| *** ijw has quit IRC | 00:02 | |
| *** ijw has joined #openstack-ironic | 00:03 | |
| *** Sukhdev has quit IRC | 00:03 | |
| *** Nisha_away has quit IRC | 00:09 | |
| *** mtanino has quit IRC | 00:13 | |
| *** naohirot has joined #openstack-ironic | 00:22 | |
| *** achanda_ has quit IRC | 00:26 | |
| *** ijw_ has joined #openstack-ironic | 00:31 | |
| *** mitchjameson has quit IRC | 00:32 | |
| *** ijw has quit IRC | 00:34 | |
| *** yuanying_ has joined #openstack-ironic | 00:36 | |
| *** yuanying has quit IRC | 00:40 | |
| *** yuanying_ has quit IRC | 00:42 | |
| *** yonglihe has joined #openstack-ironic | 00:43 | |
| *** yuanying has joined #openstack-ironic | 00:45 | |
| *** Pradip has quit IRC | 00:48 | |
| *** cing has joined #openstack-ironic | 00:50 | |
| *** ijw_ has quit IRC | 00:50 | |
| *** ijw has joined #openstack-ironic | 00:51 | |
| *** yuanying has quit IRC | 00:52 | |
| *** davideagnello has quit IRC | 00:52 | |
| *** davideagnello has joined #openstack-ironic | 00:54 | |
| *** yuanying has joined #openstack-ironic | 01:12 | |
| *** lazy_prince has joined #openstack-ironic | 01:14 | |
| *** openstack has joined #openstack-ironic | 01:25 | |
| *** ijw has quit IRC | 01:32 | |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic: Remove requirements.txt from tox.ini deps https://review.openstack.org/202349 | 01:36 |
|---|---|---|
| *** ijw has joined #openstack-ironic | 01:36 | |
| *** puranamr has joined #openstack-ironic | 01:36 | |
| *** lazy_prince has quit IRC | 01:43 | |
| *** ijw has quit IRC | 01:49 | |
| *** puranamr has quit IRC | 02:03 | |
| *** puranamr has joined #openstack-ironic | 02:18 | |
| *** ijw has joined #openstack-ironic | 02:24 | |
| jroll | lifeless: am I doing this right https://review.openstack.org/#/c/202349/ | 02:34 |
| jroll | tests passed, I assume that's a good thing :P | 02:34 |
| lifeless | jroll: yup, commented there for you | 02:37 |
| jroll | thanks | 02:37 |
| jroll | lifeless: yeah, I wondered about that | 02:37 |
| jroll | it's basically generating a picture in our docs | 02:37 |
| jroll | and it's late here | 02:37 |
| jroll | so I figured deal with it later | 02:38 |
| lifeless | yeah | 02:38 |
| lifeless | I'd just submit the g-r change to get the ball rolling on it | 02:39 |
| lifeless | nothing more can be done right away anyhow | 02:39 |
| jroll | lifeless: well the thing is, that never runs in CI | 02:40 |
| jroll | and devs likely never run it, maybe occassionally | 02:40 |
| *** puranamr has quit IRC | 02:40 | |
| jroll | I'll let others decide if they want to deal with that | 02:41 |
| lifeless | jroll: so its docs right? | 02:41 |
| jroll | kind of | 02:41 |
| lifeless | jroll: why wouldn't it get checked for validity in CI ? | 02:42 |
| jroll | I'm not 100% sure, I just learned about it | 02:42 |
| lifeless | anyhow, sure | 02:42 |
| *** hakimo_ has joined #openstack-ironic | 02:53 | |
| *** hakimo has quit IRC | 02:54 | |
| *** ijw has quit IRC | 02:56 | |
| *** bradjones has quit IRC | 03:14 | |
| *** yuanying_ has joined #openstack-ironic | 03:16 | |
| *** yuanying_ has quit IRC | 03:18 | |
| *** yuanying has quit IRC | 03:18 | |
| *** yuanying_ has joined #openstack-ironic | 03:18 | |
| *** bradjones has joined #openstack-ironic | 03:19 | |
| *** bradjones has quit IRC | 03:19 | |
| *** bradjones has joined #openstack-ironic | 03:19 | |
| *** pal has joined #openstack-ironic | 03:26 | |
| openstackgerrit | Zhenguo Niu proposed openstack/python-ironicclient: Filtering nodes by provision state https://review.openstack.org/197012 | 03:27 |
| *** yuanying_ has quit IRC | 03:30 | |
| *** yuanying has joined #openstack-ironic | 03:32 | |
| *** hblixt has joined #openstack-ironic | 03:35 | |
| *** yuanying has quit IRC | 03:37 | |
| *** yuanying has joined #openstack-ironic | 03:38 | |
| *** lazy_prince has joined #openstack-ironic | 03:38 | |
| openstackgerrit | Merged openstack/bifrost: Minor README fix for supported drivers https://review.openstack.org/200837 | 03:42 |
| openstackgerrit | Merged openstack/bifrost: Correct requirements https://review.openstack.org/202295 | 03:43 |
| openstackgerrit | Merged openstack/bifrost: Sync with global requirements https://review.openstack.org/202297 | 03:43 |
| *** zsmithnyc has quit IRC | 03:45 | |
| *** lekha has quit IRC | 03:45 | |
| *** h00327910_ has quit IRC | 03:46 | |
| *** Nisha has joined #openstack-ironic | 03:46 | |
| *** boris-42 has quit IRC | 03:46 | |
| *** Ng has quit IRC | 03:47 | |
| *** aweeks has quit IRC | 03:47 | |
| *** coolsvap|away is now known as coolsvap | 03:47 | |
| *** zsmithnyc has joined #openstack-ironic | 03:57 | |
| *** lekha has joined #openstack-ironic | 03:58 | |
| *** yuanying has quit IRC | 04:00 | |
| *** yuanying has joined #openstack-ironic | 04:01 | |
| *** achanda has joined #openstack-ironic | 04:02 | |
| *** yuanying has quit IRC | 04:02 | |
| *** aweeks has joined #openstack-ironic | 04:04 | |
| *** yuanying has joined #openstack-ironic | 04:07 | |
| *** boris-42 has joined #openstack-ironic | 04:07 | |
| *** ijw has joined #openstack-ironic | 04:10 | |
| *** Ng has joined #openstack-ironic | 04:10 | |
| *** h00327910_ has joined #openstack-ironic | 04:10 | |
| *** hblixt has quit IRC | 04:11 | |
| *** hemna has joined #openstack-ironic | 04:11 | |
| *** hblixt has joined #openstack-ironic | 04:14 | |
| openstackgerrit | Merged openstack/ironic: Updated from global requirements https://review.openstack.org/202279 | 04:14 |
| openstackgerrit | Merged openstack/python-ironicclient: Updated from global requirements https://review.openstack.org/200903 | 04:20 |
| *** cing has quit IRC | 04:27 | |
| *** hemna has quit IRC | 04:30 | |
| *** Nisha_away has joined #openstack-ironic | 04:33 | |
| *** Nisha has quit IRC | 04:33 | |
| *** rameshg87 has joined #openstack-ironic | 04:34 | |
| *** cing has joined #openstack-ironic | 04:35 | |
| *** sandhya has joined #openstack-ironic | 04:46 | |
| *** yog__ has joined #openstack-ironic | 04:47 | |
| openstackgerrit | Ramakrishnan G proposed openstack/ironic: Refactor pxe - New PXEBoot and ISCSIDeploy interfaces https://review.openstack.org/166513 | 04:53 |
| openstackgerrit | Ramakrishnan G proposed openstack/ironic: Refactor pxe - New PXEBoot and ISCSIDeploy interfaces https://review.openstack.org/166513 | 04:53 |
| openstackgerrit | Ramakrishnan G proposed openstack/ironic: Refactor agent driver with pxe boot interface https://review.openstack.org/166521 | 05:07 |
| *** hblixt_ has joined #openstack-ironic | 05:08 | |
| *** hblixt has quit IRC | 05:11 | |
| *** bigjools has quit IRC | 05:14 | |
| *** killer_prince has joined #openstack-ironic | 05:16 | |
| *** lazy_prince has quit IRC | 05:19 | |
| *** bigjools has joined #openstack-ironic | 05:27 | |
| *** praneshp has joined #openstack-ironic | 05:37 | |
| *** itamarl has joined #openstack-ironic | 05:39 | |
| *** praneshp_ has joined #openstack-ironic | 05:40 | |
| *** praneshp has quit IRC | 05:42 | |
| *** praneshp_ is now known as praneshp | 05:42 | |
| *** vishwana_ has joined #openstack-ironic | 05:43 | |
| *** vishwanathj has quit IRC | 05:47 | |
| *** puranamr has joined #openstack-ironic | 05:55 | |
| *** puranamr has quit IRC | 05:57 | |
| *** dlpartain has joined #openstack-ironic | 05:57 | |
| *** dlpartain has left #openstack-ironic | 05:57 | |
| *** dlpartain has joined #openstack-ironic | 05:58 | |
| *** dlpartain has quit IRC | 05:58 | |
| *** lazy_prince has joined #openstack-ironic | 05:59 | |
| *** e0ne has joined #openstack-ironic | 06:00 | |
| *** yuikotakada has joined #openstack-ironic | 06:01 | |
| *** killer_prince has quit IRC | 06:02 | |
| *** e0ne has quit IRC | 06:02 | |
| *** saripurigopi has joined #openstack-ironic | 06:07 | |
| saripurigopi | hello Ironic | 06:07 |
| yuikotakada | saripurigopi, hi :) | 06:08 |
| saripurigopi | yuikotakada: o/ | 06:10 |
| *** degorenko has quit IRC | 06:10 | |
| *** Sukhdev has joined #openstack-ironic | 06:11 | |
| *** ifarkas has joined #openstack-ironic | 06:14 | |
| *** max_lobur has joined #openstack-ironic | 06:23 | |
| *** puranamr has joined #openstack-ironic | 06:26 | |
| *** praneshp has quit IRC | 06:39 | |
| *** praneshp has joined #openstack-ironic | 06:45 | |
| *** Sukhdev has quit IRC | 06:45 | |
| *** pal has quit IRC | 06:51 | |
| *** itamarl_ has joined #openstack-ironic | 06:51 | |
| *** itamarl has quit IRC | 06:52 | |
| *** itamarl_ is now known as itamarl | 06:52 | |
| *** ifarkas has quit IRC | 06:52 | |
| *** ifarkas has joined #openstack-ironic | 06:52 | |
| *** max_lobur has quit IRC | 06:53 | |
| *** puranamr has quit IRC | 06:55 | |
| *** dtantsur|afk is now known as dtantsur | 06:56 | |
| dtantsur | Morning folks! | 06:56 |
| yuikotakada | dtantsur, o/ | 06:56 |
| dtantsur | o/ | 06:56 |
| *** pal has joined #openstack-ironic | 07:05 | |
| *** itamarl has quit IRC | 07:18 | |
| *** itamarl has joined #openstack-ironic | 07:20 | |
| *** romainh has joined #openstack-ironic | 07:20 | |
| *** foexle has joined #openstack-ironic | 07:27 | |
| *** itamarl has quit IRC | 07:34 | |
| saripurigopi | dtantsur: o/ | 07:35 |
| dtantsur | o/ | 07:35 |
| *** itamarl has joined #openstack-ironic | 07:35 | |
| *** max_lobur has joined #openstack-ironic | 07:37 | |
| *** itamarl has quit IRC | 07:41 | |
| *** pal_ has joined #openstack-ironic | 07:42 | |
| *** stendulker has joined #openstack-ironic | 07:42 | |
| *** pal has quit IRC | 07:44 | |
| dtantsur | Haomeng, o/ could you please review/approve https://review.openstack.org/#/c/194722/ ? | 07:45 |
| dtantsur | 2x +2 already, bunch of +1 | 07:45 |
| *** yuikotak_ has joined #openstack-ironic | 07:45 | |
| *** dlpartain has joined #openstack-ironic | 07:45 | |
| *** achanda has quit IRC | 07:46 | |
| *** yuikotakada has quit IRC | 07:47 | |
| Haomeng | dtantsur: sure, let me have a look | 07:49 |
| Nisha_away | dtantsur, inspector when used as a devstack plugin will place the ramdisk and kernel at what location ? | 07:51 |
| *** itamarl has joined #openstack-ironic | 07:52 | |
| dtantsur | Nisha_away, the same as ironic one, lemme find | 07:52 |
| dtantsur | Nisha_away, /opt/stack/data/ironic/tftpboot/ | 07:53 |
| Haomeng | dtantsur: Ramakrishnan think that his comments can be addressed with follow-up patch, so +a now:) | 07:58 |
| dtantsur | Haomeng, thanks! | 07:59 |
| * dtantsur adds a reminder to address follow-up comments | 07:59 | |
| Haomeng | dtantsur: yw:) | 07:59 |
| *** lucasagomes has joined #openstack-ironic | 08:00 | |
| *** dlpartain has quit IRC | 08:00 | |
| *** romcheg has joined #openstack-ironic | 08:01 | |
| *** pal_ has quit IRC | 08:07 | |
| *** UForgotten has quit IRC | 08:08 | |
| *** UForgotten has joined #openstack-ironic | 08:10 | |
| openstackgerrit | Merged openstack/ironic-python-agent: Updated from global requirements https://review.openstack.org/200674 | 08:11 |
| openstackgerrit | Merged openstack/ironic: Remove deprecated code for driver vendor passthru https://review.openstack.org/200547 | 08:12 |
| openstackgerrit | Merged openstack/ironic: Replace common.fileutils with oslo_utils.fileutils https://review.openstack.org/201397 | 08:13 |
| *** enikanorov_ has joined #openstack-ironic | 08:15 | |
| *** jistr has joined #openstack-ironic | 08:16 | |
| *** itamarl has quit IRC | 08:17 | |
| *** enikanorov has quit IRC | 08:17 | |
| *** pal has joined #openstack-ironic | 08:18 | |
| *** itamarl has joined #openstack-ironic | 08:20 | |
| *** derekh has joined #openstack-ironic | 08:27 | |
| *** ifarkas has quit IRC | 08:30 | |
| openstackgerrit | Cameron.C proposed openstack/ironic: Inspector driver in standalone mode doesn't work https://review.openstack.org/202435 | 08:30 |
| *** ifarkas has joined #openstack-ironic | 08:31 | |
| *** dlpartain has joined #openstack-ironic | 08:32 | |
| *** boris-42 has quit IRC | 08:32 | |
| *** dlpartain has left #openstack-ironic | 08:32 | |
| *** itamarl has quit IRC | 08:37 | |
| *** saripurigopi has quit IRC | 08:38 | |
| *** itamarl has joined #openstack-ironic | 08:47 | |
| *** tiagogomes_ has joined #openstack-ironic | 08:47 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic: Allow abort for {CLEAN,DEPLOY}WAIT states https://review.openstack.org/201552 | 08:48 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic: Add CLEANWAIT state https://review.openstack.org/200152 | 08:48 |
| openstackgerrit | Merged openstack/ironic: Start using new ENROLL state https://review.openstack.org/194722 | 08:50 |
| tiagogomes_ | hi all, what is the status of integrating Cinder with Ironic? | 08:55 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic: Allow abort for {CLEAN,DEPLOY}WAIT states https://review.openstack.org/201552 | 08:59 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic: Add CLEANWAIT state https://review.openstack.org/200152 | 08:59 |
| lucasagomes | tiagogomes_, we currently have no integration. There are some specs/ideas around it | 09:00 |
| *** foexle has quit IRC | 09:00 | |
