_, brazil is so close to equator that it doesn't really matter | 21:41 |
| wirehead_ | OH, right. | 21:41 |
| wirehead_ | But those of us not in Phoenix / Texas / California would appreciate the change, I’m sure. | 21:42 |
| wirehead_ | If we did it at Sint Maarten, we could settle technical debates not with a sword duel but with a “who can hang on while a 747 takes off” contests. | 21:42 |
| openstackgerrit | Michal Jastrzebski (inc0) proposed openstack/kolla: Macro install packages https://review.openstack.org/326734 | 21:43 |
| sdake | inc0 i was thinking you would add the cmplexity of handling repos from raw too | 21:43 |
| sdake | inc0 so we can validate the approach works | 21:43 |
| inc0 | sdake, just test it | 21:44 |
| inc0 | there is no magic | 21:44 |
| inc0 | anway, I'm off, cya tomorrow | 21:44 |
| sdake | later | 21:45 |
| *** inc0 has quit IRC | 21:45 | |
| Mech422 | inc0: doesn't that macro need to spit out 'RUN' too ? | 21:48 |
| Mech422 | or have RUN in the dockerfile? | 21:49 |
| *** ayoung has quit IRC | 21:54 | |
| *** sdake_ has joined #openstack-kolla | 22:03 | |
| *** sdake has quit IRC | 22:06 | |
| *** sdake has joined #openstack-kolla | 22:06 | |
| *** ayoung has joined #openstack-kolla | 22:09 | |
| *** sdake_ has quit IRC | 22:10 | |
| *** dwalsh has quit IRC | 22:13 | |
| *** vhosakot has quit IRC | 22:22 | |
| *** diogogmt has quit IRC | 22:22 | |
| *** salv-orl_ has joined #openstack-kolla | 22:24 | |
| wirehead_ | sdake: so, I’m at a bit of an impasse with kubernetes + kolla + ansible. The memcached config.json file is currently hardcoding to 10.0.2.15. I want that to be 0.0.0.0 when I’m starting the memcached container (so it listens to everything, mediated by a kube service) but then when other services want to talk to memcached, they need to talk to ‘memcached.openstack' | 22:25 |
| sdake | wirehead_ not sure but i am taking a break right now for 15-20 mins | 22:26 |
| sdake | bbiaf | 22:26 |
| *** salv-orlando has quit IRC | 22:27 | |
| *** zhiwei has joined #openstack-kolla | 22:27 | |
| *** zhiwei has quit IRC | 22:32 | |
| *** Mr_Broken has joined #openstack-kolla | 22:34 | |
| *** Mr_Broken has quit IRC | 22:39 | |
| *** sdake has quit IRC | 22:43 | |
| *** tzn has joined #openstack-kolla | 22:51 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Convert Memcached to work without HostNetwork=True https://review.openstack.org/326757 | 22:51 |
| *** signed8bit is now known as signed8bit_Zzz | 22:55 | |
| *** tzn has quit IRC | 22:57 | |
| *** ravig has quit IRC | 22:58 | |
| *** absubram has quit IRC | 23:04 | |
| *** rahuls has quit IRC | 23:05 | |
| *** ravig has joined #openstack-kolla | 23:08 | |
| *** signed8bit_Zzz is now known as signed8bit | 23:09 | |
| *** rahuls has joined #openstack-kolla | 23:09 | |
| *** kproskurin has quit IRC | 23:11 | |
| *** sdake has joined #openstack-kolla | 23:12 | |
| *** sacharya_ has joined #openstack-kolla | 23:15 | |
| *** sacharya has quit IRC | 23:15 | |
| *** rahuls has quit IRC | 23:15 | |
| *** ravig has quit IRC | 23:22 | |
| *** ravig has joined #openstack-kolla | 23:22 | |
| *** sacharya_ has quit IRC | 23:25 | |
| *** rahuls has joined #openstack-kolla | 23:26 | |
| *** signed8bit is now known as signed8bit_Zzz | 23:28 | |
| *** salv-orl_ has quit IRC | 23:43 | |
| *** salv-orlando has joined #openstack-kolla | 23:43 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Convert RabbitMQ to work without HostNetwork=True https://review.openstack.org/326763 | 23:45 |
| *** sdake has quit IRC | 23:46 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Convert MariaDB to work without HostNetwork=True https://review.openstack.org/325613 | 23:47 |
| *** ravig has quit IRC | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!