| NobodyCam | wanders afk for a while | 00:05 |
|---|---|---|
| *** yuanying has quit IRC | 00:06 | |
| *** yuanying has joined #openstack-ironic | 00:15 | |
| mrda | jroll, comstud: There's been some comments on your caching stuff over here: https://review.openstack.org/#/c/115540/4/nova/virt/ironic/driver.py | 00:17 |
| mrda | Any comments welcomed from you both | 00:17 |
| *** r-daneel has quit IRC | 00:17 | |
| mrda | Just figured you might like to opportunity to comment before I do | 00:18 |
| jroll | ugh | 00:18 |
| mrda | Thanks jroll and comstud | 00:18 |
| jroll | UGH | 00:18 |
| jroll | mrda: I'm trying to run out the door, do you know the answers to those? | 00:21 |
| mrda | jroll: I can answer them if you like (ie. read the code and work it out) | 00:22 |
| jroll | mrda: tldr the cache is refreshed every loop of the resource tracker, it won't get hit twice for the same node, etc | 00:22 |
| mrda | ok | 00:22 |
| mrda | just figured you might like to comment. | 00:22 |
| jroll | or you can bug comstud, he's supposedly working :) | 00:22 |
| jroll | I would | 00:22 |
| mrda | If you think everything is ok, I can comment to that effect | 00:22 |
| jroll | but late for a thing already | 00:23 |
| jroll | yes, ironic peolpe asked the same question | 00:23 |
| jroll | (so maybe we want to add a comment, idk) | 00:23 |
| mrda | thanks | 00:23 |
| mrda | lifeless: poke | 00:38 |
| lifeless | poke | 00:38 |
| mrda | So, back on cpu_arch | 00:38 |
| mrda | The Tl;DR is that we'll leave it as "NotFound" | 00:39 |
| mrda | But we will canonicalize() it if it is set | 00:39 |
| mrda | Agreed? | 00:39 |
| lifeless | no | 00:39 |
| mrda | ...which means TripleO may need to canonicalize as well | 00:39 |
| lifeless | we will leave it as is. | 00:39 |
| mrda | hmmm | 00:39 |
| lifeless | If we want to add canonicalize, we have to do that to the scheduler as well. | 00:39 |
| lifeless | since this data is late bound | 00:40 |
| lifeless | (it lives in Ironic, and Ironic hasn't been canonicalising - if there is bad data out there, its in users systems already) | 00:40 |
| mrda | You realise that when this blows up I'm going to point to this eavesdrop :) | 00:40 |
| mrda | But I understand | 00:40 |
| lifeless | we seemed to have consensus on that here yesterday with jroll and JayF | 00:41 |
| mrda | yeah, we were certain on the NotFound bit | 00:41 |
| lifeless | I don't remember whether it was 'and do the canonicalise' or 'and do nothing', but it was clear that | 00:42 |
| lifeless | if we canonicalize we do that in the scheduler too | 00:42 |
| mrda | I wasn't sure we were on the canonicalize() bit | 00:42 |
| JayF | I'd be willing to bet that what we have in our database isn't canonialize()'d | 00:42 |
| JayF | and I'd really rather not have the merged nova driver break me without database work | 00:42 |
| lifeless | and frankly, I think having canonicalize called shouldn't be a driver responsibility | 00:42 |
| lifeless | if nova wants things canonicalized | 00:42 |
| mrda | ok, just wanted to be sure | 00:42 |
| lifeless | thats a separate patch - change the scheduler and the compute manager at the same time, fixing anything the driver might have wrong, *for all* drivers. | 00:43 |
| mrda | Thanks lifeless and JayF! | 00:44 |
| *** chuckC has quit IRC | 00:45 | |
| *** eghobo has quit IRC | 01:10 | |
| *** eghobo has joined #openstack-ironic | 01:11 | |
| *** eguz has joined #openstack-ironic | 01:21 | |
| *** eguz has quit IRC | 01:21 | |
| *** eghobo has quit IRC | 01:25 | |
| *** nosnos has joined #openstack-ironic | 01:49 | |
| *** openstackgerrit has quit IRC | 02:02 | |
| *** openstackgerrit has joined #openstack-ironic | 02:03 | |
| *** yuanying has quit IRC | 02:17 | |
| *** eghobo has joined #openstack-ironic | 02:32 | |
| *** eghobo has quit IRC | 02:37 | |
| *** eghobo has joined #openstack-ironic | 02:38 | |
| *** chuckC has joined #openstack-ironic | 02:40 | |
| *** dlaube has quit IRC | 02:43 | |
| *** killer_prince is now known as lazy_prince | 02:46 | |
| *** eguz has joined #openstack-ironic | 02:55 | |
| *** eghobo has quit IRC | 02:58 | |
| *** eguz has quit IRC | 03:01 | |
| openstackgerrit | Haomeng,Wang proposed a change to openstack/ironic: Add send-data-to-ceilometer support for pxe_ipminative driver https://review.openstack.org/112486 | 03:02 |
| lazy_prince | Hi All.. Has anyone looked at https://github.com/rackerlabs/ironic-neutron-plugin.. i have few doubts regarding it.. can someone please help..? | 03:11 |
| lazy_prince | does it need a corresponding agent running on the compute node..? | 03:12 |
| JoshNang | lazy_prince: hey i wrote the ironic side code to control that plugin. it doesn't require an agent on the compute node. it's a plugin for neutron and then talks directly to switches during the deploy/tear_down process to configure vlans | 03:14 |
| JoshNang | morgabra wrote the actual code there (but doesn't look like he's in this channel) | 03:15 |
| JoshNang | nvm, tab complete failed me | 03:15 |
| lazy_prince | okay.. but as i understand, its an extension of ml2 plugin from Neutron, which as per my understanding, needs an agent on the compute side.. correct me if i am wrong.. | 03:17 |
| lazy_prince | i posted it on neutron channel but no one responded.. | 03:18 |
| JoshNang | lazy_prince: it definitely doesn't need an agent on the deployed nodes. i'm not sure the details | 03:18 |
| JoshNang | i think we're just not sending any of the calls that would normally go to the agent | 03:19 |
| lazy_prince | so basically, when using this plugin, all physical switch management will be handled by this plugin in neutron itself.. | 03:19 |
| JoshNang | lazy_prince: correct, at least as far as flipping vlans on ports and ip assignments | 03:20 |
| lazy_prince | but IP assignment is still done by DHCP agent of neutron.. right.. | 03:25 |
| lazy_prince | sorry, i am new to neutron and hence so many questions.. | 03:26 |
| *** rloo has quit IRC | 03:26 | |
| *** nosnos has quit IRC | 03:26 | |
| mrda | lazy_prince: don't be embarrassed by asking questions. We all learn this way. | 03:27 |
| *** nosnos has joined #openstack-ironic | 03:27 | |
| lazy_prince | I also has a look at https://review.openstack.org/#/c/84795 | 03:30 |
| lazy_prince | JoshNang: I see that there is a agent_utils.py introduced to talk to Neutron.. Can you please tell me why did we write this and not used the current neutron client (bundled with ironic). | 03:31 |
| *** nosnos has quit IRC | 03:32 | |
| JoshNang | lazy_prince: (sorry, hosting an event at the office) agent_utils is a wrapper for neutron client. it adds some of the extra params a neutron plugin that deals with hardware needs | 03:55 |
