| spandhe | so https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ipxe_config.template#L10 is to make sure it boots after that? | 00:00 |
|---|---|---|
| *** ryanpetrello has quit IRC | 00:00 | |
| spandhe | so if I dont want it to boot, e.g. for localboot or something, I need to keep exit 0 or something? | 00:01 |
| *** Masahiro has joined #openstack-ironic | 00:01 | |
| NobodyCam | http://ipxe.org/cmd/chain?redirect=1 | 00:01 |
| NobodyCam | thats info on the ipxe boot command | 00:02 |
| jroll | ipxe is for pxe booting | 00:02 |
| jroll | you don't want ipxe if you are local booting | 00:02 |
| spandhe | NobodyCam: great! thanks! | 00:02 |
| NobodyCam | why whould you not want it to boot | 00:02 |
| spandhe | jroll: so how to localboot after ipxe booting? | 00:03 |
| NobodyCam | spandhe: jroll makes a very valid point depending on what your trying to do | 00:03 |
| jroll | spandhe: why would you do that? just set it to boot from disk instead? | 00:03 |
| NobodyCam | spandhe: are you deploying a whole disk image? | 00:03 |
| spandhe | NobodyCam: yes.. | 00:04 |
| *** dlaube has quit IRC | 00:04 | |
| *** smoriya has joined #openstack-ironic | 00:04 | |
| spandhe | jroll: thats what I meant by localboot.. boot from disk | 00:04 |
| NobodyCam | are you using IPA? | 00:04 |
| jroll | spandhe: ipxe is for pxe booting (network booting) | 00:05 |
| *** Masahiro has quit IRC | 00:05 | |
| jroll | spandhe: if you're booting from disk, you don't use pxe at all | 00:05 |
| NobodyCam | spandhe: take a look at the agent_ipmitool driver | 00:06 |
| spandhe | jroll: after network boot, the host reboots, right? | 00:06 |
| * NobodyCam assumes ipmi and real hardware | 00:07 | |
| jroll | spandhe: no, it just gets a kernel and ramdisk and boots | 00:07 |
| jroll | it downloads those from the network | 00:08 |
| jroll | and boots them | 00:08 |
| spandhe | jroll: ok I see your point | 00:08 |
| NobodyCam | if it would auto update I'd love a driver feature matrix page | 00:08 |
| jroll | you can have pxe or local, not both | 00:08 |
| spandhe | I tried putting boot without kernel and initrd… it failed | 00:08 |
| *** yuanying has joined #openstack-ironic | 00:08 | |
| jroll | (though I wouldn't be surprised if ipxe could bootstrap from a disk | 00:08 |
| jroll | ) | 00:08 |
| spandhe | so even with pxe, when we want to boot from disk, we used the switched pxe config, right? | 00:09 |
| jroll | spandhe: you use driver.management.set_boot_device('disk') | 00:09 |
| jroll | no | 00:09 |
| jroll | no | 00:09 |
| jroll | pxe isn't not involved with booting from disk at all | 00:09 |
| jroll | I have to step away for a bit, sorry | 00:09 |
| spandhe | jroll: me too.. ill talk to you guys later.. dont think we are on the same page.. or my understanding is completely wrong :P | 00:10 |
| *** ryanpetrello has joined #openstack-ironic | 00:11 | |
| *** yuanying_ has quit IRC | 00:11 | |
| NobodyCam | brb | 00:17 |
| spandhe | btw I am doing boot from disk because I am using Anaconda | 00:19 |
| spandhe | for Anaconda, we need to specify localboot 0 | 00:19 |
| spandhe | after the pxebooting is done | 00:20 |
| *** lucas-dinner has quit IRC | 00:23 | |
| devananda | spandhe: what do you maen, you're using Anaconda? | 00:25 |
| spandhe | devananda: I am using Anaconda installer.. | 00:25 |
| devananda | that's an installer .... | 00:25 |
| spandhe | right.. not using ironic python agent.. | 00:25 |
| devananda | spandhe: so, Ironic doesn't use installers. it writes a machine image to a machine | 00:26 |
| *** andreykurilin_ has quit IRC | 00:28 | |
| devananda | spandhe: does that help clear up why you don't need to localboot? | 00:32 |
| * NobodyCam is back | 00:33 | |
| spandhe | devananda have to head out urgently.. will be back again later.. | 00:41 |
| NobodyCam | is there a public picture of Pixie Boots? | 00:43 |
| *** Marga_ has quit IRC | 00:46 | |
| *** ryanpetrello has quit IRC | 00:46 | |
| smoriya | Hello Ironic, | 00:47 |
| NobodyCam | hello there smoriya | 00:47 |
| smoriya | I'm new to Ironic and curious to know about the reason why Ironic doesn't use installer? | 00:48 |
| smoriya | I mean, in general, we can install OS after using pxe boot but Ironic doesn't. Why? | 00:48 |
| NobodyCam | we deploy images.. generally folks like to build a image for a task and then deploy that | 00:49 |
| *** spandhe has quit IRC | 00:49 | |
| NobodyCam | one could deploy a image with something like a kickstart or preceed type file backed in | 00:49 |
| NobodyCam | smoriya: have you seen any tools like DIck image builder? | 00:50 |
| devananda | smoriya: repeatability and performance | 00:50 |
| smoriya | NobodyCam: Yes | 00:50 |
| devananda | smoriya: when you start an instance in OpenStack, or in Amazon, or $anywhere in a cloud, does it use an installer? | 00:50 |
| NobodyCam | ieek that was a really bad type-oh | 00:50 |
| devananda | NobodyCam: lol | 00:51 |
| NobodyCam | :( *blush* | 00:51 |
| *** ryanpetrello has joined #openstack-ironic | 00:51 | |
| smoriya | @devananda: Hmm, OK. | 00:51 |
| devananda | smoriya: Ironic is not just a bare metal management tool. It's specifically geared towards enabling a cloud-like usage and workflow with physical servers | 00:52 |
| devananda | smoriya: there are plenty of the former already around :) | 00:52 |
| smoriya | @devananda: Thank you! I understand :) | 00:53 |
| NobodyCam | smoriya: thank you for asking about Ironic | 00:55 |
| devananda | smoriya: np! glad to help :) | 00:55 |
| *** Masahiro has joined #openstack-ironic | 00:55 | |
| *** ryanpetrello has quit IRC | 00:56 | |
| rloo | hi devananda, before I forget again, did NobodyCam or jogo ask you about https://bugs.launchpad.net/nova/+bug/1379373 ? | 00:57 |
| smoriya | BTW, I'm interested in https://blueprints.launchpad.net/ironic/+spec/cinder-integration | 00:57 |
| smoriya | Does someone active on this BP for kilo? | 00:57 |
| rloo | devananda: jogo's question was whether this is still needed: nova/api/openstack/compute/contrib/baremetal_nodes.py | 00:57 |
| devananda | rloo: yes and yes. | 01:00 |
| NobodyCam | smoriya: we had some folk looking at it eariler | 01:00 |
| rloo | devananda: great. My job as messenger is done then ;) | 01:00 |
| devananda | rloo: NobodyCam: we should continue to maintain the proxy for this cycle, at a minimum. It's up to Nova cores whether that should remain beyond the Kilo cycle or not | 01:02 |
| NobodyCam | devananda: ack I have a question posted on the bug. I think I know what happened | 01:02 |
| NobodyCam | also have a ping for dafna in NOva | 01:03 |
| *** pcrews has joined #openstack-ironic | 01:03 | |
| devananda | NobodyCam: yea. afaict, the proxy shouldn't ever generate the error mentioned in that report | 01:04 |
| *** arif-ali has quit IRC | 01:05 | |
| devananda | I have changed the bug status to incomplete | 01:06 |
| *** arif-ali has joined #openstack-ironic | 01:06 | |
| NobodyCam | this patch removed a check we had: https://github.com/openstack/nova/commit/46ed619b9ab1d61582f36155eea0d4a88f31fd50#diff-75f16f54de93fdcfd1fadd84b04247a7 | 01:06 |
| NobodyCam | so that error is very strange | 01:06 |
| *** Masahiro has quit IRC | 01:07 | |
| NobodyCam | but its clearly looking for the nova_bm tables | 01:07 |
| smoriya | NobodyCam: Thank you | 01:07 |
| NobodyCam | smoriya: I say that and also add that there is no one "actively" working on it. | 01:08 |
| NobodyCam | it needs a kilo spec | 01:09 |
| *** Masahiro has joined #openstack-ironic | 01:10 | |
| NobodyCam | thou not in channel BertieFulton was the one looking into it around the summit | 01:10 |
| smoriya | Thanks. I'll make contact with hemna to ask whether I can do something:) | 01:11 |
| NobodyCam | :) | 01:11 |
| devananda | NobodyCam: I've commentd on the bug | 01:16 |
| *** penick has quit IRC | 01:16 | |
| NobodyCam | devananda: ack.. Thank you :) .. I'll keep an eye on it if it gets updated | 01:17 |
| *** ChuckC has quit IRC | 01:17 | |
| *** ChuckC has joined #openstack-ironic | 01:18 | |
| *** ryanpetrello has joined #openstack-ironic | 01:18 | |
| *** ChuckC has quit IRC | 01:22 | |
| *** ChuckC has joined #openstack-ironic | 01:31 | |
| *** rloo has quit IRC | 01:32 | |
| *** ryanpetrello_ has joined #openstack-ironic | 01:33 | |
| *** ryanpetrello has quit IRC | 01:34 | |
| *** ryanpetrello_ is now known as ryanpetrello | 01:34 | |
| *** dlaube has joined #openstack-ironic | 01:39 | |
| *** dlaube has quit IRC | 01:40 | |
| *** Kamilion has left #openstack-ironic | 01:47 | |
| *** nosnos has joined #openstack-ironic | 01:52 | |
| *** ryanpetrello has quit IRC | 01:55 | |
| *** lazy_prince is now known as killer_prince | 02:09 | |
| *** killer_prince is now known as lazy_prince | 02:09 | |
| *** Marga_ has joined #openstack-ironic | 02:18 | |
| *** Marga_ has quit IRC | 02:19 | |
| *** Masahiro has quit IRC | 02:20 | |
| *** Masahiro has joined #openstack-ironic | 02:23 | |
| *** marcoemorais1 has quit IRC | 02:25 | |
| *** lazy_prince is now known as killer_prince | 02:26 | |
| *** Marga_ has joined #openstack-ironic | 02:33 | |
| *** Marga_ has quit IRC | 02:33 | |
| *** Marga_ has joined #openstack-ironic | 02:34 | |
| *** r-daneel has quit IRC | 02:36 | |
| *** ryanpetrello has joined #openstack-ironic | 02:37 | |
| *** Masahiro has quit IRC | 02:38 | |
| *** igordcard has quit IRC | 02:39 | |
| *** Masahiro has joined #openstack-ironic | 02:41 | |
| *** ramineni has joined #openstack-ironic | 02:47 | |
| *** Masahiro has quit IRC | 03:18 | |
| *** Masahiro has joined #openstack-ironic | 03:19 | |
| *** nosnos has quit IRC | 03:23 | |
| *** Masahiro has quit IRC | 03:27 | |
| *** Masahiro has joined #openstack-ironic | 03:30 | |
| *** naohirot has quit IRC | 03:32 | |
| *** chenglch has joined #openstack-ironic | 03:49 | |
| *** naohirot has joined #openstack-ironic | 04:03 | |
| *** pensu has joined #openstack-ironic | 04:04 | |