| lucasagomes | it was also discussed in the vancouver summit, both teams are aligned about what needs to be done (re attaching a volume and booting from volume) | 09:00 |
| lucasagomes | but I don't know if there's someone actually working on it | 09:00 |
| *** foexle has joined #openstack-ironic | 09:01 | |
| tiagogomes_ | lucasagomes ok, thanks | 09:02 |
| lucasagomes | there's an etherpad,trying to find the link | 09:02 |
| *** killer_prince has joined #openstack-ironic | 09:03 | |
| lucasagomes | tiagogomes_, https://etherpad.openstack.org/p/ironic-and-cinder | 09:04 |
| *** yonglihe has quit IRC | 09:06 | |
| *** sandhya has quit IRC | 09:06 | |
| *** lazy_prince has quit IRC | 09:06 | |
| *** lazy_prince has joined #openstack-ironic | 09:07 | |
| *** killer_prince has quit IRC | 09:08 | |
| openstackgerrit | Cameron.C proposed openstack/ironic: Allow inspector driver to work in standalone mode https://review.openstack.org/202435 | 09:09 |
| *** rvasilets_ has left #openstack-ironic | 09:12 | |
| *** ndipanov has joined #openstack-ironic | 09:14 | |
| tiagogomes_ | thanks lucasagomes. So currently, is there a nice way to update the baremetal OS without destroying its user data? | 09:18 |
| *** athomas has quit IRC | 09:20 | |
| *** dtantsur is now known as dtantsur|brb | 09:24 | |
| *** alex_xu_ is now known as alex_xu | 09:27 | |
| *** athomas has joined #openstack-ironic | 09:27 | |
| lucasagomes | tiagogomes_, I wouldn't say nice... But the way tripleo does it is by suing ephemeral partition | 09:32 |
| lucasagomes | you can preserve an ephemeral partition when you rebuild the instance | 09:32 |
| *** yog__ has quit IRC | 09:32 | |
| lucasagomes | using* | 09:33 |
| openstackgerrit | Roman Podoliaka proposed openstack/ironic: db: use new EngineFacade feature of oslo.db https://review.openstack.org/191801 | 09:33 |
| *** ukalifon1 has joined #openstack-ironic | 09:35 | |
| *** lazy_prince has quit IRC | 09:39 | |
| openstackgerrit | Merged openstack/ironic: Remove requirements.txt from tox.ini deps https://review.openstack.org/202349 | 09:41 |
| *** degorenko has joined #openstack-ironic | 09:46 | |
| *** Nisha_away has quit IRC | 09:49 | |
| sambetts | Morning all | 09:50 |
| yuikotak_ | sambetts, o/ | 09:50 |
| sambetts | yuikotak_: o/ Hey1 | 09:50 |
| sambetts | Hey! | 09:50 |
| *** yuikotak_ is now known as yuikotakada | 09:51 | |
| *** naohirot has quit IRC | 09:54 | |
| lucasagomes | sambetts, yuikotakada good ugt morning | 09:56 |
| *** yog__ has joined #openstack-ironic | 09:56 | |
| tiagogomes_ | lucasagomes won't node cleaning erase all the partitions? | 09:56 |
| sambetts | Hey lucasagomes | 09:57 |
| lucasagomes | tiagogomes_, it will but only when the instance is unprovisioned | 09:57 |
| lucasagomes | tiagogomes_, rebuild happens with an ACTIVE instance and it goes to deploying again | 09:57 |
| lucasagomes | it doesn't pass through cleaning | 09:58 |
| *** pal has quit IRC | 09:59 | |
| *** praneshp has quit IRC | 09:59 | |
| *** e0ne has joined #openstack-ironic | 10:02 | |
| *** zhenguo has quit IRC | 10:03 | |
| tiagogomes_ | lucasagomes ah, so will ` nova rebuild --preserver-ephemeral` not format the partitions on the baremetal? | 10:07 |
| lucasagomes | tiagogomes_, nop, it will keep the ephemeral one | 10:07 |
| tiagogomes_ | lucasagomes ok thanks! | 10:08 |
| lucasagomes | it will basically replace the OS image with a new one | 10:08 |
| tiagogomes_ | overwriting existing files? | 10:11 |
| *** lazy_prince has joined #openstack-ironic | 10:12 | |
| openstackgerrit | Shivanand Tendulker proposed stackforge/proliantutils: Add RIS support for updating boot device https://review.openstack.org/201420 | 10:14 |
| *** pal has joined #openstack-ironic | 10:17 | |
| lucasagomes | tiagogomes_, it just copy the image onto the root partition | 10:19 |
| *** e0ne is now known as e0ne_ | 10:25 | |
| *** e0ne_ is now known as e0ne | 10:28 | |
| *** amotoki has joined #openstack-ironic | 10:32 | |
| *** stendulker has quit IRC | 10:34 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Allow 'abort' verb for node-set-provision-state https://review.openstack.org/202184 | 10:45 |
| *** coolsvap has quit IRC | 10:58 | |
| *** coolsvap has joined #openstack-ironic | 10:58 | |
| *** yog__ has quit IRC | 11:07 | |
| rameshg87 | lucasagomes: h | 11:08 |
| rameshg87 | hi | 11:08 |
| lucasagomes | rameshg87, hi there | 11:11 |
| rameshg87 | lucasagomes: regarding https://review.openstack.org/#/c/193439/23/ironic/api/controllers/v1/port.py | 11:12 |
| TheJulia | Good morning | 11:12 |
| lucasagomes | TheJulia, good ugt morning | 11:12 |
| rameshg87 | lucasagomes: that patch was actually meant just to allow node name also to be passed while creating a new port, right ? | 11:12 |
| lucasagomes | rameshg87, I think that is the intention, use the node name to be able to create a port | 11:13 |
| lucasagomes | but it will later be translated into the node's uuid | 11:13 |
| lucasagomes | that's what I got from that patch at least | 11:13 |
| rameshg87 | lucasagomes: yeah, so why do we really need to deprecate the node_uuid ? | 11:13 |
| *** e0ne is now known as e0ne_ | 11:13 | |
| lucasagomes | rameshg87, the patch says it does "and deprecate 'node_uuid'." | 11:14 |
| lucasagomes | in the commit message | 11:14 |
| lucasagomes | maybe we should just allow one inputing the name in "node_uuid" | 11:14 |
| rameshg87 | lucasagomes: oh, but that conflicts in mind (_uuid accepting something not a uuid) :) | 11:15 |
| lucasagomes | yeah, that's why he added "node" | 11:15 |
| lucasagomes | my point is that by keeping node_uuid "deprecated" doesn't work for API | 11:15 |
| lucasagomes | because the user can't see it's deprecated | 11:15 |
| *** Haomeng|2 has joined #openstack-ironic | 11:15 | |
| rameshg87 | lucasagomes: yeah | 11:16 |
| rameshg87 | lucasagomes: and as per that patch it is supposed to allow only while POSTing | 11:16 |
| lucasagomes | rameshg87, yeah that's my understanding | 11:16 |
| rameshg87 | lucasagomes: so I suggested that *hack* to just put getter for node attribute return wtypes.Unset always, so that it's never returned, but always accepted in POST | 11:17 |
| rameshg87 | but I don't think it's a neat way | 11:17 |
| lucasagomes | I see the value on it, when scripting it may be easier to do a "node-create -name 'blah'" "port-create -n 'blah'" | 11:17 |
| rameshg87 | yeah, | 11:17 |
| lucasagomes | rameshg87, yeah :-/ it's hacky... with the API microversions I can see we remoiving node_uuid in favor of node | 11:18 |
| rameshg87 | lucasagomes: but what what do we really return for node when it's a GET ? | 11:18 |
| lucasagomes | as a break change, but we need to gather info see if this is a change that people wants to do | 11:18 |
| rameshg87 | lucasagomes: a uuid or a name ? | 11:18 |
| *** Haomeng has quit IRC | 11:18 | |
| *** thrash|g0ne is now known as thrash | 11:18 | |
| lucasagomes | rameshg87, hmm it's hard. The thing I like about the UUID is that it doesn't change | 11:19 |
| lucasagomes | that's the canonical identification for a resource | 11:19 |
| lucasagomes | the name on the other hand is complicated | 11:19 |
| *** pal has quit IRC | 11:19 | |
| lucasagomes | cause it can change | 11:19 |
| lucasagomes | rameshg87, I haven't thought deeply about that change tho | 11:20 |
| *** yog__ has joined #openstack-ironic | 11:20 | |
| lucasagomes | maybe we should bring it up on the next meeting and discuss it with more people | 11:20 |
| rameshg87 | lucasagomes: yeah, I think so | 11:20 |
| lucasagomes | cause I can see some benefit with it, but again... it's not like a huge thing... | 11:21 |
| rameshg87 | lucasagomes: to not to cause such a confusion, I kind of thought we can add a new node_name attribute where we can both get and set values with it | 11:21 |
| lucasagomes | right, yeah that may be a way then | 11:21 |
| *** sambetts has quit IRC | 11:21 | |
| lucasagomes | have both there | 11:21 |
| lucasagomes | and in the db we just keep the numeral id | 11:21 |
| lucasagomes | numeric* | 11:22 |
| rameshg87 | lucasagomes: yeah, it's just in the API, returning current values | 11:22 |
| rameshg87 | exactly | 11:22 |
| rameshg87 | people who want to keep seeing names, let them see names | 11:22 |
| lucasagomes | yeah | 11:22 |
| rameshg87 | people who want to continue seeing uuids, they are not affected | 11:22 |
| lucasagomes | indeed | 11:22 |
| rameshg87 | anyway, let's wait for other's opinions :) | 11:22 |
| rameshg87 | lucasagomes: thanks for the feedback | 11:23 |
| rameshg87 | need to leave now .. | 11:23 |
| * rameshg87 goes home | 11:23 | |
| lucasagomes | rameshg87, thank you! I will ocmment on the patch | 11:23 |
| *** rameshg87 has quit IRC | 11:23 | |
| *** dtantsur|brb is now known as dtantsur | 11:24 | |
| *** lucasagomes is now known as lucas-hungry | 11:25 | |
| *** hblixt_ has quit IRC | 11:30 | |
| *** e0ne_ is now known as e0ne | 11:30 | |
| *** sambetts has joined #openstack-ironic | 11:37 | |
| *** Haomeng has joined #openstack-ironic | 11:51 | |
| *** Haomeng|2 has quit IRC | 11:53 | |
| *** lazy_prince has quit IRC | 11:59 | |
| *** lazy_prince has joined #openstack-ironic | 12:01 | |
| *** cing has quit IRC | 12:02 | |
| *** dprince has joined #openstack-ironic | 12:02 | |
| *** amotoki has quit IRC | 12:04 | |
| *** ifarkas has quit IRC | 12:06 | |
| *** ifarkas has joined #openstack-ironic | 12:09 | |
| *** [1]cdearborn has joined #openstack-ironic | 12:17 | |
| *** jcoufal has joined #openstack-ironic | 12:19 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic-inspector: Create a handler for uncaught 404 errors https://review.openstack.org/202532 | 12:30 |
| *** jjohnson2 has joined #openstack-ironic | 12:34 | |
| *** trown|outttypeww is now known as trown | 12:42 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic-inspector: Provide more meaningful message for error 500 https://review.openstack.org/202535 | 12:42 |
| *** krtaylor has quit IRC | 12:44 | |
| *** jjohnson2 has quit IRC | 12:46 | |
| *** lucas-hungry is now known as lucasagomes | 12:49 | |
| *** jcoufal has quit IRC | 12:50 | |
| lucasagomes | jroll, http://mjg59.dreamwidth.org/35969.html should we worry? (re canonical IP and distributing IPA with ubuntu) | 12:50 |
| *** krtaylor has joined #openstack-ironic | 12:51 | |
| *** mjturek1 has joined #openstack-ironic | 12:52 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: Address minor comments on the ENROLL patch https://review.openstack.org/202540 | 12:53 |
| *** jcoufal has joined #openstack-ironic | 12:55 | |
| *** jcoufal has quit IRC | 12:59 | |
| *** zhenguo has joined #openstack-ironic | 13:01 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: DO NOT MERGE https://review.openstack.org/202544 | 13:01 |
| *** jjohnson2 has joined #openstack-ironic | 13:03 | |
| *** pal has joined #openstack-ironic | 13:05 | |
| *** pal_ has joined #openstack-ironic | 13:08 | |
| *** pal has quit IRC | 13:10 | |
| *** jjohnson2 has quit IRC | 13:10 | |
| *** coolsvap is now known as coolsvap|away | 13:13 | |
| *** bnemec has joined #openstack-ironic | 13:14 | |
| *** bizarrochristy has joined #openstack-ironic | 13:22 | |
| NobodyCam | good morning ironicers | 13:25 |
| lucasagomes | NobodyCam, good ugt morning | 13:26 |
| yuikotakada | NobodyCam, o/ | 13:27 |
| yuikotakada | lucasagomes, o/ | 13:27 |
| BadCub | morning folks | 13:29 |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: Allow upgrading shared mock to an exclusive one https://review.openstack.org/202558 | 13:29 |
| sambetts | o/ BadCub | 13:30 |
| NobodyCam | good ugt morning lucasagomes yuikotakada dtantsur sambetts jroll and everyone not mentioned | 13:30 |
| dtantsur | morning NobodyCam, BadCub, jroll, TheJulia, sambetts, and everyone else :) | 13:30 |
| NobodyCam | hehehe :) | 13:30 |
| BadCub | heya | 13:30 |
| sambetts | NobodyCam o/ | 13:30 |
| NobodyCam | :) | 13:32 |
| *** jcoufal has joined #openstack-ironic | 13:33 | |
| *** jaypipes has joined #openstack-ironic | 13:35 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: [WIP] Only take exclusive lock in sync_power_state if node is updated https://review.openstack.org/202562 | 13:39 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Fix version negotiation https://review.openstack.org/202565 | 13:41 |
| lucasagomes | NobodyCam, dtantsur devananda ^ funny one for you guys | 13:41 |
| dtantsur | oh lol | 13:42 |
| * lucasagomes fixes pep8 | 13:42 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Fix version negotiation https://review.openstack.org/202565 | 13:42 |
| dtantsur | In the meanwhile I've started converting https://etherpad.openstack.org/p/ironic-locking-reform to patches :) | 13:43 |
| devananda | g'morning! | 13:43 |
| dtantsur | devananda, morning | 13:43 |
| devananda | lucasagomes, dtantsur: either if you coming to the midcycle? | 13:44 |
| *** e0ne is now known as e0ne_ | 13:44 | |
| dtantsur | devananda, not me | 13:44 |
| lucasagomes | devananda, I don't think so :-( | 13:44 |
| devananda | :-( | 13:44 |