| JoshNang | all 4 of the main calls (configure_instance_networks, add_provisioning_network, etc) are just ways to grab the physical port descriptions and add them into the calls, with some error checking | 03:56 |
| JoshNang | so the way we do ip assignment is a static ip sent to the node via configdrive. neutron allocates the ip to nova, nova makes a configdrive, ironic writes the configdrive to the node, the node uses cloud-init to read what it's ip is supposed to be. semi-dynamic, but static on the hosts | 03:57 |
| *** rameshg87 has joined #openstack-ironic | 03:59 | |
| *** mitz_ has quit IRC | 03:59 | |
| *** mitz has joined #openstack-ironic | 03:59 | |
| lazy_prince | static IP via config drive..? why not use DHCP (dnsmasq)for that matter.. I just want to understand.. | 04:02 |
| *** eghobo has joined #openstack-ironic | 04:07 | |
| JoshNang | we don't do dhcp on our public cloud. we didn't make that decision | 04:12 |
| *** eghobo has quit IRC | 04:12 | |
| lazy_prince | aha... so if one chooses to do dhcp, is config drive still needed..? | 04:13 |
| lifeless | nope | 04:14 |
| lifeless | (but you do need to also deploy the metadata server at that point) | 04:14 |
| lazy_prince | yup, i understand that metadata server will be needed if config drive is not used.. | 04:15 |
| JoshNang | correct | 04:16 |
| *** pcrews has quit IRC | 04:19 | |
| rameshg87 | JoshNang, lazy_prince, but we don't use config driver right now in ironic, right ? are there plans to use them ? | 04:26 |
| rameshg87 | JoshNang, lazy_prince, s/config driver/config drives/ | 04:26 |
| *** nosnos has joined #openstack-ironic | 04:27 | |
| lazy_prince | rameshg87: i guess, its a user choice to use configdrives or metadata server | 04:28 |
| JoshNang | rameshg87: right. i think we have a patch proposed to add it after the nova driver lands | 04:28 |
| rameshg87 | lazy_prince, right, but there is no "choice* of config drives available in ironic right now :) | 04:29 |
| rameshg87 | JoshNang, okay | 04:29 |
| lazy_prince | JoshNang: can you post the patch link..? | 04:29 |
| rameshg87 | lazy_prince, +1, would like to look at patch link :) | 04:30 |
| lazy_prince | if you have it handy.. | 04:30 |
| *** ramineni has joined #openstack-ironic | 04:30 | |
| JoshNang | err maybe not | 04:31 |
| JoshNang | lazy_prince: rameshg87 here's the spec https://review.openstack.org/#/c/99235/ | 04:31 |
| rameshg87 | JoshNang, thanks | 04:32 |
| rameshg87 | JoshNang, thanks for the mention of the ilo :) | 04:32 |
| JoshNang | rameshg87: no prob! :) | 04:32 |
| JoshNang | lazy_prince: rameshg87 here's the code that we're running for configdrive | 04:33 |
| JoshNang | https://review.openstack.org/#/c/86192/ | 04:33 |
| JoshNang | (heh copy paste fail) | 04:33 |
| lazy_prince | thanks.. | 04:33 |
| *** eghobo has joined #openstack-ironic | 04:34 | |
| JoshNang | no problem! i'm sorry it's not all integrated and easy to set up yet. but soon, hopefully! i'm gonna head home...have a good day/night! | 04:35 |
| lazy_prince | we hope so too.. | 04:36 |
| *** lazy_prince is now known as killer_prince | 04:49 | |
| *** rakesh_hs has joined #openstack-ironic | 05:01 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/117423 | 05:35 |
| *** k4n0 has joined #openstack-ironic | 06:03 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/115878 | 06:10 |
| *** foexle has joined #openstack-ironic | 06:17 | |
| *** dtantsur|afk is now known as dtantsur | 06:30 | |
| dtantsur | Morning Ironic | 06:30 |
| *** Haomeng has quit IRC | 06:32 | |
| *** killer_prince is now known as lazy_prince | 06:33 | |
| mrda | Morning dtantsur | 06:34 |
| dtantsur | mrda, g'evening | 06:35 |
| mrda | :) | 06:35 |
| *** foexle_ has joined #openstack-ironic | 06:40 | |
| *** foexle_ has quit IRC | 06:40 | |
| openstackgerrit | Syed Ismail Faizan Barmawer proposed a change to openstack/ironic: Add UEFI based deployment support in Ironic https://review.openstack.org/114357 | 06:54 |
| *** dtantsur is now known as dtantsur|bbl | 07:03 | |
| *** Haomeng has joined #openstack-ironic | 07:08 | |
| *** eghobo has quit IRC | 07:08 | |
| *** harlowja_ is now known as harlowja_away | 07:14 | |
| *** todd_dsm has quit IRC | 07:24 | |
| *** jcoufal has joined #openstack-ironic | 07:49 | |
| openstackgerrit | Mike Durnosvistov proposed a change to openstack/ironic: Check whether specified FS is supported https://review.openstack.org/98102 | 07:54 |
| *** jistr has joined #openstack-ironic | 07:56 | |
| *** jistr has quit IRC | 07:56 | |
| *** jistr has joined #openstack-ironic | 08:08 | |
| *** rushiagr_away is now known as rushiagr | 08:08 | |
| *** rushiagr is now known as rushiagr_away | 08:12 | |
| yuriyz | Morning Ironic | 08:14 |
| mrda | Hi yuriyz | 08:17 |
| openstackgerrit | A change was merged to openstack/ironic: properly format user guide in RST https://review.openstack.org/116168 | 08:18 |
| *** lucasagomes has joined #openstack-ironic | 08:18 | |
| *** derekh has joined #openstack-ironic | 08:20 | |
| *** boris-42 has quit IRC | 08:21 | |
| Haomeng | yuriyz: morning:) | 08:27 |
| yuriyz | evening Haomeng :) | 08:28 |
| Haomeng | yuriyz: :) | 08:28 |
| *** boris-42 has joined #openstack-ironic | 08:30 | |
| *** ramineni has quit IRC | 08:37 | |
| lucasagomes | morning yuriyz Halacs | 08:37 |
| lucasagomes | ouch | 08:37 |
| lucasagomes | Haomeng, morning* | 08:37 |
| Halacs | morning lucasagomes | 08:38 |
| Haomeng | morning lucasagomes:) | 08:38 |
| lucasagomes | :) | 08:38 |
| *** ramineni has joined #openstack-ironic | 08:46 | |
| *** nosnos has quit IRC | 08:47 | |
| *** nosnos has joined #openstack-ironic | 08:47 | |
| *** nosnos has quit IRC | 08:52 | |
| *** MattMan has quit IRC | 08:57 | |
| *** MattMan has joined #openstack-ironic | 08:58 | |
| *** yongli has quit IRC | 09:06 | |
| *** athomas has joined #openstack-ironic | 09:11 | |
| *** nosnos has joined #openstack-ironic | 09:24 | |
| *** romcheg1 has joined #openstack-ironic | 09:32 | |
| romcheg1 | Good morning Ironic! | 09:32 |
| mrda | Hi romcheg1 | 09:37 |
| lucasagomes | morning mrda romcheg1 | 09:52 |
| romcheg1 | Now I'm with you again :) | 09:52 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: DRAC wsman_{enumerate, invoke}() to return an ElementTree object https://review.openstack.org/116871 | 09:53 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Implements the DRAC ManagementInterface for get/set boot device https://review.openstack.org/116913 | 09:53 |
| *** jcoufal has quit IRC | 10:09 | |
| *** jcoufal has joined #openstack-ironic | 10:12 | |
| *** igordcard has joined #openstack-ironic | 10:14 | |
| *** nosnos has quit IRC | 10:20 | |