| *** pcrews has quit IRC | 04:08 | |
| *** chenglch has quit IRC | 04:13 | |
| *** yuanying_ has joined #openstack-ironic | 04:15 | |
| *** nosnos has joined #openstack-ironic | 04:16 | |
| *** achanda has joined #openstack-ironic | 04:17 | |
| *** yuanying has quit IRC | 04:18 | |
| *** pensu has quit IRC | 04:18 | |
| *** Nisha has joined #openstack-ironic | 04:23 | |
| *** pcrews has joined #openstack-ironic | 04:28 | |
| *** Masahiro has quit IRC | 04:31 | |
| *** enterprisedc has quit IRC | 04:35 | |
| *** pcrews has quit IRC | 04:35 | |
| *** krtaylor has quit IRC | 04:41 | |
| *** Masahiro has joined #openstack-ironic | 04:42 | |
| *** rushiagr_away is now known as rushiagr | 04:48 | |
| ramineni | JoshNang: hi | 04:50 |
| *** krtaylor has joined #openstack-ironic | 04:54 | |
| *** chenglch|3 has joined #openstack-ironic | 04:58 | |
| *** Hefeweiz1n has joined #openstack-ironic | 05:04 | |
| *** Hefeweizen has quit IRC | 05:04 | |
| *** Masahiro has quit IRC | 05:06 | |
| *** pensu has joined #openstack-ironic | 05:08 | |
| *** jmank has quit IRC | 05:09 | |
| *** vipul has quit IRC | 05:09 | |
| *** jmank has joined #openstack-ironic | 05:09 | |
| *** jmank has quit IRC | 05:10 | |
| *** vipul has joined #openstack-ironic | 05:10 | |
| *** jmank has joined #openstack-ironic | 05:10 | |
| *** achanda has quit IRC | 05:10 | |
| *** achanda has joined #openstack-ironic | 05:11 | |
| *** achanda_ has joined #openstack-ironic | 05:15 | |
| *** achanda_ has quit IRC | 05:16 | |
| *** Haomeng|2 has joined #openstack-ironic | 05:16 | |
| *** Masahiro has joined #openstack-ironic | 05:16 | |
| *** achanda has quit IRC | 05:16 | |
| *** achanda_ has joined #openstack-ironic | 05:16 | |
| *** Haomeng has quit IRC | 05:16 | |
| *** achanda_ has quit IRC | 05:21 | |
| pensu | Hi all, So I have been looking into Ironic code, could anyone tell me where exactly DHCP transfer images, I mean the time when DHCP request comes and pxelinux.0 is transferred? | 05:25 |
| pensu | Figured it would be in neutron.agent, but couldn't find it.. | 05:26 |
| *** Masahiro has quit IRC | 05:26 | |
| *** Masahiro has joined #openstack-ironic | 05:32 | |
| *** killer_prince is now known as lazy_prince | 05:35 | |
| lintan | pensu: ironic update the dhcp options to tell Neutron the pxelinux.0 will be transfer, check ironic/dhcp/neutron.py update_dhcp_opts() and ironic/drivers/modules/pxe.py deploy() | 05:44 |
| *** Marga_ has quit IRC | 05:49 | |
| *** Masahiro has quit IRC | 05:52 | |
| *** Masahiro has joined #openstack-ironic | 05:55 | |
| *** Marga_ has joined #openstack-ironic | 06:00 | |
| *** Nisha has quit IRC | 06:01 | |
| pensu | lintan: that's very helpful, thank you...:) | 06:04 |
| *** Masahiro has quit IRC | 06:05 | |
| *** Masahiro has joined #openstack-ironic | 06:05 | |
| lintan | pensu: :) | 06:06 |
| *** Nisha has joined #openstack-ironic | 06:08 | |
| *** harlowja_ is now known as harlowja_away | 06:19 | |
| *** mrda is now known as mrda-away | 06:29 | |
| *** Marga_ has quit IRC | 06:31 | |
| *** ryanpetrello has quit IRC | 06:33 | |
| *** subscope has joined #openstack-ironic | 06:47 | |
| *** k4n0 has joined #openstack-ironic | 06:54 | |
| *** rakesh_hs has joined #openstack-ironic | 06:55 | |
| *** yuanying has joined #openstack-ironic | 07:01 | |
| *** yuanying_ has quit IRC | 07:04 | |
| *** yuanying has quit IRC | 07:09 | |
| *** yuanying has joined #openstack-ironic | 07:13 | |
| *** Masahiro has quit IRC | 07:38 | |
| *** Masahiro has joined #openstack-ironic | 07:49 | |
| *** dlpartain has joined #openstack-ironic | 07:49 | |
| *** ujuc has joined #openstack-ironic | 07:53 | |
| *** rameshg87 has joined #openstack-ironic | 07:54 | |
| *** pradipta_away is now known as pradipta | 08:05 | |
| *** andreykurilin_ has joined #openstack-ironic | 08:09 | |
| *** dtantsur|afk is now known as dtantsur | 08:13 | |
| dtantsur | Morning! | 08:13 |
| *** andreykurilin_ has quit IRC | 08:14 | |
| *** andreykurilin_ has joined #openstack-ironic | 08:15 | |
| *** yuriyz has quit IRC | 08:19 | |
| *** yuriyz has joined #openstack-ironic | 08:26 | |
| rameshg87 | dtantsur, morning :) | 08:28 |
| rameshg87 | dtantsur, i have some questions on https://review.openstack.org/#/c/137933/1 | 08:28 |
| *** jcoufal has joined #openstack-ironic | 08:32 | |
| *** ChuckC has quit IRC | 08:38 | |
| *** pradipta is now known as pradipta_away | 08:46 | |
| yuriyz | morning Ironic | 08:51 |
| *** andreykurilin_ has quit IRC | 08:55 | |
| *** jistr has joined #openstack-ironic | 09:05 | |
| *** ujuc has quit IRC | 09:06 | |
| *** tatyana has joined #openstack-ironic | 09:10 | |
| *** romcheg has joined #openstack-ironic | 09:10 | |
| *** athomas has joined #openstack-ironic | 09:12 | |
| *** chenglch|3 has quit IRC | 09:15 | |
| *** ChuckC has joined #openstack-ironic | 09:16 | |
| dtantsur | rameshg87, oh sorry, missed your message! | 09:16 |
| dtantsur | shoot | 09:17 |
| rameshg87 | dtantsur, hi | 09:17 |
| dtantsur | yuriyz, rameshg87, morning :) | 09:17 |
| rameshg87 | dtantsur, regarding https://review.openstack.org/#/c/137933/1/specs/backlog/vmedia_deploy_vm.rst | 09:17 |
| *** jcoufal has quit IRC | 09:18 | |
| rameshg87 | dtantsur, do you think shared filesystem is a bad idea indeed ? :) | 09:18 |
| rameshg87 | dtantsur, it is actually *not shared* between conductors | 09:18 |
| *** jcoufal has joined #openstack-ironic | 09:18 | |
| rameshg87 | dtantsur, i mean not for sharing file between conductors | 09:18 |
| rameshg87 | dtantsur, the shared filesystem was just for simplifying things so that there is a common way of providing files by conductors to vm hosts | 09:19 |
| dtantsur | rameshg87, at least it should be documented, how it's going to work. | 09:19 |
| dtantsur | I lack understanding of the whole procedure: what is done by a user manually, what is done via API, how it is all done etc | 09:20 |
| rameshg87 | dtantsur, yeah, so you mean the short-spec should be expanded a little more ? | 09:20 |
| dtantsur | rameshg87, yeah, looks like it's a bit too short :) | 09:20 |
| rameshg87 | dtantsur, okay, let me throw in a new *short* spec, thanks :) | 09:21 |
| dtantsur | cool! | 09:21 |
| dtantsur | devananda, rloo and others, I wonder if we need https://github.com/stackforge/tooz (will be under Oslo soon) for implementing new locking | 09:28 |
| *** igordcard has joined #openstack-ironic | 09:30 | |
| *** lucasagomes has joined #openstack-ironic | 09:31 | |
| *** foexle has joined #openstack-ironic | 09:31 | |
| *** jistr is now known as jistr|trng | 09:35 | |
| rameshg87 | lucasagomes, hi | 09:41 |
| lucasagomes | rameshg87, hi there | 09:42 |
| rameshg87 | lucasagomes, i have a question regarding https://review.openstack.org/#/c/138729/3/specs/kilo/root-device-hints.rst | 09:42 |
| lucasagomes | rameshg87, sure | 09:42 |
| rameshg87 | lucasagomes, problem statement "This22 | 09:42 |
| rameshg87 | may result in devices like /dev/sda and /dev/sdb switching around on23 | 09:42 |
| rameshg87 | each boot and Ironic picking different disk every time the machine is24 | 09:42 |
| rameshg87 | being deployed." | 09:42 |
| rameshg87 | lucasagomes, why is ironic picking different disk each time a problem ? | 09:42 |
| rameshg87 | lucasagomes, as long as deploy goes through fine and user's demands are satisfied :) | 09:43 |
| rameshg87 | lucasagomes, i thought the actual problem was ironic might pick up a disk which is not of enough size | 09:43 |
| lucasagomes | rameshg87, it's a problem for some people that care about it, some people may have servers with idk a cheap ssd and a big hd and they may want to be able to select the hd all the time | 09:44 |
| lucasagomes | rameshg87, plus the RAID example | 09:44 |
| lucasagomes | if you build a RAID to be ur root device, you may want to guarantee that Ironic always pick that | 09:44 |
| rameshg87 | lucasagomes, oh okay, shouldn't the problem statement be that then ? :) | 09:44 |
| lucasagomes | and other things, if you deploy a full disk image twice on the same server and each time it picks a disk | 09:44 |
| lucasagomes | now you got 2 disks with a valid bootloader | 09:44 |
| rameshg87 | lucasagomes, ah okay .. that's a problem | 09:45 |
| lucasagomes | rameshg87, sure I can add some use cases :) | 09:45 |
| rameshg87 | lucasagomes, may be these things we can add in there | 09:45 |
| rameshg87 | lucasagomes, yeah :) | 09:45 |
| lucasagomes | +1 will do | 09:45 |
| dtantsur | lucasagomes, o/ can I use you today for a couple more discoverd reviews? :) | 09:46 |
| lucasagomes | dtantsur, heh sure but I need to catch up with that ^ first | 09:46 |
| lucasagomes | next week is my holidays and I have spent most of my time talking on IRC about state machines and not actually coding anything | 09:47 |
| dtantsur | yeah ack | 09:47 |
| dtantsur | lucasagomes, if you have some time, it's https://review.openstack.org/#/c/137374/ and https://review.openstack.org/#/c/138056/ | 09:47 |
| lucasagomes | +1 | 09:48 |
| *** pelix has joined #openstack-ironic | 09:50 | |
| *** tatyana has left #openstack-ironic | 09:51 | |
| *** dlpartain has left #openstack-ironic | 09:55 | |
| *** viktors|afk is now known as viktors | 10:09 | |
| *** yuanying_ has joined #openstack-ironic | 10:14 | |
| *** Masahiro has quit IRC | 10:14 | |
| *** naohirot has quit IRC | 10:17 | |
| *** yuanying has quit IRC | 10:17 | |
| * dtantsur brb | 10:31 | |
| *** ndipanov_gone is now known as ndipanov | 10:41 | |
| *** sambetts has joined #openstack-ironic | 10:43 | |
| *** derekh has joined #openstack-ironic | 10:55 | |
| *** ramineni has quit IRC | 10:57 | |
| *** smoriya has quit IRC | 11:04 | |
| *** rameshg87 has quit IRC | 11:06 | |
| *** pensu has quit IRC | 11:06 | |
| *** athomas has quit IRC | 11:12 | |
| *** Masahiro has joined #openstack-ironic | 11:15 | |
| *** naohirot has joined #openstack-ironic | 11:18 | |
| *** Masahiro has quit IRC | 11:19 | |