| lucasagomes | devananda, good morning | 13:44 |
| dtantsur | were it not for 2 long travels to the summit this year... | 13:44 |
| dtantsur | anyway, could you folks have a quick look at https://review.openstack.org/#/c/202562/ despite it being wip? | 13:44 |
| dtantsur | just wanna sanity-check the idea | 13:45 |
| * dtantsur goes afk for an hour | 13:45 | |
| jroll | morning all :) | 13:45 |
| *** jjohnson2 has joined #openstack-ironic | 13:46 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: [WIP] Only take exclusive lock in sync_power_state if node is updated https://review.openstack.org/202562 | 13:46 |
| lucasagomes | jroll, morning | 13:47 |
| jroll | lucasagomes: holy cow that port-by-node-name patch | 13:48 |
| lucasagomes | jroll, heh I know | 13:48 |
| *** e0ne_ is now known as e0ne | 13:49 | |
| NobodyCam | good morning devananda | 13:49 |
| *** jcoufal has quit IRC | 13:52 | |
| *** jjohnson2 has quit IRC | 13:52 | |
| lucasagomes | jroll, btw take also a look at http://mjg59.dreamwidth.org/35969.html I worry about us building IPA on every commit and making it available due the ubuntu IP | 13:52 |
| lucasagomes | canonical IP* | 13:53 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Fix version negotiation https://review.openstack.org/202565 | 13:54 |
| jroll | lucasagomes: oh my god | 13:54 |
| * jroll rages | 13:54 | |
| jroll | mordred: ^^ thoughts? | 13:54 |
| jroll | we can hit the legal list if needed | 13:54 |
| lucasagomes | I want to talk to you before going to -infra | 13:54 |
| lucasagomes | yeah | 13:54 |
| jroll | yeah, let's bounce over there | 13:55 |
| lucasagomes | because we make it available I wonder that it's considered "distributing", tho I believe so | 13:55 |
| jroll | I believe so too | 13:55 |
| jroll | it's a coreos image with a ubuntu container inside | 13:55 |
| lucasagomes | exactly | 13:56 |
| devananda | oh srsly | 13:57 |
| devananda | yah, we'll almost definitely need to stop that, even just because ... | 13:57 |
| jroll | soooo who is up for converting that to fedora? | 13:59 |
| jroll | and/or debian | 13:59 |
| lucasagomes | centos | 13:59 |
| lucasagomes | ++ | 13:59 |
| jroll | I'm going to push a patch to switch it to debian | 13:59 |
| jroll | just because I don't have to think about yum | 13:59 |
| jroll | lucasagomes: wanna file a bug? | 14:00 |
| lucasagomes | jroll, sure will do | 14:00 |
| trown | jroll: I have been hacking on the DIB element for fedora/centos | 14:00 |
| NobodyCam | lucasagomes: why did you take out 1.10 from test | 14:00 |
| trown | still WIP though | 14:00 |
| lucasagomes | NobodyCam, it wasn't actually testing anything with that 1.10, it was left over | 14:00 |
| lucasagomes | NobodyCam, + that version of the client the default version is 1.8 | 14:01 |
| jroll | trown: I'm (personally) not really interested in DIB, and really the coreos images we provide are the concern here | 14:01 |
| * TheJulia reads the post and feels very sad | 14:01 | |
| jroll | trown: but I appreciate the work :) | 14:01 |
| NobodyCam | but testing for 1.10 > 1.8, at least to me is a valid test | 14:02 |
| openstackgerrit | Roman Podoliaka proposed openstack/ironic: db: use new EngineFacade feature of oslo.db https://review.openstack.org/191801 | 14:03 |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic-python-agent: Change Dockerfile to use Debian as a base https://review.openstack.org/202578 | 14:05 |
| jroll | let's see how that goes | 14:05 |
| *** kkoski has joined #openstack-ironic | 14:06 | |
| lucasagomes | jroll, https://bugs.launchpad.net/ironic/+bug/1475325 | 14:07 |
| openstack | Launchpad bug 1475325 in Ironic "IPA: Stop using ubuntu by default as the base OS" [Undecided,New] | 14:07 |
| lucasagomes | NobodyCam, right, I can add it back if u want | 14:08 |
| jroll | thanks lucasagomes | 14:08 |
| * jroll adds | 14:08 | |
| NobodyCam | lucasagomes: only becayse it was oe of the examples you point out in the commit message | 14:08 |
| lucasagomes | NobodyCam, ++ ok | 14:08 |
| NobodyCam | :) | 14:08 |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic-python-agent: Change Dockerfile to use Debian as a base https://review.openstack.org/202578 | 14:08 |
| mordred | jroll, lucasagomes: moving to debian - fine by me -however, I do want to repeat what I said in infra, that I do _not_ think there is an actual problem here | 14:10 |
| mordred | that said - I think moving to debian is an awesome choice | 14:11 |
| NobodyCam | mordred: thank you great info here and what in -infra :) | 14:14 |
| mordred | NobodyCam: my pleasure! | 14:15 |
| NobodyCam | :) | 14:15 |
| lucasagomes | yeah thanks | 14:15 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Fix version negotiation https://review.openstack.org/202565 | 14:29 |
| *** pal has joined #openstack-ironic | 14:29 | |
| lucasagomes | NobodyCam, ^ that test would actually fail with the old code | 14:29 |
| *** pal_ has quit IRC | 14:30 | |
| lucasagomes | In [8]: min('99.99', '1.6') | 14:31 |
| lucasagomes | Out[8]: '1.6' | 14:31 |
| lucasagomes | In [9]: min('1.10', '1.6') | 14:31 |
| lucasagomes | Out[9]: '1.10' | 14:31 |
| lucasagomes | just for ref | 14:31 |
| jroll | hmm, when did we add py34 to the ipa dockerfile? | 14:36 |
| jroll | also E: Unable to locate package python3.4 | 14:36 |
| jroll | with debian | 14:36 |
| * lucasagomes no idea | 14:36 | |
| *** dlpartain has joined #openstack-ironic | 14:36 | |
| lucasagomes | currently IPA doesn't even work with py3, so I believe it can be deleted | 14:36 |
| *** e0ne is now known as e0ne_ | 14:36 | |
| jroll | oh heh | 14:37 |
| jroll | we purge python3, not install it | 14:37 |
| lucasagomes | oh it's purge yeah | 14:38 |
| lucasagomes | it's grand to not list it there | 14:38 |
| jroll | yeah, doing it | 14:39 |
| jroll | testing before I post it | 14:39 |
| NobodyCam | lucasagomes: ya | 14:39 |
| NobodyCam | lucasagomes: hummm: http://logs.openstack.org/65/202565/4/check-tripleo/gate-tripleo-ironic-undercloud-precise-nonha/811bd26/console.html#_2015-07-16_14_36_42_686 | 14:41 |
| jroll | surprise, tripleo ci is still broke as heck | 14:42 |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic-python-agent: Change Dockerfile to use Debian as a base https://review.openstack.org/202578 | 14:42 |
| jroll | lucasagomes: ^ that built fine for me, I haven't ran it but don't expect surprises | 14:42 |
| lucasagomes | NobodyCam, yeah ooo CI has been broken | 14:42 |
| lucasagomes | for a while actually | 14:43 |
| NobodyCam | :( | 14:43 |
| lucasagomes | jroll, w00t! the gate will build from source and run it for us | 14:43 |
| jroll | lucasagomes: indeed :) | 14:43 |
| *** itamarl has left #openstack-ironic | 14:44 | |
| lucasagomes | NobodyCam, lemme ping #tripleo | 14:47 |
| *** cing has joined #openstack-ironic | 14:49 | |
| *** dlpartain has quit IRC | 14:50 | |
| *** pal_ has joined #openstack-ironic | 14:59 | |
| *** pal has quit IRC | 14:59 | |
| *** cing has quit IRC | 15:00 | |
| *** karimb has joined #openstack-ironic | 15:02 | |
| *** pal_ has quit IRC | 15:05 | |
| *** e0ne_ is now known as e0ne | 15:05 | |
| *** coolsvap|away is now known as coolsvap | 15:06 | |
| *** pal has joined #openstack-ironic | 15:07 | |
| *** amotoki has joined #openstack-ironic | 15:10 | |
| *** absubram has joined #openstack-ironic | 15:11 | |
| dtantsur | so about hardware: got a report from QE about HP Proliant DL170e G6 where power on via pxe_ipmitool takes around 25 seconds of exclusive locking | 15:14 |
| *** pal has quit IRC | 15:14 | |
| dtantsur | which is enough to get some troubles to the next command in the script, if it checks for "reservation" without using task_manager... | 15:15 |
| *** pal has joined #openstack-ironic | 15:17 | |
| *** hemna has joined #openstack-ironic | 15:18 | |
| *** trown is now known as trown|brb | 15:19 | |
| *** hemna has quit IRC | 15:20 | |
| *** trown|brb is now known as trown | 15:24 | |
| *** pal_ has joined #openstack-ironic | 15:24 | |
| *** pal has quit IRC | 15:25 | |
| devananda | dtantsur: holding the lock internally? or blocking the conductor process / api response? | 15:25 |
| devananda | dtantsur: also yes, we should always assume that any actual call to the BMC might be VERY SLOW | 15:25 |
| *** e0ne has quit IRC | 15:26 | |
| *** e0ne has joined #openstack-ironic | 15:26 | |
| devananda | dtantsur: but I see your point -- anything could have changed in 25 seconds if a non-exclusive lock was used | 15:26 |
| *** e0ne has quit IRC | 15:27 | |
| *** e0ne has joined #openstack-ironic | 15:27 | |
| *** e0ne has quit IRC | 15:28 | |
| *** _________ has joined #openstack-ironic | 15:30 | |
| *** _________ has quit IRC | 15:31 | |
| *** e0ne has joined #openstack-ironic | 15:32 | |
| *** yog__ has quit IRC | 15:34 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/python-ironicclient: Allow 'abort' verb for node-set-provision-state https://review.openstack.org/202184 | 15:35 |
| *** pal has joined #openstack-ironic | 15:36 | |
| *** pal_ has quit IRC | 15:36 | |
| dtantsur | yeah, right. | 15:37 |
| lucasagomes | people can you look and post ur opnions at https://review.openstack.org/#/c/193495/ ? | 15:38 |
| lucasagomes | tl;dr it's basically renaming pxe.http_root to deploy.http_server_root (same for http_url). While changing groups from 'pxe' to 'deploy' is fine because other drivers can use the same http server | 15:39 |
| lucasagomes | I find it the name inconsistent with other config options such as tftp_root | 15:39 |
| lucasagomes | anyway, if you have a time please take a look and post ur opnions there | 15:39 |
| *** ig0r_ has quit IRC | 15:40 | |
| * dtantsur agrees with http_root | 15:41 | |
| *** dprince has quit IRC | 15:42 | |
| *** ifarkas has quit IRC | 15:43 | |
| * TheJulia ponders | 15:45 | |
| *** vishwanathj has joined #openstack-ironic | 15:45 | |
| lucasagomes | folks and if u have more free time, please take a look https://review.openstack.org/#/c/197141/ :-) | 15:47 |
| lucasagomes | the fix for the agent_* drivers are already merged | 15:47 |
| lucasagomes | is* | 15:47 |
| *** vishwana_ has quit IRC | 15:47 | |
| jroll | devananda: btw, the ENROLL change landed, so much for waiting until after a release :P | 15:49 |
| devananda | urgh | 15:49 |
| devananda | i knew i should have -2'd that | 15:49 |
| *** pal_ has joined #openstack-ironic | 15:50 | |
| devananda | oh jeez | 15:50 |
| devananda | i just found the tab | 15:50 |
| lucasagomes | devananda, jroll yeah I noticed it was merged this morning... The idea was to do a release prior to merging it? | 15:51 |
| devananda | i have a comment written and just didn't hit send :( | 15:51 |
| * devananda fails | 15:51 | |
| devananda | lucasagomes: yes | 15:51 |
| jroll | lol | 15:51 |
| lucasagomes | oh lol | 15:51 |
| devananda | The change itself LGTM, however, I would like us to hold it back until: | 15:51 |
| devananda | - we fully embrace the new release process | 15:51 |
| openstackgerrit | Satoru Moriya proposed openstack/ironic-specs: Add attributes about volume conneciton into nodes table https://review.openstack.org/200496 | 15:51 |
| devananda | - we do a server and client release that are API compatible with Kilo | 15:51 |
| devananda | - then roll this (and perhaps other API incompatible changes) into the subsequent server release, along with a major release version bump | 15:51 |
| *** pal has quit IRC | 15:51 | |
| devananda | is what i THOUGHT I POSTED :( | 15:51 |
| dtantsur | devananda, eek? you didn't want to release ENROLL? | 15:51 |
| devananda | not yet | 15:52 |
| dtantsur | but why, we're still backward compatible | 15:52 |
| jroll | yeah, I don't see a problem with releasing this now, I just remembered devananda mentioning he didn't want to | 15:52 |
| devananda | because it isn't | 15:52 |
| jroll | it is | 15:52 |
| jroll | versioning | 15:52 |
| devananda | see my updates to the versioning spec. i know ya'll dont agree yet though | 15:53 |
| jroll | it isn't compatible for folks using 'latest' | 15:53 |
| dtantsur | devananda, sorry, I don't get you. We got through all this versioning fun to make sure we're always and absolutely compatible | 15:53 |
| jroll | link? | 15:53 |
| devananda | it isn't compatible the moment we release a new client | 15:53 |
| devananda | that starts using the new version # | 15:53 |
| lucasagomes | jroll, https://review.openstack.org/#/c/196320/ | 15:53 |
| devananda | which we all know | 15:53 |
| dtantsur | devananda, on another note, what prevents you from tagging release from a previous commit? | 15:53 |
| devananda | is going to break people | 15:53 |
| devananda | dtantsur: good question ... | 15:54 |
| jroll | devananda: release notes, or there's the option of, ya know, not defaulting to the highest possible version | 15:54 |
| dtantsur | devananda, I dunno how it works for you, but for inspector relase is $ git push gerrit TAG | 15:54 |
| jroll | I still think the client should require a version from the user | 15:54 |
| dtantsur | which means it's not necessary HEAD | 15:54 |