| *** nosnos has joined #openstack-ironic | 10:21 | |
| *** romcheg1 has quit IRC | 10:23 | |
| *** romcheg1 has joined #openstack-ironic | 10:24 | |
| *** nosnos has quit IRC | 10:25 | |
| *** pelix1 has joined #openstack-ironic | 10:32 | |
| *** nosnos has joined #openstack-ironic | 10:34 | |
| *** HenryG has quit IRC | 10:43 | |
| *** rameshg87 has quit IRC | 10:59 | |
| *** rakesh_hs has quit IRC | 11:09 | |
| *** nosnos has quit IRC | 11:18 | |
| *** nosnos has joined #openstack-ironic | 11:20 | |
| *** igordcard has quit IRC | 11:36 | |
| *** dtantsur|bbl is now known as dtantsur | 11:44 | |
| *** Halacs has left #openstack-ironic | 11:46 | |
| *** HenryG has joined #openstack-ironic | 11:52 | |
| lucasagomes | jroll, ping | 12:12 |
| *** lazy_prince is now known as killer_prince | 12:20 | |
| *** nosnos has quit IRC | 12:22 | |
| *** nosnos has joined #openstack-ironic | 12:23 | |
| *** nosnos has quit IRC | 12:27 | |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Log ERROR power state in node_power_action() https://review.openstack.org/117499 | 12:55 |
| *** k4n0 has quit IRC | 13:10 | |
| lucasagomes | yuriyz, ping re bug 1358816. What it means by checked? | 13:11 |
| lucasagomes | are we taking some action in case it returns ERROR like putting the node in maintenance or something? | 13:11 |
| Shrews | Is it safe to do rechecks yet? | 13:20 |
| Shrews | infra says "yes" | 13:25 |
| yuriyz | lucasagomes, before we put node in maintenance after CONF.conductor.power_state_sync_max_retries only if driver raises exception, now we do it also if ERROR state returned | 13:30 |
| lucasagomes | Shrews, seems so | 13:30 |
| lucasagomes | yuriyz, right, so that's what the bug is proposing? to check for those states and mark the node as in maintenance in case it occurs | 13:32 |
| lucasagomes | cause when I read it says like ERROR state should be checked, but it doesn't say which action we would take in case the ERROR state appears | 13:33 |
| yuriyz | lucasagomes, we should do something with ERROR power state from drivers, agree with https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ipminative.py#L214 | 13:37 |
| lucasagomes | yuriyz, I c | 13:38 |
| lucasagomes | yeah | 13:38 |
| yuriyz | lucasagomes https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ipmitool.py#L476 also | 13:40 |
| lucasagomes | +1 to return ERROR from the driver and check that one layer above | 13:43 |
| lucasagomes | I was just wondering what would be the action on the layer above (which is not explicity in that bug) | 13:43 |
| lucasagomes | bug description* | 13:43 |
| yuriyz | lucasagomes, IMO exception from get_power_state() - fatal error (network, etc), states.ERROR - we got reply from node but with strange date | 13:47 |
| yuriyz | data | 13:47 |
| jroll | morning lucasagomes Shrews yuriyz :) | 13:52 |
| lucasagomes | jroll, hey... was waiting for ya | 13:52 |
| lucasagomes | jroll, there's a race condition on the DHCP patch | 13:52 |
| Shrews | morning jroll | 13:52 |
| lucasagomes | jroll, mostly because we are using a global var there | 13:52 |
| lucasagomes | jroll, I have a fix locally for that, I basically converted that factory file in a class and so on | 13:53 |
| lucasagomes | jroll, it's ok if I submit it on top of your patch? | 13:53 |
| lucasagomes | I mean ellen's/your patch | 13:53 |
| NobodyCam | good morning ironic | 13:53 |
| lucasagomes | factory file == factory functions | 13:53 |
| NobodyCam | morning Shrews jroll lucasagomes and yuriyz | 13:54 |
| lucasagomes | morning NobodyCam | 13:54 |
| NobodyCam | :) | 13:54 |
| jroll | lucasagomes: sure, always ok :) | 13:55 |
| lucasagomes | jroll, ack | 13:55 |
| jroll | lucasagomes: curious what the race is but I guess I'll watch for your patch | 13:55 |
| jroll | morning NobodyCam :) | 13:55 |
| lucasagomes | jroll, yeah... I put a paste in the patch | 13:55 |
| lucasagomes | I saw it because py26 was failing with that error | 13:55 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Make DHCP provider pluggable https://review.openstack.org/112351 | 13:56 |
| lucasagomes | ^ tested locally it works | 13:56 |
| lucasagomes | ew why git is not seems ironic/tests/dhcp/test_neutron.py as renamed from ironic/tests/test_neutron.py >.< | 13:58 |
| *** pcrews has joined #openstack-ironic | 13:59 | |
| jroll | lucasagomes: I see, I was also thinking about just loading the driver at conductor startup at one point | 13:59 |
| jroll | which would likely fix this as well | 13:59 |
| lucasagomes | yup | 13:59 |
| lucasagomes | like a singleton | 13:59 |
| lucasagomes | using global vars is hard to make tests | 13:59 |
| lucasagomes | it's hard to clean states etc | 13:59 |
| lucasagomes | better to encapsulate it in a class | 13:59 |
| jroll | I guess we would still need locking | 14:00 |
| lucasagomes | I put one | 14:00 |
| jroll | but only for tests, I think | 14:00 |
| jroll | yeah | 14:00 |
| jroll | I was just thinking how to make this simplere | 14:00 |
| lucasagomes | yeah that was my first attempt, to put a lock on that _set_dhcp_provider function | 14:00 |
| *** linggao has joined #openstack-ironic | 14:01 | |
| lucasagomes | jroll, yeah it's not the simplest code to read :/ but races is hard | 14:01 |
| lucasagomes | if we can make it simpler I would +1 it :) | 14:02 |
| jroll | right | 14:03 |
| jroll | so loading on startup would fix the race if you're just running the conductor | 14:03 |
| jroll | but parallel tests could still break it | 14:03 |
| lucasagomes | jroll, well now I've a lock on the _set_dhcp... | 14:03 |
| lucasagomes | to prevent parallel tests to break it | 14:03 |
| jroll | I think it's fine, it's all in one place | 14:03 |
| jroll | right | 14:03 |
| jroll | what you did is fine :) | 14:04 |
| jroll | it's needed to fix parallel tests | 14:04 |
| lucasagomes | yeah | 14:04 |
| romcheg1 | I like it the way it is now | 14:04 |
| romcheg1 | It's simple enough, as for me | 14:04 |
| jroll | a simpler one line thing would fix the conductor, but not tests | 14:04 |
| jroll | agree | 14:04 |
| * jroll +2's | 14:04 | |
| lucasagomes | yup | 14:04 |
| lucasagomes | yeah... the tests were the problem | 14:04 |
| lucasagomes | parallel tests* | 14:04 |
| jroll | right | 14:04 |
| *** r-daneel has joined #openstack-ironic | 14:05 | |
| *** chuckC has quit IRC | 14:08 | |
| lucasagomes | gate's super slow today :( | 14:11 |
| *** romcheg1 has left #openstack-ironic | 14:11 | |
| NobodyCam | :( | 14:12 |
| jroll | s/today// | 14:13 |
| Shrews | jroll: lol... i was *just* typing that | 14:13 |
| Shrews | witch | 14:13 |
| lucasagomes | heh | 14:13 |
| Shrews | get out of my head | 14:13 |