| *** ppalucki has joined #openstack-ironic | 11:30 | |
| *** ppalucki has quit IRC | 11:31 | |
| *** ppalucki has joined #openstack-ironic | 11:32 | |
| *** ppalucki has quit IRC | 11:35 | |
| *** ppalucki has joined #openstack-ironic | 11:36 | |
| *** ppalucki_ has joined #openstack-ironic | 11:36 | |
| *** ppalucki has quit IRC | 11:36 | |
| *** ppalucki_ has quit IRC | 11:36 | |
| *** viktors has quit IRC | 11:37 | |
| *** ppalucki has joined #openstack-ironic | 11:37 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic-specs: Root device hints https://review.openstack.org/138729 | 11:38 |
| *** jcoufal_ has joined #openstack-ironic | 11:59 | |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic-specs: Root device hints https://review.openstack.org/138729 | 12:01 |
| *** yjiang5 has quit IRC | 12:01 | |
| *** jcoufal has quit IRC | 12:02 | |
| *** yjiang5 has joined #openstack-ironic | 12:03 | |
| *** igordcard has quit IRC | 12:08 | |
| *** pensu has joined #openstack-ironic | 12:22 | |
| *** ppalucki has quit IRC | 12:38 | |
| *** pensu has quit IRC | 12:43 | |
| *** dprince has joined #openstack-ironic | 13:02 | |
| lucasagomes | dtantsur, reviewed | 13:07 |
| dtantsur | aha thanks, will update soon | 13:07 |
| lucasagomes | :) ack | 13:07 |
| *** erwan_taf has joined #openstack-ironic | 13:08 | |
| dtantsur | lucasagomes, not returning just a node, because it's not extensible. if we're going to return some instructions to the ramdisk, we'll have to put them inside node object, which is weird | 13:08 |
| lucasagomes | dtantsur, right, gotcha. That's why I asked cause idk it looked weird to have a single key in a dict | 13:09 |
| lucasagomes | with a single node as value | 13:09 |
| dtantsur | for now - yes :) | 13:09 |
| lucasagomes | yup that's all good | 13:09 |
| lucasagomes | plumbing work yay | 13:09 |
| dtantsur | ok lemme split the patch now | 13:09 |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Move code updating Node.properties to a plugin https://review.openstack.org/138056 | 13:13 |
| dtantsur | lucasagomes, updated example.conf ^^^ | 13:13 |
| * lucasagomes looks | 13:13 | |
| dtantsur | and yes, now discoverd patches spam into this channel \o/ | 13:13 |
| erwan_taf | hello world | 13:13 |
| dtantsur | erwan_taf, o/ | 13:13 |
| lucasagomes | erwan_taf, yo | 13:14 |
| erwan_taf | dtantsur: _o/ | 13:14 |
| erwan_taf | lucasagomes: _o/ | 13:14 |
| *** ppalucki has joined #openstack-ironic | 13:17 | |
| *** Masahiro has joined #openstack-ironic | 13:17 | |
| *** Haomeng has joined #openstack-ironic | 13:19 | |
| *** Haomeng|2 has quit IRC | 13:20 | |
| *** Masahiro has quit IRC | 13:22 | |
| *** athomas has joined #openstack-ironic | 13:24 | |
| *** pensu has joined #openstack-ironic | 13:25 | |
| *** nosnos has quit IRC | 13:27 | |
| *** pensu has quit IRC | 13:43 | |
| openstackgerrit | Merged stackforge/ironic-discoverd: Move code updating Node.properties to a plugin https://review.openstack.org/138056 | 13:46 |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Return serialized node to the ramdisk https://review.openstack.org/137374 | 13:50 |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Add option `power_off_after_discovery` https://review.openstack.org/139054 | 13:50 |
| dtantsur | lucasagomes, ^^^ | 13:50 |
| dtantsur | `git add -p` rules! :) | 13:50 |
| *** dlaube has joined #openstack-ironic | 13:52 | |
| *** shardy has joined #openstack-ironic | 13:53 | |
| shardy | Hey all, other than the extra_specs for the baremetal flavor, which specify the deploy kernel and ramdisk, do I need to do anything to the image I specify when booting a node via nova? | 13:54 |
| shardy | I'm getting Image foo is missing the following properties: kernel_id, ramdisk_id (HTTP 500) in n-cond logs | 13:55 |
| shardy | Having followed http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#deploying-ironic-with-devstack | 13:55 |
| shardy | I'm probably missing something obvious :) | 13:55 |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Remove 'glance://' prefix strip from image hrefs https://review.openstack.org/139057 | 13:56 |
| *** athomas has quit IRC | 13:58 | |
| vdrok | shardy, seems that your image that you use as image_source does not have kernel_id and ramdisk_id properties | 13:58 |
| openstackgerrit | Tan Lin proposed openstack/ironic-specs: Bare Metal Trust Using Intel TXT https://review.openstack.org/133902 | 13:59 |
| vdrok | shardy, you can check it with glance image-show $your_image_uuid | 14:00 |
| dtantsur | shardy, http://docs.openstack.org/developer/ironic/deploy/install-guide.html#image-requirements may help | 14:00 |
| *** rakesh_hs has quit IRC | 14:01 | |
| shardy | dtantsur, vdrok: aha, thanks - I was missing the --property arguments to image-create, thanks! | 14:02 |
| *** rakesh_hs has joined #openstack-ironic | 14:08 | |
| *** rloo has joined #openstack-ironic | 14:12 | |
| *** jistr|trng has quit IRC | 14:13 | |
| *** athomas has joined #openstack-ironic | 14:15 | |
| openstackgerrit | Yuriy Zveryanskyy proposed openstack/ironic-specs: Add a new driver for Fuel Agent https://review.openstack.org/138115 | 14:19 |
| *** jistr has joined #openstack-ironic | 14:22 | |
| NobodyCam | good morning Ironic | 14:22 |
| *** jistr is now known as jistr|trng | 14:22 | |
| sambetts | morning NobodyCam | 14:22 |
| NobodyCam | morning sambetts | 14:22 |
| NobodyCam | :) | 14:22 |
| lucasagomes | NobodyCam, sambetts morning | 14:23 |
| NobodyCam | morning lucasagomes :) | 14:23 |
| rloo | morning NobodyCam. Aft sambetts, lucasagomes | 14:26 |
| NobodyCam | morning rloo :) | 14:26 |
| erwan_taf | lo NobodyCam | 14:27 |
| erwan_taf | heya rloo | 14:27 |
| erwan_taf | dtantsur: git add -p is the way to go ;) | 14:27 |
| erwan_taf | dtantsur: glad to see you use it | 14:27 |
| lucasagomes | rloo, hello there! | 14:28 |
| dtantsur | :) | 14:28 |
| NobodyCam | good morning erwan_taf | 14:28 |
| dtantsur | NobodyCam, sambetts, rloo _o/ | 14:28 |
| NobodyCam | morning dtantsur | 14:28 |
| erwan_taf | dtantsur: {checkout|reset} -p are nice too | 14:28 |
| dtantsur | oh have to try next time | 14:28 |
| erwan_taf | checkout -p is nice when you fuck up part of your local changes | 14:29 |
| *** ppalucki has quit IRC | 14:29 | |
| erwan_taf | reset -p is nice when you staged unexpected parts with add -p | 14:29 |
| *** jogo has left #openstack-ironic | 14:29 | |
| dtantsur | the latter happens to me :) | 14:30 |
| rloo | hi erwan_taf ;) | 14:31 |
| rloo | hi dtantsur! | 14:31 |
| *** k4n0 has quit IRC | 14:33 | |
| *** ppalucki has joined #openstack-ironic | 14:33 | |
| *** lazy_prince is now known as killer_prince | 14:34 | |
| * lucasagomes just watched a video about brazilians, can't stop laughing | 14:43 | |
| lucasagomes | https://www.youtube.com/watch?v=uru-_Zt7UJc | 14:43 |
| *** Marga_ has joined #openstack-ironic | 14:45 | |
| *** ryanpetrello has joined #openstack-ironic | 14:45 | |
| NobodyCam | omg | 14:48 |
| NobodyCam | lol | 14:48 |
| lucasagomes | hah the sepultura music on the background and all | 14:49 |
| lucasagomes | damn so funny | 14:49 |
| *** jcoufal_ is now known as jcoufal | 14:49 | |
| NobodyCam | yea | 14:49 |
| NobodyCam | cute | 14:49 |
| NobodyCam | :) | 14:49 |
| *** mjturek has joined #openstack-ironic | 14:50 | |
| victor_lowther | morning all | 14:53 |
| victor_lowther | ooh, lots of comments on the state machine spec | 14:53 |
| lucasagomes | victor_lowther, morning | 14:54 |
| dtantsur | victor_lowther, morning :) as usual, I guess | 14:55 |
| dtantsur | lucasagomes, please have a look at https://review.openstack.org/#/c/139054/ and https://review.openstack.org/#/c/137374/ once you have some time | 14:55 |
| victor_lowther | well, sort of a foggy/rainy morning | 14:55 |
| NobodyCam | mornign victor_lowther | 14:55 |
| victor_lowther | so a little unusual there. :) | 14:55 |
| lucasagomes | dtantsur, will do | 14:56 |
| dtantsur | thnx | 14:56 |
| NobodyCam | victor_lowther: look at all the +1's | 14:57 |
| NobodyCam | :) | 14:57 |
| victor_lowther | yeah, I saw those. | 14:57 |
| victor_lowther | Hoorays. | 14:57 |
| victor_lowther | I will update the spec to fix the nits and add the clarifications people have pointed out. | 14:58 |
| dtantsur | lucasagomes, lol awesome video | 14:59 |
| lucasagomes | :D | 14:59 |
| erwan_taf | heya victor_lowther | 14:59 |
| victor_lowther | hi | 14:59 |
| * victor_lowther works on finishing the first cup of coffee first | 15:00 | |
| *** alexpilotti has joined #openstack-ironic | 15:00 | |
| *** luisjariz has joined #openstack-ironic | 15:01 | |
| NobodyCam | victor_lowther: ++++++ | 15:02 |
| erwan_taf | do we have a state machine for that ? | 15:04 |
| * erwan_taf hides | 15:04 | |
| erwan_taf | & run | 15:04 |
| NobodyCam | lol | 15:04 |
| *** andreykurilin_ has joined #openstack-ironic | 15:05 | |
| *** Masahiro has joined #openstack-ironic | 15:06 | |
| victor_lowther | 1: drink coffee -> 2: do work 3: goto 1 | 15:06 |
| victor_lowther | because circular graphs are the bestest graphs! | 15:06 |
| *** igordcard has joined #openstack-ironic | 15:10 | |
| *** Masahiro has quit IRC | 15:11 | |
| erwan_taf | while(1): { drink_coffee(); sleep($RANDOM) } | 15:11 |
| NobodyCam | 10 for each fullCup: eval $drink: goto 10 | 15:13 |
| *** rakesh_hs has quit IRC | 15:15 | |
| openstackgerrit | Merged stackforge/ironic-discoverd: Add option `power_off_after_discovery` https://review.openstack.org/139054 | 15:16 |
| devananda | morning, all | 15:16 |
| *** Marga_ has quit IRC | 15:16 | |
| NobodyCam | good morning devananda | 15:16 |
| *** athomas has quit IRC | 15:16 | |
| dtantsur | devananda, o/ | 15:17 |
| *** Marga_ has joined #openstack-ironic | 15:17 | |
| lucasagomes | devananda, morning | 15:17 |
| *** r-daneel has joined #openstack-ironic | 15:18 | |
| *** athomas has joined #openstack-ironic | 15:21 | |
| *** romcheg has quit IRC | 15:24 | |