| devananda | dhellmann: is it possible to tag a release at an (arbitrary, recent, not tip-of-master) commit SHA ? | 15:54 |
| devananda | dtantsur: yea, there's more automation around release-managed services | 15:55 |
| jroll | devananda: please make sure it includes https://review.openstack.org/#/c/200153/ whatever you do | 15:55 |
| jroll | that fixes a major bug in upgrades with agent driver | 15:55 |
| devananda | cool | 15:56 |
| jroll | actually, shit, that might not fix it | 15:56 |
| devananda | :( | 15:56 |
| jroll | so here's the thing: | 15:56 |
| jroll | when conductor starts, it assumes anything in DEPLOYING has a lock, it releases it's own locks and kills anything in DEPLOYING that isn't locked | 15:57 |
| jroll | this kills most ongoing agent deploys, until that patch is applied that makes it use DEPLOYWAIT | 15:57 |
| devananda | dtantsur: so the main thing is, i want us to do a release using the new numbering system, post-kilo, that is in fact compatible with kilo | 15:57 |
| jroll | so upgrading to get that fix will also kill deploys in flight :( | 15:58 |
| jroll | devananda: it's totally compatible. the problem is you want to release a *client* that is *by default* compatible with kilo | 15:58 |
| dtantsur | devananda, in a sense of backward compatibility? yeah, got it | 15:58 |
| *** trown is now known as trown|lunch | 15:58 | |
| jroll | devananda: so to accomplish this, you should release tip of ironic + client that defaults to (version before ENROLL) | 15:59 |
| jroll | right? | 15:59 |
| dtantsur | devananda, if there are not other options, I can only suggest quick revert, followed by quick revert of revert after release | 15:59 |
| dtantsur | devananda, what I don't wanna do is to chase after reviewers again :( | 15:59 |
| devananda | dtantsur: totally -- no need for that :) | 15:59 |
| devananda | (chasing reviewers | 15:59 |
| devananda | jroll: the very question of "is this client compatible with that server" is the problem | 16:00 |
| devananda | we should never release a client that isn't compatible *by default* with all server versions we support | 16:02 |
| jroll | devananda: so release a client that doesn't know about ENROLL yet | 16:02 |
| dtantsur | client is compatible, it's user that will break :D | 16:02 |
| devananda | dtantsur: heh, well put | 16:02 |
| lucasagomes | yeah the only problem I forsee with that is if we decide to include the clean stuff in the release | 16:03 |
| lucasagomes | with the abort etc, because that's also bumping the api version | 16:03 |
| *** kozhukalov__ has joined #openstack-ironic | 16:03 | |
| dtantsur | devananda, so you want to release a version that will work as expected even with a new client, right? | 16:03 |
| dtantsur | devananda, so the client would default to 1.11 but server will ask to downgrade, and everything will work as before? | 16:04 |
| lucasagomes | devananda, jroll re version negotiation in the client https://review.openstack.org/#/c/202565/ | 16:04 |
| lucasagomes | it's actually broken | 16:04 |
| devananda | dtantsur: as you said, it's not the client<->server that fails, it's user's expectation | 16:04 |
| devananda | lucasagomes: yea, saw that, waiting for tests | 16:04 |
| *** coolsvap is now known as coolsvap|away | 16:05 | |
| lucasagomes | ack | 16:05 |
| dtantsur | versioning is hard, negotiation is even harder... | 16:05 |
| *** pal_ has quit IRC | 16:05 | |
| *** ijw has quit IRC | 16:07 | |
| devananda | http://specs.openstack.org/openstack/ironic-specs/specs/liberty/enroll-node-state.html | 16:08 |
| devananda | so this was called out several times as a breaking change | 16:08 |
| *** pal has joined #openstack-ironic | 16:09 | |
| devananda | meh. i'm never happy breaking users | 16:10 |
| dtantsur | nobody is.. | 16:10 |
| jroll | sorry, got pulled into a meeting | 16:11 |
| devananda | getting pulled into a meeting now() too :( | 16:11 |
| jroll | so IMO we should release a new version of the server, including enroll. | 16:11 |
| jroll | include really fantastic messaging about the impact of the api version stuff. YOU NEED TO THINK ABOUT THIS etc | 16:12 |
| jroll | if you want to make breaking things harder, release the new client with a default of 1.10 or whatever that will not do ENROLL by default | 16:12 |
| jroll | devananda: dtantsur ^ | 16:13 |
| devananda | lucasagomes: https://review.openstack.org/#/c/196320/ | 16:13 |
| tiagogomes_ | does file injection work on the baremetal? | 16:13 |
| devananda | tiagogomes_: nope | 16:13 |
| jroll | and we also need to update our docs on "how to enroll nodes" | 16:13 |
| jroll | devananda: what, yes it does | 16:13 |
| lucasagomes | devananda, yes? reading it now | 16:14 |
| jroll | tiagogomes_: it should, there was a bug in nova kilo where it did not but I think it was backported | 16:14 |
| devananda | jroll: modifying files in the guest image after deployment? | 16:14 |
| lucasagomes | tiagogomes_, if you mean injecting file in the configdrive yes it does | 16:15 |
| jroll | devananda: ok let's back up | 16:15 |
| jroll | ^ that | 16:15 |
| devananda | configdrive != nova file injection | 16:15 |
| jroll | here was the stable backport https://review.openstack.org/#/c/176396/ | 16:15 |
| tiagogomes_ | without config drive | 16:15 |
| jroll | oh no, not without configdrive | 16:15 |
| * devananda goes afk for reals | 16:15 | |
| lucasagomes | tiagogomes_, so no | 16:15 |
| *** yuikotakada has quit IRC | 16:15 | |
| *** absubram has quit IRC | 16:15 | |
| lucasagomes | jroll, https://review.openstack.org/#/c/197141/ (if you have time) | 16:16 |
| jroll | looking | 16:17 |
| tiagogomes_ | out of curiosity, if I use the config-drive and the OS has cloud-init, will the file be placed where I intended | 16:17 |
| jroll | tiagogomes_: yep :) | 16:18 |
| *** mitchjameson has joined #openstack-ironic | 16:18 | |
| jroll | lucasagomes: +A | 16:18 |
| lucasagomes | w00t! | 16:18 |
| dtantsur | see you tomorrow | 16:23 |
| NobodyCam | night dtantsur | 16:23 |
| dtantsur | lemme know what you folks decide about ENROLL patch | 16:23 |
| *** dtantsur is now known as dtantsur|afk | 16:23 | |
| *** pal has quit IRC | 16:25 | |
| lucasagomes | dtantsur|afk, night | 16:27 |
| TheJulia | goodnight dtantsur|afk | 16:28 |
| *** achanda has joined #openstack-ironic | 16:28 | |
| *** achanda has quit IRC | 16:29 | |
| *** Nisha has joined #openstack-ironic | 16:29 | |
| *** romcheg has quit IRC | 16:31 | |
| *** Nisha has quit IRC | 16:34 | |
| *** Nisha has joined #openstack-ironic | 16:35 | |
| *** achanda has joined #openstack-ironic | 16:35 | |
| *** praneshp has joined #openstack-ironic | 16:42 | |
| *** rwsu has quit IRC | 16:44 | |
| *** Pradip has joined #openstack-ironic | 16:44 | |
| lucasagomes | folks I'm calling it a day here. Have a good night everyone | 16:47 |
| *** lucasagomes is now known as lucas-dinner | 16:47 | |
| NobodyCam | night lucas-dinner | 16:48 |
| *** absubram has joined #openstack-ironic | 16:53 | |
| openstackgerrit | Devananda van der Veen proposed openstack/ironic: Allow upgrading shared lock to an exclusive one https://review.openstack.org/202558 | 16:54 |
| *** romainh has left #openstack-ironic | 16:56 | |
| *** derekh has quit IRC | 17:00 | |
| *** jistr has quit IRC | 17:00 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic: Updated from global requirements https://review.openstack.org/202699 | 17:01 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic-inspector: Updated from global requirements https://review.openstack.org/202700 | 17:01 |
| *** kozhukalov_ has quit IRC | 17:03 | |
| *** kozhukalov__ is now known as kozhukalov_ | 17:03 | |
| *** jcoufal has joined #openstack-ironic | 17:09 | |
| *** e0ne has quit IRC | 17:09 | |
| *** kozhukalov has joined #openstack-ironic | 17:10 | |
| NobodyCam | TheJulia: is this about how far you expected I would get? http://paste.openstack.org/show/BP5m1xQcRxMPTeBbJUgx/ | 17:10 |
| *** absubram has quit IRC | 17:12 | |
| *** krtaylor has quit IRC | 17:13 | |
| TheJulia | sweet | 17:15 |
| TheJulia | NobodyCam: base os missing ansible? | 17:15 |
| TheJulia | err | 17:15 |
| TheJulia | maybe guest? | 17:15 |
| * TheJulia has never really used vagrant | 17:15 | |
| NobodyCam | TheJulia: looking into it now | 17:16 |
| *** kozhukalov has quit IRC | 17:16 | |
| jroll | NobodyCam: looks like your host doesn't have ansible installed | 17:16 |
| NobodyCam | si | 17:16 |
| NobodyCam | :) | 17:16 |
| *** tiagogomes_ has quit IRC | 17:19 | |
| *** praneshp has quit IRC | 17:20 | |
| *** praneshp has joined #openstack-ironic | 17:20 | |
| Pradip | Hi, i was reading this blueprint https://review.openstack.org/#/c/188528/6/specs/liberty/ironic-ml2-integration.rst. I have question. in this context, are the bare metal servers needs to be connected to a switch? I mean can there be scenarios where the bare metals are connected to the controller using hub or direct connection | 17:21 |
| jroll | Pradip: it would need to be a managed switch | 17:21 |
| *** athomas has quit IRC | 17:22 | |
| Pradip | jroll: that's what i thought. thanks | 17:22 |
| jroll | np | 17:22 |
| Pradip | jroll: another thing is how are you planning to identify the provisioning network | 17:22 |
| jroll | Pradip: in the config file | 17:22 |
| jroll | Pradip: see also http://specs.openstack.org/openstack/ironic-specs/specs/liberty/network-provider.html | 17:23 |
| *** praneshp has quit IRC | 17:24 | |
| *** achanda has quit IRC | 17:25 | |
| *** mtanino has joined #openstack-ironic | 17:25 | |
| Pradip | jroll: what i mean is, when the dhcp for the pxe boots comes in, you need to feed it to the provisioning network dhcp, but when you are switching to tenant network it needs to be fed to the tenant dhcp. for that i beleive you will need to tag the packets differently, am i right? | 17:25 |
| *** kozhukalov has joined #openstack-ironic | 17:25 | |
| jroll | Pradip: this will only be supported for instances that boot from disks, we can't rely on being able to control DHCP on the tenant network | 17:26 |
| *** mitchjameson has quit IRC | 17:26 | |
| *** athomas has joined #openstack-ironic | 17:26 | |
| Pradip | jroll: okay but you need the dhcp for the provisioning network, right? | 17:26 |
| jroll | Pradip: yes, ironic will put the node on the provisioning network and configure dhcp | 17:27 |
| jroll | the switch will probably need a dhcp relay set up; that's up to the operator and out of scope here | 17:27 |
| Pradip | jroll: okay got it. the only thing unclear to me is, how the baremetal is gonna get a tenant ip then | 17:28 |
| *** ukalifon1 has quit IRC | 17:28 | |
| Pradip | is the pexe boot going to set it up? | 17:28 |
| Pradip | *pxe boot | 17:28 |
| *** puranamr has joined #openstack-ironic | 17:29 | |
| jroll | Pradip: nova (or the user, in the case without nova) will initially create a neutron port on the tenant network, which allocates the IP. after deployment, ironic will remove the provisioning network port and tell the tenant port to activate | 17:30 |
| jroll | and the configuration may get to the instance via configdrive or metadata service | 17:30 |
| *** karimb has quit IRC | 17:30 | |
| Pradip | jroll: ah got it. thanks a lot for clearing this up :) | 17:30 |
| jroll | yep np :) | 17:31 |
| Pradip | :) | 17:31 |
| *** achanda has joined #openstack-ironic | 17:35 | |
| *** max_lobur has quit IRC | 17:39 | |
| *** achanda has quit IRC | 17:40 | |
| *** davideagnello has quit IRC | 17:40 | |
| *** lazy_prince has quit IRC | 17:41 | |
| *** puranamr has quit IRC | 17:42 | |
| Pradip | jroll: just another question. will the interfaces file on the baremetal will also be properly set during the pxe boot so that it doesn't send any dhcp packets after the bootup | 17:42 |
| devananda | JoshNang: so https://review.openstack.org/#/c/152703/5 seems to be doing not what i expected us to do | 17:42 |
| mordred | devananda: so - here's my use case | 17:42 |
| mordred | devananda: I have 48 machines in a rack in a data center | 17:43 |
| *** lucas-dinner has quit IRC | 17:43 | |
| mordred | devananda: they are all connected to a switch/router I do not control, and that has all of the machines connected to vlan25 | 17:43 |
| *** zhenguo has quit IRC | 17:43 | |
| mordred | devananda: I want to boot those machines using ironic/bifrost | 17:43 |
| mordred | devananda: so, as part of booting them, I need to communicate to them that they are/should be associated with vlan25 | 17:44 |
| mordred | devananda: I'm communicating the _rest_ of their network info via network_info.json that bifrost creates compatibly with the nova spec above | 17:44 |
| devananda | ahh | 17:44 |
| mordred | (assuming I might at some point want to use nova instead of bifrost) | 17:44 |
| devananda | (right - but that's irrelevant) | 17:45 |
| mordred | sure - just being complete as to why I care about teh nova spec if I'm using bifrost | 17:45 |
| devananda | so the difference between this patch and, say, https://review.openstack.org/#/c/187829/6 or https://review.openstack.org/#/c/188528/6 | 17:46 |