| lucasagomes | I mean slower than normal | 14:14 |
| lucasagomes | :) | 14:14 |
| jroll | :D | 14:14 |
| lucasagomes | jroll, https://review.openstack.org/#/c/105795/ when you get some time | 14:16 |
| * jroll rechecks everything to speed the gate up | 14:16 | |
| jroll | sure | 14:16 |
| lucasagomes | I'm not very familiar with swift I suppose you may understang it better, but code LGTM | 14:16 |
| jroll | I like to pretend I might know things | 14:18 |
| *** viktors is now known as viktors|afk | 14:19 | |
| jroll | landed | 14:20 |
| *** killer_prince is now known as lazy_prince | 14:22 | |
| jroll | adam_g: we should land https://review.openstack.org/#/c/112134 soon pls :) | 14:23 |
| openstackgerrit | A change was merged to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/115878 | 14:24 |
| *** jistr has quit IRC | 14:29 | |
| lucasagomes | oh | 14:30 |
| * lucasagomes wonders about we should a gate test for iPXE and all | 14:31 | |
| lucasagomes | I think so | 14:31 |
| NobodyCam | jroll: Jenkins -1 that job? | 14:31 |
| jroll | NobodyCam: yeah | 14:32 |
| jroll | NobodyCam: that ping was more like "hey we should fix that and get it landed" :) | 14:32 |
| NobodyCam | :) | 14:33 |
| *** rainya has joined #openstack-ironic | 14:40 | |
| *** viktors|afk is now known as viktors | 14:48 | |
| *** romcheg1 has joined #openstack-ironic | 14:52 | |
| *** jcoufal has quit IRC | 14:53 | |
| Shrews | adam_g, romcheg1: I left a note on 112402 about the preserve_ephemeral option not being carried over, but I'm not sure if that's a big deal or not since it's doubtful that would be used (don't think rebuild was ever supported for nova bm). | 14:55 |
| *** chuckC has joined #openstack-ironic | 14:55 | |
| romcheg1 | adam_g: I'm going to take a look on my patches this evening. Thanks! | 14:56 |
| NobodyCam | Shrews: we do need that | 14:56 |
| romcheg1 | Whoops, it saw Shrews :) | 14:56 |
| romcheg1 | %s/saw/was | 14:56 |
| NobodyCam | :) morning romcheg1 :) | 14:56 |
| Shrews | NobodyCam: for migration? | 14:56 |
| romcheg1 | Good morning NobodyCam! | 14:56 |
| NobodyCam | and Shrews too | 14:57 |
| lucasagomes | any bash expers can explain to me why this works? echo hi > /tmp/data ya ; cat /tmp/data | 14:57 |
| NobodyCam | oh sorry nope ... for rebuild | 14:57 |
| lucasagomes | expert* | 14:57 |
| romcheg1 | lucasagomes: I think it's better poke lifeless about ^ | 14:58 |
| NobodyCam | lol wow that neet | 14:58 |
| lucasagomes | yeah | 14:58 |
| lucasagomes | super weird right? trying to google to understand the reasons, but I don't even know hw to google it | 14:58 |
| *** jistr has joined #openstack-ironic | 15:01 | |
| *** ndipanov has joined #openstack-ironic | 15:01 | |
| NobodyCam | lucasagomes: http://wiki.bash-hackers.org/syntax/redirection not an exact match but take a lok at the discussion stuff on the bottom | 15:01 |
| * lucasagomes checks | 15:02 | |
| openstackgerrit | A change was merged to openstack/ironic: Add support for interacting with swift https://review.openstack.org/105795 | 15:03 |
| openstackgerrit | Szymon Wróblewski proposed a change to openstack/ironic-python-agent: Remove unnecessary argument injection in agent sync and async decorators. https://review.openstack.org/117529 | 15:03 |
| lucasagomes | looks like it's just the order things gets evaluated... the redirection runs after so "echo hi > /tmp/data ya" has two pieces "> /tmp/data" which is the redirection and "echo hi ya"... so it runs echo and then the redirection | 15:04 |
| lucasagomes | idk | 15:04 |
| *** coolsvap has quit IRC | 15:07 | |
| NobodyCam | lucasagomes: ya thats my thought on it | 15:10 |
| NobodyCam | lucasagomes: echo > /tmp/data hi ya ; cat /tmp/data also works | 15:11 |
| lucasagomes | yup | 15:11 |
| lucasagomes | yeah... heh just an odd syntax :) | 15:11 |
| lucasagomes | jroll, NobodyCam https://review.openstack.org/#/c/116772 | 15:12 |
| lucasagomes | we gotta sort that, the last patch-set in nova has a -1 because of that I will try to clarify it | 15:12 |
| NobodyCam | I agree with jroll | 15:14 |
| NobodyCam | arch for the nova-compute host is not the right answer for us | 15:14 |
| NobodyCam | we could default to i386 if not found, but even that seem wrong to me | 15:15 |
| jroll | lucasagomes: we've talked about it a couple times now | 15:19 |
| jroll | I think NotFound is the right thing to do | 15:19 |
| lucasagomes | jroll, right... so I will comment on the nova patch about it | 15:20 |
| lucasagomes | https://review.openstack.org/#/c/111425/14/nova/virt/ironic/driver.py | 15:20 |
| jroll | lucasagomes: we also can't use canonicalize() because it's not backwards compat for tripleo | 15:20 |
| *** coolsvap has joined #openstack-ironic | 15:20 | |
| jroll | :( | 15:20 |
| lucasagomes | oh I was about to suggest us to canonicalize that | 15:20 |
| lucasagomes | jroll, but idk if it's a valid point for us | 15:20 |
| jroll | yeah, I tend to agree, but it will make lifeless rage :P | 15:20 |
| lucasagomes | I mean... is it an Ironic problem or an TripleO problem? (sure we need to work that out) | 15:20 |
| jroll | thing is they don't canonicalize the flavor | 15:21 |
| lucasagomes | hmm | 15:21 |
| jroll | so if you have e.g. i386 flavors, those won't match to i386 machines | 15:21 |
| jroll | because i386 machines will be read as i686 | 15:21 |
| lucasagomes | jroll, right I will add it in the comment as well | 15:21 |
| jroll | ok, thanks | 15:22 |
| NobodyCam | awesome TY lucasagomes | 15:22 |
| jroll | might be good to talk to daniel on irc | 15:22 |
| jroll | check queue is over 24 hours | 15:24 |
| jroll | amazing | 15:24 |
| lucasagomes | jroll, +1 to talk to him on IRC | 15:24 |
| *** HenryG has quit IRC | 15:28 | |
| jroll | lucasagomes: https://review.openstack.org/#/c/109590/8/nova/scheduler/filters/image_props_filter.py | 15:31 |
| jroll | lucasagomes: so we can use canonicalize() | 15:31 |
| jroll | actually, have to | 15:31 |
| lucasagomes | oh | 15:31 |
| lucasagomes | alright that's good | 15:31 |
| *** dlaube has joined #openstack-ironic | 15:32 | |
| lucasagomes | will make our life easier... problem is we don't have any way to say that the arch is unknown >.< | 15:32 |
| lucasagomes | jroll, canonicalize() returns None if the name is None | 15:33 |
| lucasagomes | maybe we should just leave it as None | 15:33 |
| lucasagomes | if it's unknown | 15:33 |
| jroll | oh huh | 15:33 |
| *** igordcard has joined #openstack-ironic | 15:34 | |
| *** foexle has quit IRC | 15:35 | |
| NobodyCam | brb | 15:44 |
| jroll | lucasagomes: so talking more, we should just not report the node's resources if the arch is unknown | 15:45 |