| *** romcheg has joined #openstack-ironic | 15:24 | |
| *** romcheg has quit IRC | 15:25 | |
| NobodyCam | lucasagomes: you test with ipxe yes? | 15:26 |
| lucasagomes | NobodyCam, sometimes yes | 15:26 |
| NobodyCam | I was just looking at rloo's comments on: https://review.openstack.org/#/c/138864 | 15:27 |
| NobodyCam | do you know if ipxe needs the tftp-map file too? | 15:27 |
| *** subscope has quit IRC | 15:28 | |
| *** derekh has quit IRC | 15:28 | |
| lucasagomes | NobodyCam, hmm not sure | 15:29 |
| lucasagomes | gotta investigate | 15:30 |
| NobodyCam | :) | 15:30 |
| NobodyCam | TY | 15:30 |
| lucasagomes | NobodyCam, we still using tftp to pass the undionly.kpxe so I bet it needs the map yes | 15:31 |
| PaulCzar | NobodyCam: I'm super close to having this thing working now | 15:31 |
| dlaube | g'morning | 15:31 |
| PaulCzar | NobodyCam: pxebooting is working, but it boots up to a 'sda: unknown partition table' | 15:32 |
| PaulCzar | which means there's probably an issue with the image | 15:32 |
| *** luisjariz has quit IRC | 15:32 | |
| PaulCzar | is there a stash of prebuilt image sets somewhere ? | 15:33 |
| jroll | morning everybody :) | 15:33 |
| cohn | hi | 15:33 |
| jroll | NobodyCam: I think lucas is right, probably do need the map | 15:33 |
| jroll | PaulCzar: there's a cirros image for testing, dunno about the rest | 15:34 |
| dtantsur | PaulCzar, I know about http://file.rdu.redhat.com/~jslagle/tripleo-images-juno-source/ (used for instack) | 15:34 |
| jroll | PaulCzar: afaik people just build through tripleo | 15:34 |
| dtantsur | but cirros should be much easier | 15:34 |
| jroll | err, s/tripleo/diskimage-builder | 15:34 |
| NobodyCam | PaulCzar: great :) | 15:36 |
| NobodyCam | hi cohn :) | 15:36 |
| NobodyCam | lucasagomes: jroll: Thank you | 15:37 |
| PaulCzar | oh, I have the wrong images listed as the deploy images | 15:37 |
| PaulCzar | that should be an easy fix! | 15:37 |
| cohn | NobodyCam: howdy! | 15:38 |
| *** rameshg87 has joined #openstack-ironic | 15:38 | |
| *** andreykurilin_ has quit IRC | 15:39 | |
| *** andreykurilin_ has joined #openstack-ironic | 15:39 | |
| *** subscope has joined #openstack-ironic | 15:43 | |
| NobodyCam | :) | 15:44 |
| lucasagomes | jroll, morning | 15:44 |
| jroll | hey lucas :) | 15:45 |
| *** pcrews has joined #openstack-ironic | 15:45 | |
| *** naohirot has quit IRC | 15:46 | |
| *** Marga_ has quit IRC | 15:47 | |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Extend node_cache.pop_node() result to be a structure https://review.openstack.org/139096 | 15:48 |
| lucasagomes | I will call it a day earlier today (have a pre-xmas dinner tonight) | 15:48 |
| lucasagomes | have a good day/night everyone! | 15:49 |
| dtantsur | pre-xmas dinner already? nice :) | 15:49 |
| dtantsur | lucasagomes, g'night | 15:49 |
| openstackgerrit | Lucas Alvares Gomes proposed openstack/ironic: Add tests to iscsi_deploy.build_deploy_ramdisk_options https://review.openstack.org/139097 | 15:49 |
| NobodyCam | have a good night lucasagomes | 15:49 |
| rloo | bye lucasagomes. Have fun! | 15:49 |
| sambetts | night lucasagomes | 15:49 |
| lucasagomes | dtantsur, yeah because I'm going to brazil next week | 15:49 |
| lucasagomes | so doing a dinner with some people in the dublin office | 15:49 |
| erwan_taf | see ya lucasagomes | 15:50 |
| *** lucasagomes is now known as lucas-dinner | 15:50 | |
| PaulCzar | alright, so I got the images in the right order ... now having issues with the build ... its trying to look up the ironic api and failing because of dns | 15:50 |
| *** mjturek has quit IRC | 15:51 | |
| *** mjturek has joined #openstack-ironic | 15:51 | |
| PaulCzar | I have dns_nameservers set in my neutron subnet ... which should be enough for dhcp to give that to the ironic build via dhcp right ? | 15:51 |
| *** lucas-dinner has quit IRC | 15:52 | |
| *** rameshg87 has quit IRC | 15:56 | |
| *** hemna__ has joined #openstack-ironic | 15:56 | |
| NobodyCam | I would think so. | 15:57 |
| cohn | hey, so question about the rackerlabs ironic neutron plugin (https://github.com/rackerlabs/ironic-neutron-plugin): what is the process flow for spinning up a bare metal server with it? I'm assuming you can't just use Horizon since you guys have API extensions in there. | 16:00 |
| *** Marga_ has joined #openstack-ironic | 16:00 | |
| jroll | cohn: it doesn't really have anything to do with horizon vs api vs whatever | 16:01 |
| cohn | ok | 16:01 |
| jroll | but code needs to be added to ironic to use it | 16:01 |
| cohn | oh! | 16:01 |
| jroll | which we have yet to open source, because priorities, but I can dig it up for you | 16:01 |
| cohn | so with the code added, it can "just work" via CLI commands or Horizon? | 16:02 |
| victor_lowther | Any suggestions on what to call the "don't power off the machine when cleaning is finished" flag? | 16:02 |
| victor_lowther | for the next rev of the state machine spec | 16:02 |
| *** ppalucki has quit IRC | 16:02 | |
| cohn | jroll: if you could dig that up i would be incredibly appreciative | 16:02 |
| jroll | victor_lowther: that will be a config option, I wouldn't worry about it for that spec, I'm working a separate spec for that | 16:02 |
| jroll | cohn: yep, give me a few and let me get back to you | 16:02 |
| victor_lowther | ya | 16:02 |
| cohn | woot! | 16:02 |
| * cohn hugz jroll | 16:03 | |
| jroll | :) | 16:03 |
| victor_lowther | jroll: I am just going through fleshing out state descriptions and adding links to other specs where appropriate | 16:03 |
| victor_lowther | and it seemed like a good thing to do | 16:03 |
| jroll | aha | 16:03 |
| jroll | sec | 16:03 |
| jroll | I'm really behind on it, but here you go: https://review.openstack.org/#/c/102405/ | 16:03 |
| victor_lowther | thanks | 16:05 |
| *** Marga_ has quit IRC | 16:07 | |
| *** romcheg has joined #openstack-ironic | 16:09 | |
| openstackgerrit | Chris Krelle proposed openstack/ironic: Add info on creating a tftp map file https://review.openstack.org/138864 | 16:10 |
| *** subscope has quit IRC | 16:11 | |
| *** Isotopp_ has joined #openstack-ironic | 16:15 | |
| Shrews | NobodyCam: if i were to suggest a change to 138864 that was not actually your fault, but you DO touch it, would you be offended? :) | 16:16 |
| Shrews | NobodyCam: ex) explicitly numbering the sections instead of using #. like the previous sections | 16:16 |
| victor_lowther | hm | 16:16 |
| rloo | Shrews: you mean not to explicitly number... (++) | 16:17 |
| Shrews | rloo: right | 16:17 |
| rloo | Shrews: although I wonder if it should be changed in the same patch. | 16:17 |
| *** bradjones has quit IRC | 16:18 | |
| victor_lowther | looking at the comments, how does adding an ACTIVE -> [DELET*/MANAGED] -> MANAGED path to the state machine sound? | 16:18 |
| *** Isotopp has quit IRC | 16:18 | |
| *** jroll has quit IRC | 16:18 | |
| *** adam_g has quit IRC | 16:18 | |
| *** sbadia has quit IRC | 16:18 | |
| *** jroll has joined #openstack-ironic | 16:18 | |
| *** adam_g has joined #openstack-ironic | 16:19 | |
| jroll | victor_lowther: that sounds like I need to manually set every node to available after it's deleted | 16:19 |
| *** adam_g has quit IRC | 16:19 | |
| *** adam_g has joined #openstack-ironic | 16:19 | |
| *** jroll has quit IRC | 16:19 | |
| *** jroll has joined #openstack-ironic | 16:19 | |
| jroll | which is a non-starter | 16:20 |
| *** bradjones has joined #openstack-ironic | 16:20 | |
| victor_lowther | for doing things like pulling a machine out of production while bypassing CLEANING | 16:20 |
| *** dguerri has joined #openstack-ironic | 16:20 | |
| NobodyCam | Shrews: lol I tested that change locally and it miss numbered so I just keep the hard coding we had | 16:20 |
| rloo | victor_lowther: that changes a lot of stuff. You mean get rid of [DELET*/AVAILABLE] ? | 16:20 |
| NobodyCam | I'm happy to correct | 16:21 |
| jroll | victor_lowther: oh, so in addition. how would it decide to go that way? I don't see the usefulness | 16:21 |
| Shrews | NobodyCam: hrm, wonder what's off with the automatic numbering | 16:22 |
| victor_lowther | nah | 16:22 |
| victor_lowther | that is not the intent of the path | 16:22 |
| *** arif-ali has quit IRC | 16:23 | |
| victor_lowther | jroll: You need to pull a machine out of production for forensic analysis | 16:23 |
| victor_lowther | RESCUE is out, because you to not intend to return it to its previous role. | 16:23 |
| JayF | makes sense to have a way to do that | 16:24 |
| JayF | however going to be VERY HARD to implement consistently | 16:24 |
| victor_lowther | CLEANING is of course right out. | 16:24 |
| JayF | because if it's a nova provisioned instance | 16:24 |
| *** sbadia has joined #openstack-ironic | 16:24 | |
| JayF | what are you going to do? Not delete it in nova and instead delete it via ironic? | 16:24 |
| rloo | can't you just put it in maintenance mode first. then do whatever? | 16:24 |
| jroll | theoretically. | 16:24 |
| jroll | though I think you should be able to CLEAN a node in maintenance mode | 16:25 |
| JayF | rloo: that's similar to the workflow we'd use in OnMetal today, except we allow decom to complete in maint mode (downstream; aiui upstream we'll not to any damn thing to a maint node) | 16:25 |
| victor_lowther | Think lawsuit and/or lat subpoena | 16:25 |
| rloo | when would you know/indicate to ironic that you want to pull a machine out of production...? | 16:25 |
| JayF | yeah, that's my issue with it | 16:25 |
| jroll | victor_lowther: sure, I understand, it's just ^ | 16:25 |
| *** penick has joined #openstack-ironic | 16:25 | |
| *** romcheg has quit IRC | 16:25 | |
| victor_lowther | yeah, I don't really know how to disambiguate from the Nova side. | 16:26 |
| victor_lowther | is why I threw it out here. :) | 16:26 |
| *** arif-ali has joined #openstack-ironic | 16:27 | |