| *** achanda has joined #openstack-ironic | 17:46 | |
| devananda | is that you don't control the switch in any way, shape, or form -- and certainly not through neutron | 17:46 |
| *** trown|lunch is now known as trown | 17:46 | |
| *** krtaylor has joined #openstack-ironic | 17:47 | |
| devananda | rather than integrate ironic with a tool that configures the switch, you need to tell ironic-managed instances about the unchangeable switch they're attached to | 17:47 |
| *** puranamr has joined #openstack-ironic | 17:47 | |
| devananda | yes? | 17:47 |
| *** jcoufal has quit IRC | 17:48 | |
| *** igordcard has joined #openstack-ironic | 17:50 | |
| jroll | Pradip: yes, that's what configdrive and metadata service is for, + cloud-init or similar | 17:52 |
| *** puranamr has quit IRC | 17:53 | |
| jroll | mordred: devananda: what problems are you seeing here? | 17:53 |
| devananda | jroll: discussing https://review.openstack.org/#/c/152703/5 | 17:54 |
| jroll | right... | 17:54 |
| devananda | jroll: i initially misunderstood the intent, as I thought it was part o the ml2 work | 17:54 |
| jroll | devananda: well, it's kind of a dependency, otherwise the instance has no way to learn what vlan it is on | 17:54 |
| jroll | the info needs to be in the configdrive -- this defines a format for that, and also populates it if you're using nova | 17:55 |
| devananda | jroll: uhh... well, with ml2 integration, the instance shouldn't know | 17:55 |
| jroll | um, it needs to know? | 17:55 |
| jroll | if you're using vlan, the instance needs to be configured with that vlan no? | 17:55 |
| devananda | the switch should be tagging and/or filtering packets coming from the machine's phys interfaces | 17:55 |
| devananda | not simply trusting them | 17:56 |
| jroll | it has to tag outgoing packets | 17:56 |
| jroll | well, sure | 17:56 |
| jroll | the switch should also do security on that. | 17:56 |
| devananda | so | 17:56 |
| mordred | devananda: so - in my end, I don't care as much about how the vlans stuff gets configured - purely how I know about that state in the instance | 17:56 |
| jroll | but for the machine to tag the packet it needs to know what vlan to tag | 17:56 |
| mordred | I understand other people _do_ care about how they get configued | 17:56 |
| devananda | the instance does not need to tag packets IF the switch is configured properly | 17:56 |
| mordred | and that's important | 17:56 |
| *** lucas-dinner has joined #openstack-ironic | 17:56 | |
| jroll | devananda: what about with multiple vlans? | 17:57 |
| mordred | devananda: the instance needs to know what vlan tag is associated with the nic | 17:57 |
| devananda | mordred: totally. in your case (we can't change the switch config AND all traffic needs to be tagged) | 17:57 |
| devananda | mordred: then yea, this is needed | 17:57 |
| mordred | k | 17:57 |
| mordred | I mean - I imagine it's _use_ will be low | 17:57 |
| mordred | in broad sense | 17:57 |
| devananda | right | 17:57 |
| jroll | I'm confused if there's a problem here, and if so what the problem is | 17:57 |
| devananda | for network isolation of multiple tenants in the same bare metal pool (rack, what ever) -- the instances should not need to knkow or care about the cloud's vlans or the switch config | 17:58 |
| devananda | passing this data in is leaky and shouldn't be needed most of the time, assuming ironic<->neutron<->switch integration | 17:58 |
| jroll | I believe that's really up to the operator | 17:58 |
| jroll | our cloud does need to care | 17:59 |
| devananda | 17:55:58 < jroll> if you're using vlan, the instance needs to be configured with that vlan no? | 17:59 |
| devananda | jroll: ^ that's my concern | 17:59 |
| devananda | but it may be a miscommunication because it's not clear to me whether you're talking about the ML2 integration or not | 17:59 |
| jroll | that's my impression, I could be wrong. in some cases it does need to know, I'm not 100% sure if that's always a deployer decision or what | 17:59 |
| TheJulia | There is a difference between the cloud's infrastucture and what the tenant is being handed | 18:00 |
| jroll | in our deployment, wthe instance does need to be configured, so there is clearly a use case | 18:00 |
| devananda | jroll: yes. mordred is in the same boat right now | 18:00 |
| jroll | morgabra: you may or may not be interested in this discussion | 18:00 |
| *** amotoki has quit IRC | 18:00 | |
| devananda | jroll: but the ML2 work AIUI is designing something different | 18:00 |
| devananda | which we all would like to converge on | 18:01 |
| jroll | devananda: the ML2 work is defining a framework for how this works -- how the switch is configured is up to the ML2 mechanism for that switch | 18:01 |
| jroll | we're defining the interactions, not the exact switch configs, because that's impossible | 18:02 |
| devananda | "Neutron ML2 mechanism drivers should support this feature by using the data passed in binding profile to dynamically configure relevant ports and port-channels on the relevant switch(es)." | 18:02 |
| jroll | mhmm | 18:03 |
| jroll | and that configuration may require the server to tag its packets. | 18:03 |
| devananda | "VLAN provisioning on switch(es) is dependent on ML2 driver functionality being developed to support this feature." | 18:03 |
| devananda | gah | 18:03 |
| devananda | so we need to explicitly call that out | 18:03 |
| jroll | yeah, that one really makes me sad | 18:03 |
| devananda | as I think that's going to be, well, problematic | 18:03 |
| devananda | to put it politely | 18:04 |
| jroll | well | 18:04 |
| morgabra | jroll: devananda: ML2 functionality needing to be developed? | 18:04 |
| jroll | that's calling out that VLAN work is happening in a separate spec, AIUI | 18:04 |
| morgabra | not really | 18:04 |
| *** Sukhdev has joined #openstack-ironic | 18:04 | |
| morgabra | if you add a vlan to the provider network it's already visible to nova | 18:04 |
| devananda | jroll: link to that spec? | 18:04 |
| morgabra | 'segmentation_id' | 18:04 |
| jroll | devananda: the source is right where you got the quote from | 18:05 |
| * jroll double checks | 18:05 | |
| devananda | jroll: sorry, i meant, is there a spec for the ML2 work? I only see lnks to your network-provider spec | 18:05 |
| jroll | devananda: oh, I misunderstood | 18:05 |
| jroll | devananda: no, that's saying that the ML2 mechanisms for thse need to understand VLANs and how to configure the switchport, I believe | 18:06 |
| *** Marga_ has quit IRC | 18:06 | |
| *** davideagnello has joined #openstack-ironic | 18:06 | |
| devananda | right - so that, to me, seems to conflict with "it may require the server to tag its packets" | 18:06 |
| devananda | what am I missing? | 18:06 |
| *** Marga_ has joined #openstack-ironic | 18:07 | |
| jroll | devananda: but I believe it is related to https://review.openstack.org/#/c/94612/ as well | 18:07 |
| *** kozhukalov_ has quit IRC | 18:07 | |
| jroll | so | 18:07 |
| jroll | morgabra: our instances require the vlan to be configured, right? | 18:07 |
| jroll | deva contests that isn't necessary | 18:07 |
| *** degorenko has quit IRC | 18:07 | |
| morgabra | right, you have sort of 2 rough options: | 18:07 |
| jroll | I'm not sure if that's a technical decision that we explicitly made | 18:07 |
| morgabra | 1) if you bond your interfaces, you have to trunk vlans if you want to access multiple networks | 18:08 |
| jroll | or a switch feature thing we lack | 18:08 |
| morgabra | 2) if not, you can attach X networks to X interfaces | 18:08 |
| morgabra | and set the access vlan ^ | 18:08 |
| jroll | right, so with (2) you could refrain from the host knowing about the vlan and do it all on the switch | 18:08 |
| jroll | for (1) you need the host to know the vlans | 18:08 |
| jroll | right? | 18:08 |
| morgabra | correct, because it can assume all traffic is for the configured vlan | 18:09 |
| morgabra | because there's only 1 | 18:09 |
| jroll | ok, cool | 18:09 |
| jroll | so, we certainly need a way for the instance to know about the vlans | 18:09 |
| jroll | devananda: does that answer your question as to why we pass it to the host? | 18:09 |
| jroll | devananda: and what problems do you see with passing it to the host? | 18:10 |
| jroll | or mordred, even ^ | 18:10 |
| devananda | hrm | 18:11 |
| openstackgerrit | Merged openstack/ironic: Periodically checks the status of nodes in DEPLOYING state https://review.openstack.org/197141 | 18:11 |
| devananda | i do not understand (1)'s logic | 18:11 |
| jroll | sorry for misunderstanding on the 1:1 case | 18:11 |
| mordred | so - it's not just about bonding | 18:12 |
| mordred | the switch could still be configured in such a way that you need to know about the vlan for access purposes on the host | 18:12 |
| jroll | devananda: so let's say you have two nics, bonded. you want four vlans to ride on that bond. how does the instance/switch decide which vlan to send which packets on? | 18:12 |
| devananda | jroll: agreed on that. | 18:13 |
| devananda | however | 18:13 |
| jroll | mordred: sure, I can agree with that | 18:13 |
| devananda | why do we need 4 vlan's for a single neutron network? | 18:13 |
| jroll | you don't | 18:13 |
| jroll | you need 4 vlans for 4 neutron networks | 18:13 |
| devananda | aiui, we only support 1 net :: 1 port-group | 18:13 |
| jroll | for *now*, that's correct | 18:14 |
| devananda | ok | 18:14 |
| jroll | because software is hard and baby steps | 18:14 |
| devananda | so for now, we don't need the instance to know anyhting about VLANs, even for bonded NICs | 18:14 |
| jroll | that depends on the switch config. which depends on the ML2 implementation for that switch. | 18:14 |
| devananda | blah | 18:14 |
| jroll | but in general yes that should be able to be avoided | 18:15 |
| devananda | cool | 18:15 |
| jroll | assuming driver authors are capable of doing that :) | 18:15 |
| jroll | that said, I don't think it does any harm to pass the vlan to the instance | 18:15 |
| devananda | if the switch strips the vlan tag, then yes, it does | 18:16 |
| *** achanda has quit IRC | 18:16 | |
| devananda | so we need to be super clear to the ML2 drivers re: when they should trunk vs when they should tag/strip | 18:16 |
| jroll | because an instance could access other vlans that way? | 18:17 |
| devananda | yup | 18:17 |
| jroll | yeah | 18:17 |
| jroll | ml2 drivers should use port security features etc | 18:17 |
| devananda | which is why I'm really saying: we should always require the switch to strip vlan tags | 18:17 |
| jroll | "drop all packets for this port that are not on vlan X" | 18:18 |
| jroll | wait, always require it to strip tags? isn't that what causes things to be able to break out? | 18:19 |
| devananda | I mean "strip tags going to the host, replace tags coming from the host" | 18:19 |
| jroll | so that is to say the host shouldn't be configuring its own VLANs? | 18:20 |
| devananda | trunking + filtering out other VLANs works, but exposes the VLAN data to the tenant, meaning bits of your physical net infrastructure are now known by the tenant | 18:20 |
| devananda | jroll: correct | 18:20 |
| jroll | which means we can't do trunking at all. | 18:20 |
| devananda | right | 18:20 |
| jroll | which is a fail IMO | 18:21 |
| jroll | partially for the fact that we'll ALWAYS have to carry patches downstream. | 18:21 |
| jroll | and partially because nobody else will be able to do trunking without also carrying patches | 18:21 |
| TheJulia | for all the user knows, it is qinq or the switch is changing the tag number. Tenant machine knowing they need to use a tag number doesn't mean that tag is used anywhere else really. | 18:22 |
| *** mtanino has quit IRC | 18:22 | |
| morgabra | I'm confused why exposing vlan ids is so bad? | 18:22 |
| jroll | yep | 18:22 |
| jroll | and hiding the vlan is just security by obscurity | 18:22 |
| TheJulia | indeed | 18:22 |
| morgabra | it's true, information leak | 18:22 |
| TheJulia | indeed | 18:23 |
| jroll | there's only 4096, it is trivial to brute force them | 18:23 |
| morgabra | but if you configure your switchports, you can't just emit tagged traffic for another vlan and have it work | 18:23 |
| TheJulia | but you can't make assumptions that it means much :) | 18:23 |
| TheJulia | yup | 18:23 |
| jroll | morgabra: right, so I think the concern is that ML2 authors may write their plugin to be insecure in this case | 18:23 |
| jroll | which is a non-argument IMO | 18:24 |
| TheJulia | it requires clear, conconcise documentation. | 18:24 |
| morgabra | I mean, if you are allowing your hosts on vlans it's not supposed to be on | 18:25 |
| morgabra | you're gonna see traffic | 18:25 |
| devananda | i agree the security can be handled in other ways - and yah, driver authors need to be aware of and adhere to those, so we should make it simple to do good, and hard to do poorly | 18:25 |
| morgabra | broadcasts, etc | 18:25 |