| jroll | from our perspective, that node is unusable | 15:45 |
| *** igordcard has quit IRC | 15:46 | |
| jroll | lucasagomes: going to add a comment on the nova review and probably push up a new ironic review | 15:47 |
| *** ndipanov has quit IRC | 15:47 | |
| jroll | going to want mrda and lifeless on board with this stuff ^ | 15:47 |
| lucasagomes | jroll, ack, please do | 15:51 |
| lucasagomes | jroll, NobodyCam ... a FYI I just had my report with thierry about J3, I marked the instance-info bp as Implemented even if the tempest is not merged yet | 15:51 |
| lucasagomes | as he instructed me to do so | 15:51 |
| jroll | \o/ | 15:52 |
| *** eghobo has joined #openstack-ironic | 15:54 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: IloVirtualMediaIscsi deploy driver https://review.openstack.org/113865 | 15:55 |
| *** todd_dsm has joined #openstack-ironic | 15:55 | |
| lucasagomes | jroll, so retuning the empty list for supported_instances seems the way to go | 15:57 |
| jroll | aha | 15:58 |
| jroll | I was just trying to figure that out | 15:58 |
| openstackgerrit | A change was merged to openstack/ironic: Update API document with BootDevice https://review.openstack.org/114415 | 16:00 |
| *** lazy_prince is now known as killer_prince | 16:00 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Nova review updates for _node_resource https://review.openstack.org/116772 | 16:02 |
| jroll | lucasagomes: ^ needs tests but wdyt | 16:02 |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: IloVirtualMediaAgent deploy driver https://review.openstack.org/115885 | 16:04 |
| *** penick has joined #openstack-ironic | 16:04 | |
| lucasagomes | jroll, ah looks simply/good | 16:05 |
| jroll | lucasagomes: fwiw, daniel says it looks good to him | 16:05 |
| jroll | :) | 16:05 |
| lucasagomes | nice one! | 16:05 |
| lucasagomes | alright... will +2 after tests | 16:05 |
| jroll | cool | 16:05 |
| jroll | gimme a few | 16:06 |
| lucasagomes | jroll, np... thanks for that | 16:06 |
| NobodyCam | awesome news lucasagomes TY | 16:07 |
| openstackgerrit | Vladyslav Drok proposed a change to openstack/ironic: Rewrite images tests with mock https://review.openstack.org/117549 | 16:10 |
| *** vdrok_ is now known as vdrok | 16:13 | |
| *** foexle has joined #openstack-ironic | 16:14 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Nova review updates for _node_resource https://review.openstack.org/116772 | 16:17 |
| JayF | jroll: I thought consensus was to /not/ canonicalize the cpu_arch? | 16:17 |
| jroll | lucasagomes: https://review.openstack.org/#/c/116772/ | 16:17 |
| jroll | JayF: read the comments on the nova review, turns out we have to do it | 16:17 |
| jroll | JayF: there's a patch up for nova right now that canonicalizes the arch on the image | 16:18 |
| lucasagomes | jroll, awesome! | 16:18 |
| lucasagomes | jroll, the link on the commit message is wrong | 16:19 |
| JayF | jroll: I don't see anything in that nova review that counteracts the point made yesterday that it will break existing Ironic installs (including ours) | 16:19 |
| lucasagomes | JayF, ^ | 16:19 |
| JayF | heh ^ maybe that is why | 16:19 |
| lucasagomes | https://review.openstack.org/#/c/111425/14 | 16:19 |
| lucasagomes | jroll, I will update it there | 16:19 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Nova review updates for _node_resource https://review.openstack.org/116772 | 16:20 |
| *** dguerri_ is now known as dguerri | 16:20 | |
| jroll | JayF: https://review.openstack.org/#/c/109590/8/nova/scheduler/filters/image_props_filter.py | 16:20 |
| jroll | lucasagomes: thanks :P | 16:21 |
| lucasagomes | np heh | 16:21 |
| *** todd_dsm has quit IRC | 16:23 | |
| NobodyCam | jumpping on conf call | 16:26 |
| *** penick has quit IRC | 16:27 | |
| *** todd_dsm has joined #openstack-ironic | 16:28 | |
| *** igordcard has joined #openstack-ironic | 16:28 | |
| lucasagomes | ack... I'm heading home now | 16:30 |
| lucasagomes | still gotta take the train back (came to the office) | 16:31 |
| lucasagomes | have a good night everybody | 16:31 |
| *** todd_dsm has quit IRC | 16:33 | |
| *** lucasagomes has quit IRC | 16:34 | |
| *** derekh has quit IRC | 16:35 | |
| JayF | mrda: NobodyCam: others, I think I have an issue with https://review.openstack.org/#/c/116772/5 -- we're going to break valid installs of Ironic | 16:43 |
| JayF | because if you weren't using the nova filter that looks at arch before | 16:44 |
| JayF | and what you set for arch isn't 'valid' per nova/libvirt standards, and you upgrade, you're going to break your installation pretty badly | 16:44 |
| JayF | comstud: ^ I moved some of our chatter into here | 16:44 |
| comstud | i'm not sure if there is a filter that looks at this | 16:45 |
| JayF | It's just hard for me to stomach us breaking some i->j migrations given we're writing code, and bending over backwards, to not break nova-bm->j | 16:45 |
| openstackgerrit | David J Hu proposed a change to openstack/python-ironicclient: Add keystone v3 CLI support https://review.openstack.org/111175 | 16:46 |
| *** jistr has quit IRC | 16:46 | |
| comstud | i guess it's the compute_capabilities_filter | 16:47 |
| comstud | ComputeCapabilitiesFilter | 16:47 |
| JayF | the filtering doesn't even matter though, afaict it just isn't going to advertise nodes that don't have a "valid" arch | 16:47 |
| JayF | just to make that one filter work more reliably I guess? | 16:47 |
| comstud | yes it does still advertise them | 16:47 |
| comstud | it catches the error and sets arch to None | 16:48 |
| JayF | comstud: https://review.openstack.org/#/c/116772/5/ironic/nova/virt/ironic/driver.py,cm it returns an empty list for supported_instances | 16:48 |
| comstud | yeah | 16:48 |
| comstud | that's not important | 16:48 |
| JayF | why not? | 16:48 |
| comstud | I guess the image_props_filter does look at that | 16:49 |
| comstud | it's not important as far as 'advertising the node' | 16:49 |
| comstud | but | 16:49 |
| comstud | it would break image_props_filter matching for people with certain configs | 16:50 |
| JayF | If someone set some 'invalid' arch (like lets say 386 vs i386) and was not filtering on arch before | 16:50 |
| comstud | from what I can tell | 16:50 |
| JayF | would their stuff continue to work? | 16:50 |
| openstackgerrit | David J Hu proposed a change to openstack/python-ironicclient: Add keystone v3 CLI support https://review.openstack.org/111175 | 16:50 |
| comstud | well | 16:52 |
| JayF | that's the crux of my concern | 16:53 |
| comstud | as far as I can tell | 16:53 |
| comstud | supported_instances is supposed to be a dict anyway | 16:53 |
| comstud | and not a list | 16:53 |