| rloo | victor_lowther: so you want to DELETE and after that put it in limbo? (or MANAGED) | 16:27 |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Support for non-Glance image references https://review.openstack.org/136741 | 16:27 |
| *** subscope has joined #openstack-ironic | 16:28 | |
| *** mikedillion has joined #openstack-ironic | 16:28 | |
| victor_lowther | rloo: ACTIVE -> [DELET*/MANAGED] -> MANAGED | 16:28 |
| victor_lowther | so MANAGED | 16:28 |
| victor_lowther | Not adding a LIMBO state. :) | 16:28 |
| NobodyCam | Shrews: I can spend more then 1 minute looking at it :) | 16:28 |
| *** Marga_ has joined #openstack-ironic | 16:28 | |
| victor_lowther | I suppose from the POV of Nova the machine just goes offline and never comes back | 16:30 |
| devananda | victor_lowther: on "dont power off machine when done cleaning" -- I think this is something the driver implements. not a flag the operator passes | 16:30 |
| jroll | victor_lowther: also. presumably the ironic operators would be the ones doing forensic analysis, not the nova user. the nova user might want to delete to hide things. can't really make it a nova thing. | 16:30 |
| rloo | devananda: I don't think he wants it to be cleaned, or to be made avail again. | 16:30 |
| victor_lowther | jroll: yes | 16:31 |
| jroll | victor_lowther: though we could use some nova admin things, like lock/stop/suspend/idk, to take the machine back or whatever | 16:31 |
| devananda | victor_lowther: to pull a node out of prod w/o CLEAN, set maintenance flag | 16:31 |
| victor_lowther | is there an expectation that it will ever return? | 16:31 |
| *** Marga_ has quit IRC | 16:31 | |
| jroll | victor_lowther: "maybe" | 16:31 |
| victor_lowther | heh | 16:31 |
| jroll | I mean | 16:32 |
| *** Marga_ has joined #openstack-ironic | 16:32 | |
| devananda | nova user deleting their instance ---> node should return to available (unless it's broken // in maintenance mode) | 16:32 |
| jroll | rackspace does similar things like this for non-payment, take it away for a bit, give it back if they pay | 16:32 |
| *** ryanpetrello has quit IRC | 16:32 | |
| victor_lowther | devananda: that text I had in mind for CLEANING and long-running deploy ramdisks goes along these lines: | 16:33 |
| victor_lowther | * Booting to a `long running deploy ramdisk | 16:33 |
| victor_lowther | <https://review.openstack.org/#/c/102405/>`, if you want the | 16:33 |
| victor_lowther | machine to stay on while in AVAILABLE. | 16:33 |
| victor_lowther | jroll: so it sounds like the maintenance flag encompasses that use case well enough, then. | 16:34 |
| victor_lowther | Ever had to deal with "the cops want the whole system intact"? | 16:35 |
| openstackgerrit | Chris Krelle proposed openstack/ironic: Add info on creating a tftp map file https://review.openstack.org/138864 | 16:35 |
| NobodyCam | Shrews: ^^^ | 16:35 |
| jroll | victor_lowther: we haven't with ironic, no idea about virt | 16:35 |
| devananda | oh, is anyone actively working on improving our policy handling? | 16:35 |
| victor_lowther | ok | 16:35 |
| jroll | devananda: roman had a patch up to pull latest oslo.policy | 16:36 |
| devananda | victor_lowther: right - no. I was just also thinking about that. | 16:36 |
| victor_lowther | I will leave things as they are for now, then | 16:36 |
| dtantsur | devananda, you mean https://review.openstack.org/126265 ? | 16:36 |
| jroll | devananda: other than that, not that I know of | 16:36 |
| victor_lowther | since it is more of an exceptional situation anyways. | 16:36 |
| victor_lowther | Something to thing about later | 16:37 |
| devananda | victor_lowther: maintenance mode is close, but we'd need to "delete" the nova instance, block in-and-outgoing networks, and power off the machine | 16:37 |
| victor_lowther | maybe after it happens to someone. :) | 16:37 |
| devananda | heh | 16:37 |
| victor_lowther | Anyways, back to edits. | 16:37 |
| jroll | devananda: in the "cops want the whole system intact" case, you may not want to power off, even | 16:37 |
| devananda | jroll: sure you do. cops want the system as it is now, not after some daemon finishes scrubbing the disks | 16:38 |
| *** achanda has joined #openstack-ironic | 16:38 | |
| NobodyCam | unless there /my disks | 16:38 |
| devananda | jroll: but anyway. details to discuss later | 16:38 |
| jroll | ok, maybe, never dealt with this :) | 16:39 |
| *** Marga_ has quit IRC | 16:39 | |
| *** Marga_ has joined #openstack-ironic | 16:39 | |
| devananda | dtantsur: that's just keeping it up to date, I think? | 16:39 |
| dtantsur | not sure | 16:39 |
| *** dprince has quit IRC | 16:39 | |
| devananda | dtantsur: I mean adding more configurable policies and privilege levels for different parts of Ironic's API | 16:39 |
| dtantsur | ah, probably nobody | 16:40 |
| *** dprince has joined #openstack-ironic | 16:40 | |
| devananda | dtantsur: like allowing read access to power,provision state and instance_info, but hiding driver_info | 16:40 |
| devananda | for example | 16:40 |
| jroll | or allowing certain folks to do certain actions | 16:40 |
| devananda | right | 16:41 |
| jroll | devananda: do you think network partitioning stuff could fit this cycle? | 16:42 |
| jroll | at least an interface and an implementation for our neutron plugin thing? | 16:42 |
| devananda | jroll: "network partitioning" ? | 16:43 |
| jroll | like, provisioning network vs tenant network | 16:43 |
| jroll | switching between them | 16:43 |
| devananda | ah | 16:44 |
| jroll | priorities are a thing | 16:44 |
| devananda | that'd be great. but yah. | 16:44 |
| jroll | we just get asked about it sometimes | 16:44 |
| devananda | I keep telling people about it, too | 16:44 |
| jroll | and I'm sick of having downstream code | 16:44 |
| *** krtaylor has quit IRC | 16:44 | |
| jroll | :P | 16:44 |
| jroll | do you know about https://etherpad.openstack.org/p/ironic-neutron-bonding | 16:44 |
| devananda | it seems like that would be orthogonal to the other big work going on | 16:44 |
| jroll | I added a gist there of our code that calls our neutron plugin | 16:44 |
| devananda | like it could be approved and reviewed independently | 16:44 |
| jroll | yeah | 16:45 |
| devananda | in which case, i'm all for it | 16:45 |
| jroll | cool | 16:45 |
| devananda | folks may or may not have bandwidth to review it -- but if it's not dependent on // blocking anything else, let's try | 16:45 |
| jroll | yeah | 16:45 |
| jroll | right on | 16:45 |
| jroll | thanks | 16:45 |
| devananda | jroll: the "flip the node between these two networks" part -- that should definitely be upstreamed, even if the ML2 plugin doesn't make it | 16:46 |
| devananda | jroll: because literally everyone wants ^ | 16:47 |
| jroll | right, we can give it an interface and an implementation that uses our plugin | 16:47 |
| devananda | and the port bonding, while cool, can be done other ways, and is not important to everyone, so I'd suggest punting on that if there's a time constraint | 16:47 |
| jroll | and if people care about using something in neutron's tree, they can add it | 16:47 |
| jroll | right | 16:47 |
| jroll | I'm not worried about bonding stuff | 16:48 |
| devananda | k k | 16:48 |
| jroll | though our plugin can trunk things, I'll need to do something with that | 16:48 |
| *** ChanServ sets mode: -o devananda | 16:48 | |
| dtantsur | Oo | 16:48 |
| * devananda drops his @ | 16:48 | |
| NobodyCam | :-p | 16:49 |
| devananda | ChuckC: the above conversation might interest you, too. also -- https://etherpad.openstack.org/p/ironic-neutron-bonding | 16:50 |
| * ChuckC looking | 16:51 | |
| openstackgerrit | Victor Lowther proposed openstack/ironic-specs: New Ironic provisioner state machine. https://review.openstack.org/133828 | 16:51 |
| victor_lowther | With that done, off to sprint planning. | 16:54 |
| *** Masahiro has joined #openstack-ironic | 16:54 | |
| *** penick has quit IRC | 16:56 | |
| openstackgerrit | Chris Krelle proposed openstack/ironic: Add info on creating a tftp map file https://review.openstack.org/138864 | 16:58 |
| *** Marga_ has quit IRC | 16:58 | |
| *** Marga_ has joined #openstack-ironic | 16:59 | |
| *** Masahiro has quit IRC | 16:59 | |
| Shrews | grr... running ironic-api in a docker container is causing me much heartburn | 17:00 |
| *** Marga_ has quit IRC | 17:01 | |
| jroll | Shrews: what's wrong? | 17:02 |
| *** Marga_ has joined #openstack-ironic | 17:02 | |
| Shrews | jroll: http://paste.openstack.org/show/144703/ | 17:04 |
| *** dtantsur is now known as dtantsur|afk | 17:05 | |
| jroll | wtf | 17:05 |
| Shrews | exactly | 17:05 |
| Shrews | i cannot reproduce that outside the container either | 17:05 |
| Shrews | same oslo library versions | 17:06 |
| jroll | strange | 17:06 |
| ChuckC | devananda: jroll: so bonding not so urgent, but 'flipping node between networks' is | 17:09 |
| devananda | ChuckC: yup | 17:10 |
| ChuckC | does that map to updating port to change network? | 17:10 |
| JayF | Shrews: you sure running "ironic-api" is using the same python as "python" | 17:10 |
| ChuckC | devananda: ^^ | 17:10 |
| JayF | Shrews: /me suspects a distro with py27 and py3* and your deps are on py27 and it's running against py3* | 17:10 |
| jroll | ChuckC: agree, and yes, it changes the vlans for the port | 17:10 |
| Shrews | JayF: /usr/local/bin/ironic-api uses /usr/bin/python, which is 2.7 | 17:11 |
| ChuckC | devananda: in the mac address update spec, i recall that yamahata was also interested in being able to update network_id, so there may be some support there | 17:12 |
| JayF | interesting. I don't wanna rabbithole right now, but I'd be interested to see your Dockerfile for poking at later :) | 17:12 |
| JayF | but you'll likely figure it out before I take the time to look | 17:12 |
| ChuckC | devananda: the problem is neutron's upcoming SPD/SAD deadlines for kilo (next week and the following) | 17:14 |