| jroll | we can't just completely block useful features on the premise vendors might do it insecurely | 18:25 |
| devananda | but also, I dont think the instance should have to know about the infrastructure provider | 18:25 |
| devananda | that's really where i'm coming from here | 18:25 |
| jroll | even when the infrastructure provider causes it to automatically know? | 18:25 |
| jroll | take cinder integration for example, to do it in-band the instance needs to know about the infra provider | 18:26 |
| jroll | even just /b 60 | 18:26 |
| jroll | urgh | 18:26 |
| *** achanda has joined #openstack-ironic | 18:26 | |
| TheJulia | yup, at some point thats an operator choice to allow to occur | 18:26 |
| jroll | it's hardware, it needs to know about the infrastructure to do certain things. | 18:27 |
| devananda | is there an openstack API where I could query the VLAN configuration that my instance also got from cloudinit? | 18:27 |
| jroll | yes, all of this is also in the metadata service | 18:27 |
| jroll | and neutron | 18:27 |
| devananda | right - ok. then i'm cool with it | 18:27 |
| devananda | I knew it's in metadata | 18:27 |
| devananda | but if it's also exposed in neutron API, great | 18:27 |
| jroll | it comes directly forom the neutron network | 18:28 |
| jroll | from* | 18:28 |
| morgabra | might want to check with a neutron guy | 18:28 |
| jroll | and soon the port :P | 18:28 |
| morgabra | I don't know if I just YOLOd putting vlan id in the sementation_id | 18:28 |
| devananda | jroll: so that's the part that seems weird to me | 18:28 |
| jroll | morgabra: lol | 18:28 |
| devananda | morgabra: you did :) | 18:28 |
| devananda | that seems quite odd to me | 18:28 |
| jroll | morgabra: you're the only bare metal ml2 mech author right now, sooo | 18:28 |
| morgabra | god | 18:28 |
| morgabra | help us | 18:28 |
| jroll | :D | 18:28 |
| devananda | s/only/original/ | 18:28 |
| jroll | devananda: which part seems weird, the port? | 18:28 |
| jroll | devananda: s/only/only open source/ | 18:29 |
| jroll | or the fact that it's on the network object instead of the port? | 18:29 |
| devananda | jroll: no. exposing the vlan id in neutron's API | 18:29 |
| devananda | but then, many things about neutron seem wierd to me :) | 18:30 |
| jroll | devananda: I mean, you have to do it to enable this stuff | 18:30 |
| devananda | jroll: oh -- to be clear, can the non-admin user see it? | 18:30 |
| devananda | I'm sure the admin user can :) | 18:31 |
| jroll | devananda: I'm 99% sure | 18:31 |
| jroll | please hold :P | 18:31 |
| devananda | http://docs.openstack.org/admin-guide-cloud/content/provider_api_workflow.html | 18:31 |
| *** ukalifon has joined #openstack-ironic | 18:31 | |
| devananda | To use the provider extension with the default policy settings, you must have the administrative role. | 18:31 |
| morgabra | oh, it is supposed to be a vlan id | 18:31 |
| morgabra | yay | 18:31 |
| *** jcoufal has joined #openstack-ironic | 18:32 | |
| *** jcoufal has quit IRC | 18:32 | |
| devananda | mordred: probably ya'll already know how to do the neutron side of what you need to do, but i do not, and i found this, which looks very similar, so i'm sharing: http://www.s3it.uzh.ch/blog/openstack-neutron-vlan/ | 18:34 |
| devananda | jroll: also ^ says that all this is, by default, only available to admin | 18:34 |
| devananda | perhaps a question for Sukhdev ? | 18:34 |
| *** absubram has joined #openstack-ironic | 18:34 | |
| mordred | devananda: actually - I have not gotten to that part yet | 18:35 |
| jroll | devananda: hmm, this is just net-show | 18:35 |
| * Sukhdev lurking around | 18:35 | |
| mordred | devananda: I actually do not need any vlan info exposed to my vms | 18:35 |
| mordred | devananda: but right now, I'm just focusing on installing an OS with ssh access onto my ironic-controlled baremetal | 18:36 |
| devananda | mordred: oh, gotcha. | 18:36 |
| mordred | devananda: (thanks those - that's helpful) | 18:36 |
| morgabra | so the actual question is do we want to support the features that require leaking vlans | 18:36 |
| mordred | devananda: although we're planning on using linuxbridge and not ovs forneutron | 18:36 |
| morgabra | or at least not block it so hard its hard to hold a downstream patch for us crazy folk | 18:36 |
| devananda | morgabra: indeed | 18:37 |
| mordred | morgabra: well, sometimes it's not leaking - sometimes it's just a physical reality of the environment that your bare metal resides in | 18:37 |
| devananda | what bothers me is requiring the instance to know something (in order to get on the network) that we (afaik) do not share (via a REST API) with the user that created it | 18:37 |
| devananda | and by bothers, i mean, i will block it if I see it | 18:37 |
| mordred | devananda: yes. I agree that the information should be availabe via a REST API | 18:38 |
| jroll | I also agree | 18:38 |
| mordred | devananda: if it is also available to the instance | 18:38 |
| devananda | great :) | 18:38 |
| *** achanda has quit IRC | 18:38 | |
| mordred | as long as the result of that is that I can get the information somehow into my instance :) | 18:38 |
| devananda | if we have to expose the phys vlan id to the instance for $reasons -- I'm willing to accept that. there are clearly cases that do require that, though I also want us to handle the cases that do NOT require it | 18:38 |
| *** romcheg has joined #openstack-ironic | 18:39 | |
| jroll | so how does this sound: | 18:39 |
| jroll | provider:segmentation_id is an admin thing. that's the canonical place for the vlan | 18:39 |
| mordred | devananda: yes! zomg - I think that most things probably do not need to know anything about the phys vlan id | 18:40 |
| jroll | an ML2 driver that configures a port in such a fashion that it needs the instance to know the vlan, may place that data in a non-admin place on the port itself. | 18:40 |
| jroll | and then the user can get that data when needed, from the port object that relates to the interface that needs to be configured. | 18:40 |
| mordred | jroll: those words sound sane to me | 18:41 |
| devananda | jroll: and that data will also be passed in to cloud-init | 18:41 |
| devananda | s/may place/must place/ | 18:42 |
| mordred | s/cloud-init/config-drive or metadata service/ and yes | 18:42 |
| mordred | and yes to must place | 18:42 |
| devananda | mordred: right | 18:42 |
| jroll | devananda: well | 18:42 |
| jroll | at the time the metadata is built | 18:42 |
| jroll | the ml2 mechanism may not have configured things yet | 18:42 |
| mordred | if there is a vlan that must be known about, it must be in teh metadata place | 18:42 |
| jroll | or almost certainly will not have | 18:42 |
| mordred | jroll: oh - that's a whole other can of fish btw | 18:42 |
| jroll | because that is at the last moment | 18:42 |
| mordred | which one should not get me started upon | 18:43 |
| jroll | mordred: idk if I want to knwo | 18:44 |
| mordred | jroll: well, the nova thing I sent you yesterday where I was getting vms from the cloud that did not yet have network info even though they were ACTIVE | 18:44 |
| Sukhdev | This is very long discussion - hard for me to follow, if you have a specific question about provider nets or tenant nets, I can try to clarify | 18:45 |
| jroll | mordred: oh, that, that's a completely different thing (also did that get resolved?) | 18:45 |
| mordred | jroll: combined with an environment that does not believe in dhcp == HOW AM I SUPPOSED TO CONFIGURE THE NETWORK ON THAT VM???? *headdesk* | 18:45 |
| mordred | jroll: dunno - I wrote patch to not consider the machien active until it has addresses | 18:45 |
| jroll | mordred: yeah let's not get started on this other than a sincere apology from me | 18:45 |
| mordred | jroll: no worries - it's certainly not your fault! | 18:46 |
| jroll | it could be! | 18:46 |
| jroll | :P | 18:46 |
| mordred | jroll: although actually I think I abandoned the patch after realizing what you just said above | 18:46 |
| mordred | (Which si the reason I brought it up) | 18:46 |
| mordred | namely - if an instance can boot before the network config has been created and thus the metadata will have already been built | 18:46 |
| mordred | then there is a fundamental problem with the boot sequence | 18:47 |
| Sukhdev | mordred: when you launch a first instance on a given network (i.e. neutron port-create) the DHCP instance is attached to the network at that moment | 18:47 |
| mordred | Sukhdev: well, in the cloud in question, there is no dhcp | 18:47 |
| morgabra | jroll: so, fun story, binding:profile is admin_only by default too | 18:47 |
| mordred | Sukhdev: which means the instances can only get network config via static config in config-drive | 18:47 |
| morgabra | I don't see a place to stick the vlan on the port | 18:47 |
| *** mtanino has joined #openstack-ironic | 18:47 | |
| morgabra | like, assuming tenants can hit the neutron api anyway | 18:48 |
| * devananda takes a stab at rewriting this more formally | 18:48 | |
| devananda | http://paste.openstack.org/raw/381557/ | 18:48 |
| devananda | jroll: ^ | 18:48 |
| morgabra | obviously nova can figure it out | 18:48 |
| openstackgerrit | jxiaobin proposed openstack/ironic: Retry ipmi commands unless the failure is not retriable https://review.openstack.org/202742 | 18:48 |
| jroll | devananda: +1 | 18:49 |
| clif_h | devananda: is your shared lock patch something that's complete or is it a WIP? | 18:49 |
| devananda | Sukhdev: would appreciate your eyes on that pastelink too | 18:49 |
| devananda | clif_h: which one? | 18:49 |
| morgabra | jroll: devananda: I'm not sure it's correct to require it on the port object | 18:49 |
| Sukhdev | devananda: reading | 18:49 |
| morgabra | lots of stuff doesn't live on the port object, like the subnet definition | 18:49 |
| morgabra | etc | 18:49 |
| clif_h | https://review.openstack.org/#/c/202558/ | 18:49 |
| devananda | morgabra: hm, ok. | 18:49 |
| jroll | morgabra: I'm not sure either - that's just an initial proposal from me | 18:49 |
| devananda | clif_h: oh - not my patch. it's dtantsur|afk's -- i just fixed a typo | 18:50 |
| morgabra | being a part of the network is fine, already a mechanism there, it just so happens viewing it is admin_only by default :/ | 18:50 |
| clif_h | devananda: oh | 18:50 |
| jroll | CI passed on this if anyone else isn't a fan of ubuntu today :) https://review.openstack.org/#/c/202578/ /cc mordred | 18:50 |
| jroll | morgabra: so maybe it's up to the deployer to make that happen? | 18:51 |
| mordred | jroll: \o/ | 18:51 |
| * devananda stamps it | 18:51 | |
| openstackgerrit | jxiaobin proposed openstack/ironic: Retry ipmi commands unless the failure is not retriable https://review.openstack.org/202742 | 18:51 |
| morgabra | sort of a bummer, but yeah. There's some fine-grained controls there that would make it sane ('if network is shared or you own it or you are an admin) | 18:51 |
| NobodyCam | lol quick +a :-p | 18:52 |
| Sukhdev | devananda: when a nova boot request will be generated, the physical port where the server connects can be put in the access mode or trunk mode - this is a configurable option | 18:52 |
| Sukhdev | devananda: if the BM will be used to provide multi-tenancy, then it should be put into trunk port mode otherwise in the access mode | 18:53 |
| Sukhdev | devananda: one of the customers requested that they want the port in access mode during the boot operation, and in the trunk mode when the network flip takes place | 18:55 |
| openstackgerrit | jxiaobin proposed openstack/ironic: Retry ipmi commands unless the failure is not retriable https://review.openstack.org/202747 | 18:57 |
| Sukhdev | devananda: am I making sense or confusing the issue? Obviously I do not have full context of your discussion :-):-) | 18:57 |
| *** puranamr has joined #openstack-ironic | 18:58 | |
| *** rbrooker has joined #openstack-ironic | 19:01 | |
| *** kkoski has quit IRC | 19:02 | |
| *** kkoski has joined #openstack-ironic | 19:02 | |
| *** trown is now known as trown|outttypeww | 19:03 | |
| *** cdearborn has joined #openstack-ironic | 19:04 | |
| openstackgerrit | Thiago Paiva Brito proposed openstack/ironic-specs: OneView Driver for Ironic https://review.openstack.org/187762 | 19:06 |
| *** kkoski has quit IRC | 19:07 | |
| openstackgerrit | jxiaobin proposed openstack/ironic: Retry ipmi commands unless the failure is not retriable https://review.openstack.org/202751 | 19:08 |
| *** puranamr has quit IRC | 19:08 | |
| *** pal has joined #openstack-ironic | 19:14 | |
| *** [1]cdearborn has quit IRC | 19:18 | |
| openstackgerrit | jxiaobin proposed openstack/ironic: Retry ipmi commands unless the failure is not retriable https://review.openstack.org/202751 | 19:22 |
| devananda | back now - sorry, pulled away by things briefly | 19:23 |
| devananda | Sukhdev: so, the BM node needs access to some services (ironic, glance, swift) within the cloud control plane while it is being provisioned. | 19:24 |