| comstud | :) | 16:53 |
| comstud | so it's already broken | 16:53 |
| comstud | therefore no one must be using it | 16:53 |
| comstud | ;) | 16:53 |
| JayF | hah | 16:53 |
| JayF | you should probably put that on the review then? | 16:53 |
| comstud | i'm going to ask it as a question | 16:53 |
| comstud | because i'm not sure | 16:53 |
| comstud | but that's sure what it looks like | 16:53 |
| JayF | and if that's true, then that means anyone who relied on that would be broken in I? | 16:53 |
| comstud | you mean J? | 16:55 |
| comstud | er | 16:55 |
| comstud | yeah | 16:55 |
| comstud | I was last release | 16:55 |
| comstud | the question is... did it even work in I ? | 16:55 |
| JayF | yeah if it never worked in I | 16:55 |
| JayF | then my concern is completely out the window | 16:55 |
| comstud | yeah | 16:56 |
| *** lsmola has quit IRC | 16:56 | |
| comstud | looking at this image filter in nova | 16:56 |
| comstud | i'll show you what AI'm looking at here | 16:56 |
| comstud | https://github.com/openstack/nova/blob/master/nova/scheduler/filters/image_props_filter.py#L79 | 16:56 |
| comstud | https://github.com/openstack/nova/blob/master/nova/scheduler/filters/image_props_filter.py#L65 | 16:56 |
| comstud | actually, maybe this is just comparing values | 16:57 |
| comstud | nevermind | 16:57 |
| comstud | So I guess this is a tuple | 16:57 |
| comstud | https://github.com/openstack/nova/blob/master/nova/scheduler/filters/image_props_filter.py#L47 | 16:57 |
| comstud | that's just... weird. | 16:57 |
| comstud | Well, ignore everything I just said | 16:57 |
| JayF | I'd be interested in your opinion about whether or not it's OK to break in this way? | 16:58 |
| *** eghobo has quit IRC | 16:59 | |
| comstud | generally i do not think this is okay | 16:59 |
| *** eghobo has joined #openstack-ironic | 16:59 | |
| JayF | can you think of a way to satisfy the nova core's concern without breaking? | 16:59 |
| JayF | because I'm at the point of I don't think we should make that change | 16:59 |
| JayF | but I don't know how to appease the nova reviewers without it | 16:59 |
| JayF | I actually think an arch of unknown or None should be okay :/ | 17:00 |
| comstud | I don't know.. I"m a nova core | 17:00 |
| comstud | I left a -1 on this | 17:00 |
| comstud | i'm not sure who you were referring to | 17:00 |
| JayF | I know you are, that's why I wanted your opinion :) | 17:00 |
| comstud | dberrange? | 17:00 |
| JayF | yeah, I just couldn't remember the name | 17:00 |
| JayF | heh | 17:00 |
| comstud | he probably didn't understand there's a behavior change here | 17:02 |
| comstud | i mean | 17:02 |
| comstud | that it would stop making image props filter match things depending on ppls configs | 17:02 |
| *** harlowja_away is now known as harlowja_ | 17:03 | |
| *** harlowja_ is now known as harlowja_away | 17:04 | |
| *** penick has joined #openstack-ironic | 17:06 | |
| *** harlowja_away is now known as harlowja_ | 17:09 | |
| *** penick has quit IRC | 17:09 | |
| kylestev | this is really starting to piss me off, anyone know what's up? https://gist.github.com/kylestev/8446116d75b42bf339fc | 17:12 |
| JayF | that usually means you need to rebase to get some requirements.txt changes | 17:13 |
| JayF | the actual diff of the config files usually can tell you the culprit | 17:13 |
| JayF | i.e. "oh this is missing some swiftclient config" means you probably have a newer/older swiftclient than you should | 17:13 |
| kylestev | hmmmmm | 17:14 |
| JayF | actually it looks like your check_uptodate script might be broken | 17:14 |
| JayF | are you running that on mac or on linux? | 17:14 |
| JayF | and if on mac, are you sure it's supposed to work there | 17:14 |
| kylestev | oh dammit... | 17:14 |
| kylestev | i do my git workflow on my mac but run tests on my vm... that is likely the culprit | 17:15 |
| kylestev | JayF: I feel really stupid, but thank you for pointing out the issue :) | 17:15 |
| JayF | no problem | 17:16 |
| openstackgerrit | Kyle Stevenson proposed a change to openstack/ironic: Add implementation of pluggable distributed sync https://review.openstack.org/115148 | 17:19 |
| *** dguerri is now known as dguerri`afk | 17:20 | |
| *** igordcard has quit IRC | 17:28 | |
| *** penick has joined #openstack-ironic | 17:33 | |
| openstackgerrit | Kyle Stevenson proposed a change to openstack/ironic: Add implementation of ZooKeeper SyncManager https://review.openstack.org/117576 | 17:37 |
| *** bandicot has joined #openstack-ironic | 17:41 | |
| *** bandicot is now known as bensonmiller | 17:41 | |
| *** shakamunyi has joined #openstack-ironic | 17:41 | |
| openstackgerrit | Kyle Stevenson proposed a change to openstack/ironic: Add implementation of ZooKeeper SyncManager https://review.openstack.org/117576 | 17:41 |
| openstackgerrit | Kyle Stevenson proposed a change to openstack/ironic: Add implementation of pluggable distributed sync https://review.openstack.org/115148 | 17:41 |
| *** dtantsur is now known as dtantsur|afk | 17:52 | |
| *** viktors is now known as viktors|afk | 17:54 | |
| openstackgerrit | Adam Gandelman proposed a change to openstack/ironic: Script to migrate Nova BM data to Ironic https://review.openstack.org/112402 | 17:58 |
| *** pelix1 has quit IRC | 17:59 | |
| *** romcheg1 has quit IRC | 18:08 | |
| *** romcheg1 has joined #openstack-ironic | 18:14 | |
| *** romcheg1 has quit IRC | 18:14 | |
| Shrews | ack-grep tells me that we don't actually use the INIT ironic state | 18:23 |
| * Shrews wonders if we should/ever did | 18:25 | |
| *** romcheg1 has joined #openstack-ironic | 18:26 | |
| Shrews | ... "Any node with non-empty `properties` is said to be "initialized", and the state is INIT.' ... | 18:27 |
| JayF | wait we have a state called INIT? | 18:28 |
| JayF | I've never seen that in the real world :) | 18:28 |
| *** romcheg1 has joined #openstack-ironic | 18:29 | |
| adam_g | Shrews, is that grepped from nova or ironic? | 18:42 |
| Shrews | ironic | 18:42 |
| Shrews | nova driver doesn't use it either | 18:42 |
| Shrews | i can't imagine any other part of nova outside of the driver referencing it | 18:43 |
| *** todd_dsm has joined #openstack-ironic | 18:44 | |
| *** todd_dsm has quit IRC | 18:44 | |
| *** romcheg1 has left #openstack-ironic | 18:47 | |
| *** ramineni has quit IRC | 18:52 | |
| NobodyCam | brb | 18:56 |
| adam_g | Shrews, i came across this last week, https://review.openstack.org/#/c/103402/ | 19:01 |
| Shrews | adam_g: oh, good find | 19:04 |
| adam_g | oh cool. looks like eventlet released 0.15.1 with the fix for our paramiko issue | 19:06 |
| *** penick has quit IRC | 19:06 | |