| jroll | ChuckC: yeah, I don't see anything actually getting into neutron to help with this in kilo | 17:14 |
| devananda | ChuckC: right. so I think jroll is hoping to upstream the Ironic components of this during Kilo | 17:14 |
| devananda | jroll: are there aspects that need to land in Neutron, or is your out-of-tree ML2 plugin enough? | 17:15 |
| jroll | devananda: our out of tree plugin should be enough, I don't see why not | 17:15 |
| devananda | if there are no other changes needed in neutron, then definitely, let's go ahead with the Ironic side | 17:15 |
| jroll | we only have a driver for one switch, but other than that it just works | 17:15 |
| jroll | I don't think we've modified neutron to use our plugin | 17:15 |
| jroll | morgabra: ^ confirm? | 17:15 |
| ChuckC | devananda: jroll: let me play with it a bit to scope it and maybe contact yamahata to see what was driving his interest | 17:16 |
| ChuckC | this will not be my top priority though | 17:17 |
| *** krtaylor has joined #openstack-ironic | 17:17 | |
| *** pensu has joined #openstack-ironic | 17:21 | |
| *** achanda has quit IRC | 17:22 | |
| *** foexle has quit IRC | 17:22 | |
| *** achanda has joined #openstack-ironic | 17:22 | |
| *** achanda has quit IRC | 17:27 | |
| *** Marga_ has quit IRC | 17:27 | |
| *** spandhe has joined #openstack-ironic | 17:27 | |
| *** Marga_ has joined #openstack-ironic | 17:27 | |
| ChuckC | devananda: jroll: regarding mac address update (replacing bad NIC), do you see a need for non-admin to be able to do this? | 17:31 |
| jroll | ChuckC: non-admin being the nova user? | 17:32 |
| ChuckC | jroll: not sure, whoever would be updating the neutron port with the new mac address | 17:32 |
| *** marcoemorais has joined #openstack-ironic | 17:33 | |
| NobodyCam | I would tend to say an update like that should require admin | 17:33 |
| jroll | yeah, I would guess an admin would do that | 17:33 |
| jroll | I mean, non-admins shouldn't be replacing nics | 17:33 |
| jroll | :P | 17:33 |
| ChuckC | jroll: ok, thanks, just a sanity check :-) | 17:33 |
| NobodyCam | :) | 17:33 |
| jroll | yep :) | 17:33 |
| *** Marga_ has quit IRC | 17:36 | |
| *** Marga_ has joined #openstack-ironic | 17:36 | |
| *** mikedillion has quit IRC | 17:40 | |
| Shrews | jroll: JayF: think i figured it out. i have an oslo.config in both /usr/lib and /usr/local/lib. The /usr/lib one is older and probably comes from installing python-ironicclient from apt-get | 17:42 |
| jroll | this is why I refuse to use distro packages :) | 17:43 |
| jroll | (one reason, anyway) | 17:43 |
| jroll | (for python) | 17:43 |
| JayF | aha | 17:45 |
| Shrews | well, it would have been fine if both ironicclient AND ironic came from the distro. Mixing the two apparently doesn't work. | 17:48 |
| * Shrews goes to lunch | 17:48 | |
| *** spandhe has quit IRC | 17:49 | |
| *** jistr|trng has quit IRC | 17:51 | |
| *** jcoufal has quit IRC | 17:52 | |
| NobodyCam | brb | 17:53 |
| *** ryanpetrello has joined #openstack-ironic | 17:57 | |
| *** Marga_ has quit IRC | 17:57 | |
| *** Marga_ has joined #openstack-ironic | 17:58 | |
| * NobodyCam is back | 18:03 | |
| *** athomas has quit IRC | 18:08 | |
| *** harlowja_away is now known as harlowja_ | 18:10 | |
| *** sambetts has quit IRC | 18:11 | |
| *** Nisha has quit IRC | 18:17 | |
| *** shakamunyi has joined #openstack-ironic | 18:18 | |
| *** spandhe has joined #openstack-ironic | 18:30 | |
| *** mikedillion has joined #openstack-ironic | 18:33 | |
| *** mjturek has quit IRC | 18:35 | |
| *** jmanko has joined #openstack-ironic | 18:41 | |
| *** mjturek has joined #openstack-ironic | 18:42 | |
| *** Masahiro has joined #openstack-ironic | 18:43 | |
| *** jmank has quit IRC | 18:44 | |
| *** Masahiro has quit IRC | 18:48 | |
| *** openstackgerrit has quit IRC | 18:50 | |
| *** openstackgerrit has joined #openstack-ironic | 18:50 | |
| *** andreykurilin_ has quit IRC | 18:50 | |
| *** Marga_ has quit IRC | 18:52 | |
| *** Marga_ has joined #openstack-ironic | 18:54 | |
| NobodyCam | do we still want the "Whats needed to Land the New State machine" line on the agenda? | 18:55 |
| rloo | NobodyCam: isn't there more that is needed? | 18:56 |
| openstackgerrit | Jarrod Johnson proposed stackforge/pyghmi: Implement server side IPMI protocol (WIP) https://review.openstack.org/138109 | 18:56 |
| devananda | probably | 18:56 |
| *** pelix has quit IRC | 18:59 | |
| openstackgerrit | Jarrod Johnson proposed stackforge/pyghmi: Implement server side IPMI protocol (WIP) https://review.openstack.org/138109 | 19:02 |
| *** marcoemorais has quit IRC | 19:03 | |
| *** marcoemorais has joined #openstack-ironic | 19:04 | |
| *** marcoemorais has quit IRC | 19:04 | |
| *** marcoemorais has joined #openstack-ironic | 19:05 | |
| *** penick has joined #openstack-ironic | 19:12 | |
| *** igordcard has quit IRC | 19:12 | |
| openstackgerrit | Jarrod Johnson proposed stackforge/pyghmi: Implement server side IPMI protocol (WIP) https://review.openstack.org/138109 | 19:13 |
| NobodyCam | going to remove the mascot name stuff from the whiteboard | 19:13 |
| NobodyCam | brb ... quick walkies | 19:22 |
| dlaube | hey guys, so I ended up building a custom image with a modified IPA using https://github.com/openstack/ironic-python-agent/tree/master/imagebuild/coreos …what do I need to to finish making the deploy image? Do I just run glance image-create against my coreos_production_pxe_image-oem.cpio.gz coreos_production_pxe.vmlinuz files? | 19:22 |
| dlaube | I think I'm missing some steps | 19:22 |
| JayF | I think those glance upload steps are in the usual ironic docs? | 19:23 |
| JayF | if you have the .vmlinuz and .cpio.gz you built the ramdisk | 19:23 |
| dlaube | ok cool.. so just this is left: | 19:24 |
| JayF | dlaube: http://docs.openstack.org/developer/ironic/deploy/install-guide.html#image-requirements I think this gets into that some | 19:24 |
| dlaube | glance image-create --name deploy-vmlinuz --public \ | 19:24 |
| dlaube | --disk-format aki < my-deploy-ramdisk.kernel | 19:24 |
| dlaube | glance image-create --name deploy-initrd --public \ | 19:24 |
| dlaube | --disk-format ari < my-deploy-ramdisk.initramfs | 19:24 |
| JayF | yup | 19:24 |
| JayF | I'm pretty sure that's right | 19:24 |
| dlaube | nice! thats what I thought but it seemed too easy :P | 19:24 |
| JayF | things are always supposed to be easy | 19:24 |
| dlaube | seems nothing is ever easy JayF | 19:25 |
| dlaube | you know how much I bother you, jroll and NobodyCam ;) | 19:25 |
| dlaube | heh | 19:25 |
| dlaube | always appreciate the help though | 19:25 |
| dlaube | thanks! | 19:25 |
| *** mjturek has quit IRC | 19:26 | |
| *** mjturek has joined #openstack-ironic | 19:26 | |
| *** penick has quit IRC | 19:26 | |
| *** penick has joined #openstack-ironic | 19:27 | |
| *** dprince has quit IRC | 19:28 | |
| *** marcoemorais has quit IRC | 19:37 | |
| *** marcoemorais has joined #openstack-ironic | 19:37 | |
| *** rushiagr is now known as rushiagr_away | 19:38 | |
| *** marcoemorais has quit IRC | 19:44 | |
| *** marcoemorais has joined #openstack-ironic | 19:44 | |
| * NobodyCam is back | 19:50 | |
| NobodyCam | dlaube: its never a bother to help :) | 19:51 |
| devananda | why does ironic require sgabios? | 19:58 |
| devananda | interesting | 19:59 |
| *** penick has quit IRC | 19:59 | |
| devananda | wonder why devstack wasn't trying to install that, but failing when it wasn't installed ... *shrug* | 19:59 |
| *** andreykurilin_ has joined #openstack-ironic | 20:01 | |
| NobodyCam | devananda: that to get all the bios boot screens from a node | 20:01 |
| *** openstackgerrit has quit IRC | 20:04 | |
| devananda | yup, figured it out :) | 20:04 |
| *** openstackgerrit has joined #openstack-ironic | 20:04 | |
| *** penick has joined #openstack-ironic | 20:04 | |
| devananda | trying to update my local devstack env, and hitting all sorts of issues :( | 20:04 |
| NobodyCam | :( | 20:05 |
| *** hemna__ has quit IRC | 20:11 | |
| *** marcoemorais has quit IRC | 20:12 | |
| *** marcoemorais has joined #openstack-ironic | 20:13 | |
| *** marcoemorais has quit IRC | 20:13 | |
| *** marcoemorais has joined #openstack-ironic | 20:14 | |
| *** pensu has quit IRC | 20:25 | |
| *** mrda-away is now known as mrda | 20:25 | |
| mrda | Morning Ironic | 20:25 |
| soren | I recently noticed that there's a way for Ironic to not depend on Neutron for dhcp... In that case, how will Ironic know the IP of the node? | 20:26 |
| NobodyCam | morning mrda | 20:26 |
| jroll | soren: why does ironic need to know the IP? | 20:27 |
| * NobodyCam is liking https://review.openstack.org/#/c/134866 | 20:27 | |
| jroll | soren: also, in that case you would use static dhcp configs | 20:27 |
| soren | jroll: Doesn't it need it to write the disk image to the node? | 20:27 |
| soren | jroll: What's the benefit of static dhcp configs? | 20:28 |
| soren | Also, how will Nova know the IP? | 20:28 |
| soren | ...which, I suppose, is the real question I was getting at. | 20:28 |
| jroll | soren: I'm not sure how ironic finds iscsi mounts, but don't think it needs the IP. maybe the ramdisk sends it back. with the agent driver, the ramdisk sends its IP to ironic so ironic can talk to it. | 20:29 |
| soren | jroll: Ok. I guess that makes sense. What about Nova? | 20:30 |
| jroll | as far as the instance, that's different. ironic only configures dhcp for pxe booting, has nothing to do with the instance IP | 20:30 |
| jroll | nova is what provisions an IP from neutron anyway, so nova always knows | 20:30 |
| jroll | ironic does not know the instance IP | 20:30 |
| *** shakamunyi has quit IRC | 20:30 | |
| devananda | could be an artefact of my env, but curious, is anyone else seeing this lately? | 20:30 |
| devananda | DEBUG oslo.messaging._drivers.impl_rabbit [-] Timed out waiting for RPC response: timed out from (pid=22077) _raise_timeout_if_deadline_is_reached | 20:30 |