| devananda | Sukhdev: after provisioning, it should be in the tenant network(s) | 19:24 |
| Sukhdev | devananda: The provisioning network can be provider network (which is admin network) and could be plumbed in a way to provide the services that are desired | 19:26 |
| devananda | exactly | 19:26 |
| *** dguerri` is now known as dguerri | 19:26 | |
| *** Nisha has quit IRC | 19:26 | |
| devananda | also not the question we were tackling :) | 19:26 |
| devananda | Sukhdev: we were discussing the mechanism for that BM node to attach to the tenant network, and whether the switchport should "trunk and filter" or whether it should strip tags as they leave the switchport | 19:27 |
| devananda | Sukhdev: *to attach once the user's instance is booting | 19:27 |
| Sukhdev | devananda: and, when nova boot request (i.e. port-create) comes to neutron, the ML2 plugin can plumb the switch port | 19:27 |
| devananda | if the switchport is in trunk mode, then the INSTANCE must know the correct segmentation IDs | 19:27 |
| devananda | but how do we get that data to the instance? -- JoshNang's patch to Nova addresses that | 19:28 |
| devananda | by passing the data via configdrive or cloud metadata service to the instance during boot | 19:28 |
| devananda | Sukhdev: that patch is here: https://review.openstack.org/#/c/152703/5 | 19:29 |
| devananda | Sukhdev: our discussion ended with this conclusion: http://paste.openstack.org/raw/381557/ | 19:29 |
| * Sukhdev looking | 19:30 | |
| devananda | note that none of this discussion hsa to do with the provisioning of the node, or how it gets access into the cloud control plane | 19:30 |
| devananda | it's _all_ about the tenant's instance | 19:30 |
| *** pal has quit IRC | 19:31 | |
| *** mitchjameson has joined #openstack-ironic | 19:36 | |
| Sukhdev | devananda: So, during the boot cycle, since server does not know about the segmentation-id, we use native VLAN (which is same as the provider network id) | 19:37 |
| Sukhdev | devananda: and when the server moves to the tenant network, ML2 driver will dynamically add tenant VLAN on the same port to support the tenant isolation - for, this server needs to be able to deal with the tagged packets | 19:38 |
| Sukhdev | devananda: and I think that is what this seems to be doing - however, I have not looked at it carefully | 19:39 |
| Sukhdev | s/this/this patch | 19:39 |
| devananda | Sukhdev: so we must remove provider network id from the switch at that point | 19:40 |
| devananda | Sukhdev: and pass the correct tenant VLAN id into the instance via the mechanism which that patch uses | 19:41 |
| devananda | Sukhdev: and we must expose that tenant VLAN id through neutron's REST API to the user who owns that instance | 19:41 |
| Sukhdev | devananda: correct - | 19:41 |
| devananda | I attempted to capture that conclusion in the pastebin I linked | 19:42 |
| Sukhdev | devananda: when the port-update or port-create is invoked it has a network ID in it, and network has the VLAN Id in it, this is how ML2 driver knows configure the port correctly | 19:42 |
| devananda | Sukhdev: that is invoked by nova or ironic, right? | 19:43 |
| devananda | Sukhdev: my concern is that it appears that the end-user-of-nova doesn't have access this information | 19:43 |
| Sukhdev | devananda: correct - presently it is done by nova - but, my understanding is jroll is going to look into doing it from ironic driver as a part of the network flip logic | 19:44 |
| jroll | no, nova will still do the port-create | 19:44 |
| jroll | ironic will do a port-update to pass the switchport info | 19:44 |
| devananda | ah, right -- nova creates it but doesn't bind it | 19:45 |
| jroll | Sukhdev: the concern is: when a user does "port-show" the response will not have the vlan id | 19:45 |
| Sukhdev | jroll: you mean neutron port-show ? | 19:46 |
| jroll | Sukhdev: yes | 19:46 |
| openstackgerrit | Merged openstack/ironic-python-agent: Change Dockerfile to use Debian as a base https://review.openstack.org/202578 | 19:46 |
| jroll | ^ rip ubuntu | 19:46 |
| Sukhdev | jroll: I believe the neutron net-show will show the VLAN ID | 19:46 |
| jroll | Sukhdev: no, it's provider:segmentation_id | 19:46 |
| jroll | provider: is admin only | 19:47 |
| jroll | as I understand it -- I haven't verified this myself | 19:47 |
| Sukhdev | jroll: oh I see what you are saying | 19:47 |
| Sukhdev | jroll: let make sure I understand the requirement correctly - then I can play around with it and provide answer/data | 19:48 |
| Sukhdev | jroll: so, you want to use the provider-net for provisioning network? why not tenant network? | 19:48 |
| jroll | Sukhdev: no | 19:49 |
| jroll | for the tenant network | 19:49 |
| jroll | I want the nova end user to be able to query neutron and get the vlan info. | 19:49 |
| *** harlowja has joined #openstack-ironic | 19:49 | |
| *** harlowja_ has quit IRC | 19:49 | |
| jroll | today I am under the impression that does not work because the fields under "provider:*" are admin-only. | 19:49 |
| jroll | but I may be wrong | 19:49 |
| Sukhdev | jroll: So, this is how it works (best way to understand this is to use horizon) - as an admin, you can create a provider network and specify a project (which is tenant) | 19:50 |
| Sukhdev | jroll: then it is visible to that tenant - even though it is created by admin. | 19:51 |
| jroll | Sukhdev: will the provider:segmentation_id fields be visible to the user? | 19:51 |
| Sukhdev | jrollL the tenant then can launch VMs on that network (i.e. port-creates) | 19:51 |
| Sukhdev | jroll: user being the tenant, right? | 19:51 |
| jroll | Sukhdev: yes, the tenant | 19:52 |
| Sukhdev | jroll: then the answer is yes | 19:52 |
| Sukhdev | jroll - we can confirm it - | 19:52 |
| *** harlowja has quit IRC | 19:52 | |
| *** harlowja has joined #openstack-ironic | 19:53 | |
| jroll | Sukhdev: ok, great. thanks! | 19:53 |
| Sukhdev | jroll: Here is the quick way to verify it - use horizon and go to admin mode, create a provider network and specify the project and see if it is visible to the tenant (under tenant tab) | 19:53 |
| Sukhdev | jroll: I do it for shared networks all the time - have not played around with provider networks - but, my understanding is that it works the same way - | 19:54 |
| Sukhdev | jroll: hence, I said I can verify it and provide the answer - or if you have a stack up, simply try it from horizon | 19:54 |
| devananda | BadCub: fyi, we're up to 20 signups for the midcycle now | 20:03 |
| *** achanda has joined #openstack-ironic | 20:03 | |
| NobodyCam | w00t :) | 20:03 |
| BadCub | devananda: coolness | 20:04 |
| *** Sukhdev has quit IRC | 20:04 | |
| * BadCub goes to check the food thing | 20:04 | |
| BadCub | only 16 responses for feeding | 20:05 |
| BadCub | devananda: when do we shut down registration? | 20:06 |
| *** achanda has quit IRC | 20:09 | |
| devananda | BadCub: a week before? I'll send another reminder next week, including "hey i'm going to close reg on the 3rd" or something | 20:10 |
| * TheJulia calls it and goes off to drink | 20:10 | |
| BadCub | devananda: should we maybe close two weeks before? I need time to coordinate with the office for lunch and [allthethings] to make sure everyone has something they can eat | 20:11 |
| BadCub | TheJulia: drink well! | 20:11 |
| devananda | BadCub: we can try :) | 20:12 |
| devananda | BadCub: though experience tells me we will have some folks who just show up unannounced, no matter what we do | 20:13 |
| devananda | *probably | 20:13 |
| BadCub | devananda: okay | 20:13 |
| BadCub | devananda: indeed. My main concern is lead time for food places. I am coordinating three different food types, so we have variety | 20:14 |
| *** Sukhdev has joined #openstack-ironic | 20:15 | |
| *** ukalifon has quit IRC | 20:17 | |
| BadCub | speaking of food... I really need to eat | 20:17 |
| devananda | so do I :) | 20:17 |
| BadCub | lol | 20:18 |
| *** achanda has joined #openstack-ironic | 20:19 | |
| *** igordcard has quit IRC | 20:25 | |
| *** e0ne has joined #openstack-ironic | 20:27 | |
| *** krtaylor has quit IRC | 20:36 | |
| devananda | mordred: quick question - does cloud-init (or simple-init) today have the capability to understand the segmentation id's being passed in by JoshNang's patch? | 20:42 |
| devananda | mordred: also, what about port bonding? | 20:43 |
| mordred | devananda: neither - but they're next on my list and should take about 20 minutes to add | 20:44 |
| devananda | cool | 20:44 |
| mordred | devananda: I added support for explicit dhcp interfaces yesterday in about 10 | 20:44 |
| devananda | but i would not be incorrect to say "these dont exist today" | 20:44 |
| mordred | devananda: I mean - I coudl have them exist today - what do you want to accomplish? | 20:45 |
| mordred | devananda: it's literally liek "if key in dict, write 2 lines to file" | 20:45 |
| mordred | I've just been waiting to make sure that we're all good with the format before I add it | 20:46 |
| devananda | mordred: sure - but that's distro specific, right? | 20:46 |
| mordred | the two lines? | 20:46 |
| mordred | kinda - so sorry - I have to do that bit twice - once for rh and once for deb :) | 20:46 |
| devananda | heh | 20:47 |
| lifeless | mordred: are you collaborating with dprince's os-net-config on the network wireup stuff? | 20:51 |
| lifeless | mordred: because, that might avoid glean bloat | 20:52 |
| lifeless | mordred: (just a thought in passing) | 20:52 |
| mordred | lifeless: no, I am not | 20:52 |
| mordred | lifeless: it has too many dependencies last I checked | 20:52 |
| *** max_lobur has joined #openstack-ironic | 20:52 | |
| mordred | yes. way too many depends - doesn't solve the problem glean is solving | 20:53 |
| lifeless | k | 20:53 |
| mordred | http://paste.openstack.org/show/381728 <-- the list of things onc installs | 20:54 |
| lifeless | wat | 20:54 |
| lifeless | thats surprising | 20:54 |
| jroll | devananda: we have patches for cloud-init to understand the vlan stuff, fwiw, idk what the status of upstreaming them is | 20:59 |
| jroll | they are in the open, somewhere | 20:59 |
| mgagne | jroll: somewhere here: https://github.com/jayofdoom/cloud-init-fedora-pkg/blob/master/cloud-init-0.7.5-onmetal-configdrive.patch | 21:01 |
| jroll | yes, ty mgagne | 21:01 |
| mgagne | jroll: patch works fine except that the link name has to be ethN instead of linkN to work on centos. unless it got fixed since. | 21:02 |
| zer0c00l | Are the disk images used for ironic (With IPA) is anyway different from the one that is used for VMs? | 21:02 |
| *** absubram has quit IRC | 21:02 | |
| zer0c00l | I mean can i re-use a qcow2 imaged used for VM with ironic-ipa? | 21:02 |
| jroll | mgagne: centos what? we're using it on 6/7 | 21:02 |
| mgagne | or I got something wrong on my side | 21:02 |
| mgagne | jroll: I didn't get a chance to look at your implementation of network_info.json so don't know if it's link0 or eth0 you are injecting. | 21:03 |
| mgagne | jroll: might just be me that fudged something in my baremetal config | 21:03 |
| zer0c00l | Also another question is the ironic-ipa installs grub2 by default, OS like RHEL6 uses grub1. Do you guys tested and found any implication for using grub2 on a OS that inherently uses grub1? | 21:03 |
| jroll | mgagne: mmm, maybe it's different with the upstream network_info.json, our patch is likely slightly different | 21:03 |
| mgagne | yours inject stuff in vendor_data.json =) | 21:04 |
| jroll | right | 21:04 |
| jroll | but surely that doesn't affect the link names :P | 21:04 |
| mgagne | yea I can't remember if it was me or not that fudge it | 21:04 |
| mgagne | I hate myself for not leaving meaningful comment in my commit :-/ | 21:06 |
| *** ijw has joined #openstack-ironic | 21:06 | |
| mgagne | "Refresh Add network_info to vendor_data.json" yea like it helps... | 21:06 |
| mgagne | what I'm wondering now is how to expose the provider network info the right way | 21:07 |
| jroll | yeah | 21:08 |
| jroll | are you at the nova midcycle next week? | 21:08 |
| jroll | mgagne: ^ | 21:08 |
| *** thiagop has joined #openstack-ironic | 21:08 | |
| mgagne | no, I'm only an operator hacking stuff. I don't know enough to feel like I *need* to get there | 21:09 |
| jroll | gotcha | 21:09 |
| jroll | I have some other networking things that are "if baremetal: else:" | 21:09 |
| mgagne | hehe | 21:09 |
| jroll | so I feel like this might come up next week | 21:09 |
| mgagne | yea, I'm trying to upstream stuff as much as I can. currently forward porting my patches and it's a pain as always | 21:10 |
| jroll | but if you have opinions I can put them in my luggage to bring them along | 21:10 |
| jroll | mhmmm | 21:10 |
| mgagne | about ~100 patches across all projects to triage ;) | 21:11 |
| mgagne | thanks god, most are ubuntu patches or backports | 21:11 |
| zer0c00l | jroll: Ping. | 21:15 |
| zer0c00l | Sorry to bother you, but do you know if i need to use any special disk images for IPA? | 21:15 |
| zer0c00l | or a qcow2 is enough ? | 21:15 |