| lifeless | jroll: +1 on not exporting unknown arch nodes. | 19:08 |
| lifeless | jroll: but | 19:08 |
| lifeless | jroll: what happens if a user makes the arch unnknown after instance is on a node | 19:08 |
| *** penick has joined #openstack-ironic | 19:12 | |
| *** openstackgerrit has quit IRC | 19:31 | |
| *** openstackgerrit has joined #openstack-ironic | 19:33 | |
| NobodyCam | brb | 19:34 |
| JayF | quick+easy ironic-specs review, just some cleanup in requirements.txt from the docs team, needs +2a --> https://review.openstack.org/#/c/117180/2 | 19:40 |
| JayF | dtantsur|afk: ^ you had originally +2'd that, but it had to be rebased | 19:40 |
| *** penick has quit IRC | 19:55 | |
| *** penick has joined #openstack-ironic | 19:59 | |
| NobodyCam | JayF: LGTM +a'd | 20:05 |
| NobodyCam | have to run in to town... bbaib | 20:09 |
| openstackgerrit | A change was merged to openstack/ironic-specs: Remove docutils pin https://review.openstack.org/117180 | 20:09 |
| *** bensonmiller has quit IRC | 20:09 | |
| NobodyCam | JayF: ^^^^^^ | 20:11 |
| JayF | ty | 20:12 |
| *** foexle has quit IRC | 20:32 | |
| *** bandicot has joined #openstack-ironic | 20:40 | |
| *** penick has quit IRC | 21:06 | |
| * NobodyCam is back | 21:06 | |
| *** linggao has quit IRC | 21:09 | |
| *** penick has joined #openstack-ironic | 21:12 | |
| *** foexle has joined #openstack-ironic | 21:15 | |
| *** dhellmann is now known as dhellmann_ | 21:17 | |
| mrda | Morning Ironic | 21:19 |
| mrda | jroll, JayF: I'm just thinking of abandoning 116772 | 21:20 |
| JayF | My primary objection is that is has a likelyhood of breaking working--although possibly semantically invalid--ironic installations | 21:21 |
| mrda | So there's enough heat on https://review.openstack.org/#/c/111425/14/nova/virt/ironic/driver.py to suggest we are going to have to do *something* | 21:24 |
| JayF | my suggestion: if arch.canonicalize() can't resolve an arch | 21:25 |
| mrda | What do you think of canonicalize() but no support for 'NotFound'. i.e. not returning an arch | 21:25 |
| JayF | it should spit out an arch that indicates unknown | 21:25 |
| JayF | or None or something along those lines | 21:25 |
| JayF | so that the people who are working will continue working similar to how they are now | 21:25 |
| JayF | either that or we eat the breakage | 21:25 |
| JayF | I just wanted to make damn sure we didn't break people without it being *well known* we could be breaking existing installs | 21:25 |
| JayF | especially after the gymnastics that have been done to not-break nova bm | 21:26 |
| mrda | So dan b's comment is that there's no unknown arch. So we shouldn't add it to extra_specs | 21:26 |
| mrda | which will result in that node not being scheduled | 21:26 |
| JayF | So if we make an unknown arch, we solve the problem, right? | 21:26 |
| JayF | heh | 21:26 |
| JayF | jroll actually put a patch into nova to avoid this impacting our install | 21:27 |
| JayF | because all the nodes for our stuff have arch in the db as 'amd64' | 21:27 |
| mrda | I agree with Matt's comments. Ifd we don't know the arch we shouldn't be scheduling paying customers to that node | 21:27 |
| JayF | https://review.openstack.org/#/c/117555/ | 21:27 |
| JayF | mrda: I think it's one of those situations where all choices are wrong. That's why my comment didn't come with a -1 | 21:28 |
| mrda | Well, if we don't cvome up with a solution, this won't merge and we'll be aiming for K | 21:28 |
| mrda | I think what I'm going to propose is this: | 21:31 |
| mrda | cpu_arch = str(node.properties.get('cpu_arch', 'NotFound')) | 21:32 |
| mrda | if cpu_arch == 'NotFound': | 21:32 |
| mrda | LOG.error("You have a borked node ${node}", node.uuid) | 21:33 |
| mrda | cpu_arch = "" | 21:33 |
| mrda | or a more pythonic equivalent | 21:34 |
| JayF | so still advertise the node or not? | 21:34 |
| JayF | er, I mean, capablities going to be [] or ["", baremetal, baremetal] | 21:34 |
| mrda | So this will push the problem along. | 21:35 |
| openstackgerrit | A change was merged to openstack/ironic: Fix unit tests with keystoneclient master https://review.openstack.org/117314 | 21:35 |
| mrda | If we want to schedule a node where the arch is "", then the sched can make that choice. | 21:36 |
| JayF | I like that solution much better, but I wonder if it'd be acceptable to daniel | 21:36 |
| mrda | We shall see | 21:37 |
| mrda | the problem I face, from a process perspective, is that I should really be pitching this to Ironic first to get all of our agreements, but the ones I need to satisfy are the nove reviewers with issues | 21:38 |
| JayF | I'd hope the reviewers would try and help out by making comments in irc (here) if they see an issue | 21:39 |
| JayF | but it doesn't appear that's going to be the case | 21:39 |
| JayF | it's just hard to "push" the conversation when people don't seem to want to take the time other than in gerritt | 21:39 |
| mrda | I think the best I can do is to subscribe ironic peeps to the patch in Nova | 21:40 |
| NobodyCam | morning mrda | 21:42 |
| mrda | hey NobodyCam | 21:42 |
| NobodyCam | what review | 21:43 |
| NobodyCam | 116772 | 21:45 |
| mrda | ya | 21:46 |
| mrda | NobodyCam: the problem is that we can discuss things in ironic, but nova has some big opinions over in https://review.openstack.org/#/c/111425/14/nova/virt/ironic/driver.py | 21:47 |
| mrda | NobodyCam: Any soluytion we come up with is unlikely to be accepted in the Nova review, so we might as well debate it over there. | 21:47 |
| mrda | NobodyCam: thoughts? | 21:47 |
| NobodyCam | We don't want paying customers scheduled to boxes that can't actually service the request | 21:48 |
| NobodyCam | i just dont get that | 21:48 |
| NobodyCam | still reading up | 21:49 |
| lifeless | mrda: sadface | 21:52 |
| mrda | lifeless: jump in if you want | 21:53 |
| mrda | lifeless: but we need a resolution | 21:53 |
| mrda | today | 21:53 |
| mrda | T-5 until J closes | 21:54 |
| NobodyCam | mrda: is Daniel on in #nova | 21:56 |
| JayF | i don't think so, looking at /names | 21:56 |
| *** chuckC has quit IRC | 21:56 | |
| * mrda is looking | 21:57 | |
| NobodyCam | ya | 21:57 |
| lifeless | commented | 21:58 |
| lifeless | so not advertising unknown arch nodes got a +1 | 21:58 |
| lifeless | from berrange | 21:58 |
| lifeless | jroll: scheduler canonicalises now, or will sooin ? | 21:58 |
| lifeless | jroll: I'd argue that if the scheduler filter will *start* canonicalising soon, then when it does the ironic driver should, but not before because it will be broken in the interim, | 21:59 |
| JayF | lifeless: jroll is OOO right now, but we're all still talking about that patch | 21:59 |