| soren | jroll: So it'll have different IP's during PXE and when it's fully deployed? | 20:31 |
| jroll | and the benefit of static DHCP configs is 1) not using neutron/dnsmasq, and 2) knowing what IP the ramdisk gets is helpful | 20:31 |
| jroll | yes | 20:31 |
| jroll | I believe so | 20:31 |
| soren | But how does that config get in there? | 20:31 |
| NobodyCam | jroll: soren just fyi: ramdisk ping back: https://github.com/openstack/diskimage-builder/blob/master/elements/deploy-ironic/init.d/80-deploy-ironic#L46-L54 | 20:31 |
| jroll | soren: someone configures a dhcp server, it isn't automatic | 20:31 |
| jroll | NobodyCam: ah, right, thanks | 20:31 |
| soren | NobodyCam: Thanks. | 20:32 |
| *** shakamunyi has joined #openstack-ironic | 20:32 | |
| *** Masahiro has joined #openstack-ironic | 20:32 | |
| *** alexpilotti has quit IRC | 20:32 | |
| soren | I'm just confused, I guess. What is the point of all this if you're still relying on Neutron for the IP assignment post install? | 20:32 |
| soren | ...and how does the IP from Neutron get set on the server? | 20:33 |
| jroll | because we don't trust dnsmasq | 20:33 |
| jroll | through the metadata service, configdrive, or dhcp | 20:33 |
| jroll | depending on your environment | 20:33 |
| jroll | our prod environment uses configdrive | 20:33 |
| jroll | with cloud-init | 20:33 |
| soren | And how does that get written? | 20:33 |
| soren | I forget the flow for that.. Does nova build it beforehand and passes it to Ironic? | 20:34 |
| jroll | the upstream parts are still a work in progress, but the code is in ironic, it just makes another partition | 20:34 |
| jroll | yes | 20:34 |
| jroll | see https://review.openstack.org/#/q/branch:master+topic:bp/expose-configdrive,n,z | 20:34 |
| jroll | and the nova spec https://review.openstack.org/#/c/98930/ | 20:34 |
| soren | I guess this might solve half of my problem.. | 20:34 |
| jroll | still need nova code | 20:34 |
| soren | So, I'm setting up a new environment. | 20:34 |
| jroll | what's your problem? what's the goal? | 20:35 |
| soren | It'll be pretty big, and I'm having to set up a bunch of subnets. | 20:35 |
| soren | So each rack will have its own subnet. | 20:35 |
| soren | Or something to that effect. | 20:35 |
| soren | ...and I'm curious how I can get that done with Ironic/Neutron/friends. | 20:35 |
| soren | So depending on the rack, I need to give it an IP from a specific subnet. | 20:36 |
| jroll | aha | 20:36 |
| soren | dhcp-relay is the typical solution, but this is hard with OpenStack. | 20:36 |
| *** Masahiro has quit IRC | 20:37 | |
| jroll | right, I don't know much about neutron's dhcp stuff, but you might be able to make it work like that | 20:37 |
| devananda | SpamapS: ^ | 20:37 |
| soren | ...because dhcp-relay acts on the DHCP request and assigns an IP from a subnet matching the relay. | 20:37 |
| jroll | like, neutron should have a way to tie mac addresses to subnets or something | 20:37 |
| soren | Yeah. | 20:37 |
| devananda | soren: I believe others share this prob^D^D^Dneed, and I'm not yet aware that anyone has addressed it | 20:37 |
| soren | Even specific IP's would be ok. | 20:37 |
| soren | So a given mac would always get the same IP. | 20:38 |
| jroll | that's called static dhcp configuration :P | 20:38 |
| devananda | because you need more than just giving each node right right IP based on what rack // subnet it's in | 20:38 |
| soren | but which instance is running on that node (with that mac) might change over time. | 20:38 |
| devananda | you also need neutron to dynamically do the right configuration for that | 20:38 |
| jroll | that said, if you use static configs, ironic can't control pxe options either, so idk | 20:38 |
| soren | jroll: That's fine for Ironic/PXE, but I need this post install, too. | 20:38 |
| jroll | right | 20:39 |
| *** hemna__ has joined #openstack-ironic | 20:39 | |
| devananda | and you need each node to be managed only by Ironic Conductor(s) which have access to that subnet | 20:39 |
| soren | How do you do it at Rackspace? Just one big, flat network? | 20:39 |
| devananda | though taht last point may depend on how you set up the networking, so may be not ... | 20:39 |
| jroll | soren: for management or tenant networks? | 20:39 |
| soren | jroll: Hm. Both, I suppose. | 20:39 |
| * soren doesn't have to deal with multitenancy | 20:40 | |
| jroll | so every tenant gets two networks, one is inter-dc and one is public internet | 20:40 |
| jroll | we're still working on private networks | 20:40 |
| jroll | s/tenant/instance | 20:40 |
| *** ChuckC has quit IRC | 20:40 | |
| soren | Ok. | 20:40 |
| soren | So just like (the virtual) cloud servers. | 20:40 |
| devananda | soren: if you trust all tenants to have access to each other's data, and to the control plane, you probably don't need to isolate mgmt <-> provisioning networks | 20:41 |
| jroll | yeah, except virt cloud has private networks too | 20:41 |
| soren | Ok. | 20:41 |
| jroll | for management stuff, our ironic cell has one provisioning network that the servers sit on | 20:41 |
| soren | So for public internet access, you just have one huge, flat network? | 20:41 |
| jroll | something like that | 20:41 |
| jroll | yes :| | 20:41 |
| soren | How many racks/nodes? | 20:42 |
| jroll | (as I understand it, I'm not a networking expert) | 20:42 |
| soren | It makes sense. | 20:42 |
| jroll | for the public network, or with ironic? | 20:42 |
| soren | This is what I wanted to do, too. We just had some Cisco guys around and they damn near fainted, when I told them. | 20:42 |
| *** dprince has joined #openstack-ironic | 20:42 | |
| jroll | maybe they dealt with us ;) | 20:42 |
| clif_h | https://bugs.launchpad.net/ironic/+bug/1367917 this bug has been assigned for awhile but it seems like there's been no activity, does anyone mind if I take it on? | 20:43 |
| jroll | clif_h: do it | 20:43 |
| JayF | +1 | 20:43 |
| soren | jroll: Unlikely. This is in India. | 20:43 |
| * jroll would love to refactor all the config stuff | 20:43 | |
| jroll | soren: that was a joke :P | 20:43 |
| JayF | soren: FWIW we do have special configs on our network to isolate tenants | 20:43 |
| JayF | soren: so it is a "flat network" but protections in place against arp poisoning and similar things | 20:44 |
| clif_h | how do I get it reassigned to me? | 20:44 |
| soren | JayF: VLANs or something? | 20:44 |
| JayF | soren: something | 20:44 |
| JayF | soren: ;) | 20:44 |
| soren | JayF: Does Neutron control that somehow? Or where did you choose to implement that? | 20:44 |
| JayF | all the protections are in switch config/firmware | 20:45 |
| JayF | but our neturon plugin is open source and you're welcome to read it | 20:45 |
| jroll | clif_h: pushing a commit would do it, or click the pencil next to the "assigned to" name | 20:45 |
| soren | Sure, but something needs to push that config to the switch. | 20:45 |
| JayF | soren: https://github.com/rackerlabs/ironic-neutron-plugin | 20:45 |
| jroll | specifically https://github.com/rackerlabs/ironic-neutron-plugin/blob/master/ironic_neutron_plugin/drivers/cisco/commands.py | 20:45 |
| clif_h | jroll: ah, that's easy | 20:45 |
| jroll | :) | 20:45 |
| *** alexpilotti has joined #openstack-ironic | 20:48 | |
| soren | JayF: Interesting, thanks. | 20:48 |
| soren | jroll, JayF: Do you have any horror stories of problems that arose due to such a big, flat network? | 20:48 |
| soren | I've not had the "pleasure" of having one before. Everywhere I've been, there have been limited broadcast domains (per rack or per few racks or something), but everyone just poops themselves when I suggest having thousands of nodes in a big, flat network. | 20:50 |
| JayF | I don't think we intend to scale out a single cell of ironic to multiple-thousands of instances | 20:51 |
| JayF | which simplifies matters greatly | 20:51 |
| jroll | yay for cells. | 20:51 |
| soren | I guess spanning tree related problems are most likely to make everything go pear shaped. | 20:51 |
| JayF | plus same VLAN != same subnet | 20:51 |
| JayF | we have our subnets broken up by /24 in neutron and routed as /24s instead of the larger contiguous grouping, partially to decrease the broadcase domain | 20:52 |
| soren | How did you manage that, then? | 20:52 |
| NobodyCam | yay for hours spent tracking down spanning tree issues :-p | 20:52 |
| JayF | soren: Not sure what you're asking ? | 20:53 |
| soren | Me neither. | 20:53 |
| JayF | Heh | 20:53 |
| soren | So you have subnets broken into /24s. That corresponds to N racks, I imagine? | 20:54 |
| JayF | No | 20:54 |
| soren | Ok, so how are you limiting your broadcast domains? | 20:54 |
| * soren seems to be missing a piece of the puzzle here | 20:54 | |
| JayF | I'm going to talk abstract a little bit | 20:55 |
| JayF | since I can't talk exact network sizes and exact numbers | 20:55 |
| soren | Sure. | 20:55 |
| JayF | assume you have a /16. That's 255 /24s. | 20:55 |
| JayF | 1.0.0.0/16 | 20:55 |
| soren | Thereabouts. | 20:55 |
| JayF | If I give out addresses from that, 1.0.1.3, that's in the same broadcast domain with 2^16 different potential IPs | 20:55 |
| JayF | if I instead break it down into 1.0.0.0/24, 1.0.1.0/24, 1.0.2.0/24, etc etc | 20:56 |
| JayF | you end up with 256 different broadcast domains (2^8) but each one has a max of 256 potential IPs in it | 20:56 |
| soren | broadcast domain is a layer below IP, so I'm not entirely sure how subnets factor into it? | 20:56 |
| JayF | depending on which broadcast you're talking about | 20:57 |
| JayF | broadcast absolutely is a concept in level 3 as well | 20:57 |
| soren | Not in any really useful sense. | 20:57 |
| *** penick has quit IRC | 20:57 | |
| JayF | I suspect, but don't know, that some of the security bits preventing mac spoofing is helping us keep from having a chatty ethernet broadcast domain | 20:57 |