| *** e0ne has quit IRC | 21:16 | |
| zer0c00l | from the looks of it, ipa supports a diskimage and a partition image? | 21:17 |
| zer0c00l | is there any docs on how to create them? | 21:17 |
| jroll | zer0c00l: the agent_* driver can use any whole disk image in qcow2 format | 21:18 |
| jroll | whether that image boots is an exercise to the user :) | 21:18 |
| jroll | and maybe check out disk-image-builder, or I think NobodyCam might have a command line for building images handy | 21:19 |
| zer0c00l | joll: i see | 21:19 |
| zer0c00l | Thanks | 21:19 |
| jroll | np | 21:19 |
| NobodyCam | huh | 21:20 |
| NobodyCam | me | 21:20 |
| jroll | heh | 21:20 |
| jroll | I can dig it up, don't worry | 21:20 |
| jroll | something like: | 21:21 |
| jroll | freenode_#openstack-ironic_20150714.log:[14:07:59] <NobodyCam> disk-image-create -a amd64 -o "{{http_boot_folder}}/{{deploy_image_filename}}" -t qcow2 "{{ | 21:21 |
| jroll | freenode_#openstack-ironic_20150714.log-[14:08:02] <NobodyCam> dib_os_element}}" vm serial-console simple-init "{{ extra_dib_elements}}" | 21:21 |
| jroll | zer0c00l: ^^ | 21:21 |
| NobodyCam | lol | 21:21 |
| NobodyCam | TY jroll :) | 21:21 |
| jroll | zer0c00l: that should at least give you the right direction | 21:22 |
| zer0c00l | i see | 21:22 |
| *** zz_natorious is now known as natorious | 21:22 | |
| jroll | zer0c00l: the 'vm' part makes it a whole disk image | 21:23 |
| jroll | dib_os_element is ubuntu or whatever | 21:23 |
| zer0c00l | jroll: i see. | 21:23 |
| jroll | you should be able to use the same images as your VM images, maybe with some driver tweaks or something | 21:23 |
| jroll | if it's a whole disk image IPA won't install grub | 21:23 |
| zer0c00l | aah, because grub is already in the full disk image | 21:24 |
| zer0c00l | and it is just a dd ? | 21:24 |
| TheJulia | zer0c00l: yeah, vm loads grub up in the image | 21:24 |
| TheJulia | zer0c00l: a little more than dd, a write out of a qcow2 to the device | 21:24 |
| TheJulia | basically going from qcow2 to raw format on the disk | 21:24 |
| zer0c00l | TheJulia: Cool! | 21:29 |
| zer0c00l | TheJulia: Sometimes the disk image size could be small right? IS it expanded to the size of the target disk as well? | 21:30 |
| jroll | not via ironic | 21:30 |
| jroll | that's generally handled via cloud-init or something else inside the process | 21:30 |
| TheJulia | zer0c00l: Ironic just writes out the image, and maybe writes a configdrive if defined. | 21:31 |
| * TheJulia has a 2/3rd written limited grow script open in another terminal she is working on | 21:31 | |
| gabriel-bezerra | jroll: thanks for the hand in the meeting. | 21:33 |
| jroll | gabriel-bezerra: totes | 21:34 |
| jroll | I don't think I actually helped, but happy to help! | 21:34 |
| *** krtaylor has joined #openstack-ironic | 21:37 | |
| *** boris-42 has joined #openstack-ironic | 21:42 | |
| *** bnemec has quit IRC | 21:46 | |
| *** max_lobur1 has joined #openstack-ironic | 21:57 | |
| *** max_lobur has quit IRC | 21:57 | |
| *** lucas-dinner has quit IRC | 21:59 | |
| *** davideagnello has quit IRC | 22:04 | |
| *** cing has joined #openstack-ironic | 22:05 | |
| *** ijw_ has joined #openstack-ironic | 22:06 | |
| Sukhdev | jroll: Hey - I just tried what I told you about the provider network - it works the way I explained it to you | 22:07 |
| jroll | Sukhdev: awesome, ty sir | 22:07 |
| jroll | mordred: devananda: ^^ user can read provider fields | 22:07 |
| jroll | er, morgabra ^ | 22:07 |
| *** puranamr has joined #openstack-ironic | 22:08 | |
| Sukhdev | jroll - I created a network for a tenant -then, as a tenant, I was able to create a subnet - i.e. able to assign the CIDR and see the vain ID, etc | 22:08 |
| morgabra | https://github.com/openstack/neutron/blob/master/etc/policy.json#L40 | 22:08 |
| morgabra | maybe this doesn't mean what I think it means | 22:08 |
| morgabra | lol | 22:08 |
| *** ijw has quit IRC | 22:09 | |
| jroll | morgabra: heh | 22:09 |
| jroll | Sukhdev: did you test on devstack or what? | 22:09 |
| mordred | jroll: aroo? | 22:09 |
| Sukhdev | morgabra: best way to do is to try yourself using horizon - I tend to use horizon (over CLI) | 22:10 |
| jroll | mordred: tab complete is hard | 22:10 |
| jroll | ignore me | 22:10 |
| *** achanda has quit IRC | 22:10 | |
| *** lucas-dinner has joined #openstack-ironic | 22:11 | |
| Sukhdev | morgabra jroll: See this paste - http://paste.openstack.org/show/381893/ | 22:13 |
| Sukhdev | morgabra jroll : See the very last command - I changed the role to demo tenant and was able to see the VLAN ID | 22:13 |
| jroll | Sukhdev: I'm skeptical of source ~/devstack/openrc admin demo | 22:13 |
| jroll | Sukhdev: can you try with "demo demo" | 22:13 |
| jroll | just to be sure | 22:13 |
| Sukhdev | jroll: hang on - let me give that a shot as well | 22:14 |
| *** Sukhdev has quit IRC | 22:14 | |
| *** kkoski has joined #openstack-ironic | 22:16 | |
| *** Sukhdev has joined #openstack-ironic | 22:16 | |
| *** natorious is now known as zz_natorious | 22:16 | |
| Sukhdev | jroll: interesting - http://paste.openstack.org/show/381894/ | 22:16 |
| jroll | Sukhdev: yep, that's about what I expected | 22:17 |
| Sukhdev | jroll: I do not see see VLAN ID - however, when I am on horizon, from demo tenan's view I can see it | 22:17 |
| Sukhdev | how interesting - and, I always use horizon :-) | 22:17 |
| jroll | well, I think you're using an admin user on the demo tenant | 22:17 |
| jroll | or something | 22:17 |
| *** kkoski has quit IRC | 22:18 | |
| Sukhdev | jroll: perhaps - | 22:18 |
| Sukhdev | jroll: so, what does it mean then - is it a blocker? | 22:18 |
| jroll | Sukhdev: not sure, let me ask a neutron person | 22:19 |
| jroll | owait | 22:19 |
| jroll | :P | 22:19 |
| jroll | it isn't a total blocker, but | 22:19 |
| jroll | it is something that needs to happen. | 22:19 |
| jroll | Sukhdev: hah http://lists.openstack.org/pipermail/openstack-dev/2015-July/069791.html | 22:20 |
| Sukhdev | jroll: seems to make sense | 22:22 |
| mgagne | yha, you have to update policy.json | 22:23 |
| Sukhdev | jroll: what is the reason that for BM provisioning, we need to use provider network? | 22:23 |
| jroll | Sukhdev: well, there might be a reason to put an instance on the provider network | 22:24 |
| jroll | for example to connect to cinder | 22:24 |
| *** puranamr has quit IRC | 22:24 | |
| *** achanda has joined #openstack-ironic | 22:27 | |
| Sukhdev | jroll: so, you are saying cinder (or similar services) will not be under the same OpenStack controller - and hence, they will be put on the provider network ? | 22:28 |
| jroll | Sukhdev: how else would the BM node reach cinder? | 22:29 |
| Sukhdev | jroll: generally, provider networks are used to patch-in the existing legacy networks | 22:29 |
| *** sinval has joined #openstack-ironic | 22:29 | |
| jroll | ok, sure | 22:29 |
| *** puranamr has joined #openstack-ironic | 22:30 | |
| jroll | there might be a reason to attach a provider network | 22:30 |
| jroll | say, the server needs to reach that legacy network | 22:30 |
| Sukhdev | jroll: so, you are saying cinder falls into legacy network category? | 22:30 |
| Sukhdev | jroll- if server needs to reach to the legacy network - that can be accomplished by adding the server to provider network after it has booted up - | 22:31 |
| Sukhdev | jroll: server can connect to multiple network, if it needs to | 22:32 |
| *** puranamr has quit IRC | 22:33 | |
| Sukhdev | jroll: typical provider network use-case is that there is a BM server or storage cluster that is sitting on a legacy network in the DC and the VMs (from the virtual world) need to access those servers | 22:34 |
| jroll | sure | 22:35 |
| Sukhdev | jroll: and those servers are not visible to the openstack controller - hence the segmentation-id is specified (as those services are already sitting on that VLAN) | 22:35 |
| jroll | Sukhdev: so if you add the server to the provider network (for any reason, not just cinder), and need a vlan configured, you need to be able to get the vlan from the provider network object in neutron | 22:36 |
| jroll | Sukhdev: another use case: public cloud providers that use the provider network to specify the shared 10.x space that is internal to a datacenter | 22:37 |
| *** cdearborn has quit IRC | 22:37 | |
| Sukhdev | jroll: understood | 22:38 |
| jroll | Sukhdev: in any case, even for non-provider networks, we need a place to put segmentation id that the user can access | 22:38 |
| Sukhdev | jroll: I think for non-provider networks, that is visible - | 22:39 |
| jroll | what is the field, just segmentation_id? | 22:39 |
| morgabra | well, it's complicated a little bit | 22:39 |
| morgabra | tenants would have to create the network with the relevant provider:* keys | 22:40 |
| morgabra | (type vlan, etc) | 22:40 |
| Sukhdev | jroll: let me look it up on my setup - I have it running - give me a sec | 22:40 |
| morgabra | there's a mechanism in the extension to allocate vlans for you | 22:40 |
| jroll | morgabra: for the purposes of nova or? | 22:40 |
| jroll | rephrasing, why provider:* | 22:40 |
| morgabra | how else are you going to allocate vlans? | 22:40 |
| jroll | idk | 22:41 |
| morgabra | if tenants are creating their own networks | 22:41 |
| jroll | just asking why because idk | 22:41 |
| morgabra | yeah, afaik it's the only mechanism that exists already | 22:41 |
| jroll | ok | 22:41 |
| *** achanda has quit IRC | 22:41 | |
| morgabra | I was thinking about this completely from the perspective of granting access to large shared networks | 22:41 |
| morgabra | whoops | 22:41 |
| jroll | when? like in the initial plugin? | 22:42 |
| morgabra | suggesting using the provider extension | 22:42 |
| morgabra | it actually maps pretty well, it's just admin only by default. | 22:43 |
| jroll | ah, yeah | 22:44 |
| *** ijw_ has quit IRC | 22:44 | |
| Sukhdev | jroll: yeah - I just checked - VLAN ID is not visible to tenant - unless tenant admin context | 22:45 |
| jroll | right | 22:45 |
| *** foexle has quit IRC | 22:46 | |
| *** lucas-dinner has quit IRC | 22:46 | |
| Sukhdev | jroll: Hey Jim. - to change the subject on you - take a look at this paste - when I restart (or start from devstack) ir-api, I get this traceback - I see the same for ir-cond as well - http://paste.openstack.org/show/382015/]\ | 22:56 |
| Sukhdev | Any idea what could be missing/wrong? | 22:56 |
| jroll | Sukhdev: I believe that's a known issue and not an actual problem | 22:57 |
| jroll | I forget the fix but there's patches around for it | 22:57 |
| Sukhdev | jroll: thanks - let me do some googling to see if I can find something | 22:58 |
| jroll | Sukhdev: I'd just ignore it | 22:59 |
| jroll | it doesn't actually break things | 22:59 |
| Sukhdev | jroll: I do not see any ironic ports or nodes - I thought this was causing it | 23:00 |
| jroll | hmm | 23:01 |
| jroll | shouldn't be | 23:01 |
| Sukhdev | jroll: when I run devstack, doesn't it create some default nodes? | 23:01 |
| jroll | Sukhdev: it should, probably depends on your localrc | 23:01 |
| jroll | Sukhdev: also may depend on which user, make sure to admin all the things | 23:02 |
| Sukhdev | jroll - see this - http://paste.openstack.org/show/382017/ | 23:03 |
| Sukhdev | jroll: this should be sufficient - isn't it? | 23:04 |
| jroll | looks correct | 23:04 |
| Sukhdev | jroll: yeah - when devstack completes, it gives the error which I posted - and when I list nodes, ports - it is empty list - hence, I thought that this error was causing some issue :-) | 23:08 |
| *** achanda has joined #openstack-ironic | 23:09 | |
| jroll | Sukhdev: right, that isn't the issue | 23:11 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic: Updated from global requirements https://review.openstack.org/202699 | 23:16 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic-lib: Updated from global requirements https://review.openstack.org/201819 | 23:16 |
| openstackgerrit | Michael Krotscheck proposed openstack/ironic: Added CORS support middleware to Ironic https://review.openstack.org/199769 | 23:17 |
| *** athomas has quit IRC | 23:21 | |
| *** zigo has quit IRC | 23:21 | |
| *** zigo has joined #openstack-ironic | 23:22 | |
| *** cing has quit IRC | 23:23 | |
| *** zz_natorious is now known as natorious | 23:23 | |
| *** openstackstatus has joined #openstack-ironic | 23:34 | |
| *** ChanServ sets mode: +v openstackstatus | 23:34 | |
| *** bizarrochristy has quit IRC | 23:34 | |
| *** natorious is now known as zz_natorious | 23:34 | |
| *** bizarrochristy has joined #openstack-ironic | 23:34 | |
| *** bizarrochristy has quit IRC | 23:38 | |
| *** ijw has joined #openstack-ironic | 23:38 | |
| *** jaypipes has quit IRC | 23:50 | |
| *** max_lobur1 has quit IRC | 23:53 | |
| *** max_lobur has joined #openstack-ironic | 23:53 | |
| *** dguerri is now known as dguerri` | 23:54 | |
| *** barra204 has joined #openstack-ironic | 23:54 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!