| openstackgerrit | Kyle Stevenson proposed a change to openstack/ironic: Add implementation of ZooKeeper SyncManager https://review.openstack.org/117576 | 22:02 |
| JayF | lifeless: also myself and comstud went through some scenarios earlier; some people with working installs now (including mine) will break even if both sides canonicalize | 22:03 |
| JayF | lifeless: if you were not using a scheduler that cared about arch + you put in an arch that isn't in nova's class (like we did; amd64) | 22:04 |
| JayF | lifeless: obviously someone can fix the DB before upgrade; but it's still something that can break a working (arguably invalid, though) install of ironic | 22:04 |
| lifeless | JayF: oh, canonicalise tossesout things it doesn't recognise rather than failing open? | 22:04 |
| JayF | it returns None if not known | 22:05 |
| lifeless | ugh | 22:05 |
| lifeless | stabbity stab stab | 22:05 |
| JayF | which 116772 would essentially discard | 22:05 |
| JayF | exactly | 22:05 |
| lifeless | this will break tripleo definitely then | 22:05 |
| JayF | it's being reviewed as if it's 'new code' but it's been running in peoples' environments for a long time | 22:05 |
| lifeless | why exactly is Nova doing this? | 22:05 |
| JayF | Read the review? I'm not sure I get it other than a drive for correctness | 22:05 |
| JayF | 'correctness' | 22:05 |
| JayF | all this chatter happening here -> https://review.openstack.org/#/c/111425/14/nova/virt/ironic/driver.py | 22:06 |
| JayF | which you have comments on so obv you've read it | 22:06 |
| NobodyCam | that seems > feature parity | 22:06 |
| NobodyCam | https://github.com/openstack/nova/blob/master/nova/virt/baremetal/driver.py#L145-L153 | 22:07 |
| comstud | JayF: thought we were in sync... our install will not break | 22:09 |
| comstud | unless you have your own install somewhere that will break | 22:10 |
| comstud | that you're referring to | 22:10 |
| comstud | :) | 22:10 |
| JayF | comstud: hm, okay, I didn't understand what you meant then | 22:10 |
| comstud | we do not use any of the capability or image property mappings | 22:10 |
| comstud | matchings | 22:10 |
| comstud | filters | 22:10 |
| JayF | 10-4 | 22:10 |
| comstud | but yes, this would certainly break people that do and have broken arch names | 22:11 |
| comstud | i think it's limited to the image property filter from what I saw | 22:11 |
| comstud | ie, someone configured "mybogusarch" for cpu_arch in ironic... had the image property sched filter loaded... and had images with "mybogusarch" set for the image 'architecture' property.. | 22:12 |
| comstud | they'd see things break | 22:12 |
| comstud | +if someone | 22:12 |
| NobodyCam | ok heres a thought, could we use the argument that 116772 would break current nova-bm installs that want to mirgrate to Ironic | 22:21 |
| NobodyCam | s/would/could/ | 22:21 |
| mrda | So lifeless, data migration patches welcoemd :) | 22:21 |
| JayF | lifeless: also fwiw --> https://review.openstack.org/#/c/117555/ | 22:22 |
| JayF | lifeless: I had no idea 'amd64' was used by others. Nice to know :) | 22:22 |
| NobodyCam | should try and get folks in #nova? | 22:25 |
| mrda | NobodyCam: they've just finished a meeting, so likely to be around | 22:26 |
| *** chuckC has joined #openstack-ironic | 22:26 | |
| NobodyCam | :) | 22:26 |
| lifeless | JayF: ./scripts/configure-tripleo~: export NODE_ARCH=amd64 | 22:27 |
| lifeless | JayF: we use that an i386 | 22:27 |
| mrda | bbs, schoolrun | 22:29 |
| mrda | . | 22:33 |
| NobodyCam | :) | 22:37 |
| NobodyCam | so I guess no one in nova up for chatting | 22:44 |
| NobodyCam | :( | 22:44 |
| mrda | :( | 22:45 |
| NobodyCam | this looks like fun: but not sure how to get my car there | 22:47 |
| NobodyCam | http://www.dubaiautodrome.com/da-track-days-cars-bikes/ | 22:47 |
| openstackgerrit | A change was merged to openstack/ironic: Updated from global requirements https://review.openstack.org/117423 | 22:47 |
| *** bandicot has quit IRC | 22:48 | |
| NobodyCam | brb | 22:49 |
| NobodyCam | mrda: are you back | 23:01 |
| mrda | yup | 23:02 |
| mrda | NobodyCam: ^^ | 23:02 |
| NobodyCam | whats your thought for our game plan if noone in #nova wants to chat? | 23:02 |
| mrda | 1) I think discussion should take place on the Nova patch itself | 23:03 |
| mrda | 2) I'm thinking of pushing a patch that doesn't set cpu_arch if it's not defined (aka not use NotFound) | 23:04 |
| mrda | 3) But will use canonicalize() if it is set | 23:04 |
| NobodyCam | oh | 23:04 |
| mrda | So I understand the scheduler won't like cpu_arch not being set, currently | 23:05 |
| mrda | But I do agree with Matt Diaz's comment that if we don't know what a node's cpu_arch is, we probably don't want to be scheduling workloads to it | 23:05 |
| NobodyCam | but that what nova bm dose : https://github.com/openstack/nova/blob/master/nova/virt/baremetal/driver.py#L145-L153 | 23:05 |
| * mrda look | 23:05 | |
| mrda | s | 23:05 |
| mrda | ok, so we could do likewise | 23:06 |
| NobodyCam | ya | 23:07 |
| NobodyCam | its whats there | 23:07 |
| NobodyCam | now | 23:07 |
| mrda | I think that node will be ignored in scheduling though | 23:07 |
| NobodyCam | which may not be a bad thing | 23:07 |
| mrda | agreed | 23:07 |
| JayF | it's ignored in scheduling IF you're using one of the schedulers that look at that | 23:08 |
| mrda | so jroll's patch to alias x86_64 to amd64 will help with the canonicalize | 23:08 |
| mrda | but there still might be some pain here | 23:08 |
| mrda | I think what we've discussed now is worth putting a patch up for | 23:09 |
| mrda | and seeing what people think of that | 23:09 |
| NobodyCam | mrda: +1 | 23:09 |
| mrda | T-5 and all that :) | 23:09 |
| NobodyCam | 1066 and all that? | 23:10 |
| NobodyCam | lol | 23:10 |
| * mrda has missed the reference | 23:10 | |
| mrda | T-5 days until Nova freezes for J | 23:10 |
| NobodyCam | http://en.wikipedia.org/wiki/1066_and_All_That | 23:11 |
| * mrda looks | 23:11 | |
| NobodyCam | lol :) | 23:11 |
| mrda | :) | 23:14 |
| NobodyCam | mrda: adding a note to 116772 to address concerns to 111425 | 23:15 |
| mrda | Thanks NobodyCam - I'll probably backport 111425 into 116772 so the drivers continue to be consistent | 23:16 |
| NobodyCam | ack until then just trying to get everyone on the same review :) | 23:17 |
| mrda | :) | 23:20 |
| *** bandicot has joined #openstack-ironic | 23:25 | |
| *** penick has quit IRC | 23:25 | |
| *** shakamunyi has quit IRC | 23:33 | |
| *** bandicot has quit IRC | 23:36 | |
| *** openstackgerrit has quit IRC | 23:46 | |
| *** openstackgerrit has joined #openstack-ironic | 23:48 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!