| JayF | but I don't have enough in-depth knowledge as to how our switches are configured to give a better response than that | 20:57 |
| soren | If you're all connected to the same ethernet broadcast domain, you can have all the subnets you want, but if I ping my subnet's broadcast address, everyone on the ethernet broadcast domain will see it and inspect it to determine if they need to bother with it. | 20:58 |
| JayF | sure, unless your switch is doing things to prevent that :) | 20:58 |
| openstackgerrit | Merged openstack/ironic: Add info on creating a tftp map file https://review.openstack.org/138864 | 20:58 |
| soren | And what would that be | 20:58 |
| soren | ? | 20:58 |
| JayF | $cisco_things that I don't know :) | 20:59 |
| soren | If not something at the ethernet level. | 20:59 |
| JayF | warned you that you were at the edge of my knowledge ;) | 20:59 |
| soren | VLANs or whatever. | 20:59 |
| adam_g | devananda, .:devananda:. and you need each node to be managed only by Ironic Conductor(s) which have access to that subnetd <- is that possible ATM? im hitting some similar issues /w multinode testing + OVS where that would help | 21:00 |
| devananda | adam_g: nope | 21:02 |
| devananda | all nodes distribute randomly across the hash ring today | 21:02 |
| adam_g | thats what i thought | 21:02 |
| jroll | that's an interesting thought | 21:02 |
| *** penick has joined #openstack-ironic | 21:03 | |
| *** romcheg has joined #openstack-ironic | 21:05 | |
| *** ChuckC has joined #openstack-ironic | 21:06 | |
| russell_h | soren: reading a little backlog, I'm even less of a network guy than JayF, but I think I can partially answer this question | 21:08 |
| *** mikedillion has quit IRC | 21:08 | |
| russell_h | soren: every cab has 2x top-of-rack switches, hooked to a pair of what we call "aggrs" | 21:09 |
| russell_h | we terminate L2 at the aggr layer | 21:09 |
| russell_h | and there is no guarantee at all that 2 OnMetal servers will be on the same aggr | 21:09 |
| russell_h | so broadcast becomes.. not useful | 21:10 |
| *** ChuckC has quit IRC | 21:11 | |
| *** ChuckC_ has joined #openstack-ironic | 21:11 | |
| soren | russell_h: So how do you make sure they're not assigned IP's in the same subnet? | 21:12 |
| soren | russell_h: ...if they're not in the same broadcast domain. | 21:12 |
| russell_h | each aggr has its own subnet | 21:12 |
| soren | Separate cells or something? | 21:12 |
| soren | Right, but how does that get mapped to the IPs that are assigned to instances by Neutron? | 21:13 |
| russell_h | today we only have a single aggr pair per cell, yeah | 21:13 |
| soren | Ok, makes sense. | 21:13 |
| russell_h | we've talked about how to handle it if we do more | 21:13 |
| russell_h | I forget if we came to a conclusion of that, it didn't seem hard, just a thing to do | 21:13 |
| russell_h | s/of/on/ | 21:13 |
| soren | Ok... Again, maybe I'm missing a piece of the puzzle. | 21:14 |
| russell_h | so thining about it | 21:14 |
| russell_h | I think our Neutron plugin can handle this | 21:14 |
| russell_h | it knows which ToR each server is hooked up to | 21:14 |
| russell_h | I'm not sure if it _does_ this, but we could map ToRs to subnets | 21:15 |
| soren | I'm trying to see if I can find where your plugin assigns IPs. | 21:16 |
| * russell_h too | 21:16 | |
| jroll | it doesn't? | 21:16 |
| russell_h | morgabra: ^ is the guy | 21:16 |
| *** marcoemorais has quit IRC | 21:16 | |
| jroll | that's handled by regular neutron afaik | 21:16 |
| jroll | nova creates the port, that provisions an IP | 21:16 |
| *** marcoemorais has joined #openstack-ironic | 21:16 | |
| jroll | nova teels neutron to create the port* | 21:17 |
| russell_h | jroll: is the IP assigned to the port before it gets bound to a server though? | 21:17 |
| russell_h | I guess it might | 21:17 |
| JayF | yes | 21:17 |
| soren | But if neutron does that without consulting your plugin, how could it map tors to subnets? | 21:17 |
| JayF | commit=false vs commit=true | 21:17 |
| JayF | russell_h: ^ the config is created uncommitted, post deploy we commit it to the switch | 21:17 |
| russell_h | I think I'm wrong, perhaps it couldn't | 21:17 |
| JayF | russell_h: I think that answers your question? | 21:17 |
| soren | So nova creates the port in Neutron. | 21:18 |
| soren | Neutron assigns the IP. | 21:18 |
| soren | (ISTR that "nova boot --nic" lets you specify an IP?) | 21:18 |
| soren | ...and somehow that all gets applied to the instance and switches are configured.. | 21:19 |
| soren | Ok. | 21:19 |
| soren | So if I were to fix this, I need to fix Neutron's IP assignment code to be subnet aware. | 21:19 |
| soren | ...and then I'd probably be golden. | 21:19 |
| russell_h | right, so when we grow to multiple aggr pairs per cell we'd need to | 21:19 |
| russell_h | right | 21:19 |
| russell_h | and somehow we'd need to advise neutron which subnet the IP should be in | 21:20 |
| russell_h | which actually sounds harder on the nova side | 21:20 |
| soren | Really? | 21:20 |
| soren | Why? | 21:20 |
| soren | I don't even see how Nova would be affected. | 21:20 |
| russell_h | well, if I recollect, the port creation actually happens before it hits our virt driver | 21:20 |
| russell_h | although I'm like 0 for 3 on recollecting things in this discussion | 21:21 |
| russell_h | in any case, wherever Nova hits Neutron to create the port, it would need to have decided which aggr pair the instance will be scheduled to already | 21:21 |
| russell_h | I suppose thats probably already the case | 21:21 |
| soren | Ah, yeah, if it's supposed to take network location into account, it needs to happen post scheduling. | 21:21 |
| russell_h | right | 21:22 |
| russell_h | I expect it already does, but we've tried not to touch too much Nova code outside the virt driver | 21:22 |
| soren | Yeah. | 21:22 |
| *** rushiagr_away is now known as rushiagr | 21:22 | |
| soren | Well, I need to wander off. | 21:22 |
| soren | Good talk! | 21:22 |
| dlaube | anyone know how I can specify some kernel= options within the deploy image? I need to set my console settings so I can see output on the my remote serial console | 21:24 |
| jroll | dlaube: pxe_append_params or something, in the conf | 21:26 |
| dlaube | thanks jroll | 21:26 |
| dlaube | I'll google that up | 21:26 |
| SpamapS | soren: I'm working on the L3 problems you're talking about right now. | 21:27 |
| dlaube | pxe_append_params='console=ttyS0,9600' | 21:27 |
| dlaube | sweet! | 21:27 |
| jroll | :) | 21:27 |
| dlaube | ;) | 21:27 |
| SpamapS | soren: and lifeless has submitted a POC for how to make Nova DTRT when physical segments are involved: https://review.openstack.org/#/c/133260/ | 21:27 |
| *** marcoemorais has quit IRC | 21:31 | |
| *** igordcard has joined #openstack-ironic | 21:32 | |
| JayF | dlaube: note that sertting exists in both [agent] and [pxe] section | 21:32 |
| JayF | dlaube: iirc | 21:32 |
| *** marcoemorais1 has joined #openstack-ironic | 21:32 | |
| *** marcoemorais1 has quit IRC | 21:32 | |
| *** marcoemorais1 has joined #openstack-ironic | 21:33 | |
| *** marcoemorais1 has quit IRC | 21:33 | |
| jroll | slightly different names, but yes | 21:33 |
| SpamapS | soren: https://review.openstack.org/#/c/133260/1/nova/virt/driver.py specifically describes exactly what you want | 21:35 |
| SpamapS | soren: then one needs to add the bit in the ironic driver which digs into the properties and returns the subnet_id that was assigned during inventory. | 21:37 |
| NobodyCam | anyone know how to tell if swift Temp-URL-Key is actually set? | 21:42 |
| NobodyCam | fyi: the answer is yes: looks like its in stat output :) | 21:46 |
| *** marcoemorais has joined #openstack-ironic | 21:54 | |
| *** hemna__ has quit IRC | 21:57 | |
| *** dprince has quit IRC | 21:58 | |
| *** hemna__ has joined #openstack-ironic | 21:59 | |
| *** penick has quit IRC | 21:59 | |
| *** hemna has joined #openstack-ironic | 22:03 | |
| *** penick has joined #openstack-ironic | 22:04 | |
| *** hemna__ has quit IRC | 22:04 | |
| *** marcoemorais has quit IRC | 22:04 | |
| *** marcoemorais1 has joined #openstack-ironic | 22:04 | |
| *** marcoemorais1 has quit IRC | 22:04 | |
| *** marcoemorais has joined #openstack-ironic | 22:05 | |
| openstackgerrit | Devananda van der Veen proposed openstack/ironic: Add fsm.py module https://review.openstack.org/139215 | 22:10 |
| openstackgerrit | Devananda van der Veen proposed openstack/ironic: Begin using the state machine for node deploy/teardown https://review.openstack.org/139216 | 22:10 |
| openstackgerrit | Devananda van der Veen proposed openstack/ironic: Refactor async helper methods in conductor/manager.py https://review.openstack.org/139217 | 22:10 |
| *** marcoemorais has quit IRC | 22:11 | |
| *** marcoemorais has joined #openstack-ironic | 22:11 | |
| *** marcoemorais has quit IRC | 22:11 | |
| *** marcoemorais has joined #openstack-ironic | 22:12 | |
| devananda | victor_lowther: ^ | 22:14 |
| *** penick has quit IRC | 22:14 | |
| *** andreykurilin_ has quit IRC | 22:15 | |
| *** andreykurilin__ has joined #openstack-ironic | 22:15 | |
| *** erwan_taf has quit IRC | 22:15 | |
| *** ryanpetrello has quit IRC | 22:18 | |
| *** Masahiro has joined #openstack-ironic | 22:21 | |
| openstackgerrit | Ruby Loo proposed openstack/ironic: Use oslo.config.cfg.CONF with oslo.config generator https://review.openstack.org/139221 | 22:24 |
| *** jcoufal has joined #openstack-ironic | 22:26 | |
| *** Masahiro has quit IRC | 22:26 | |
| *** krtaylor has quit IRC | 22:33 | |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic: Remove python-tox from dev-quickstart docs https://review.openstack.org/139223 | 22:35 |
| * devananda takes a break | 22:35 | |
| jroll | clif_h: ^ there's the docs fix for that thing you ran into | 22:35 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic: Updated from global requirements https://review.openstack.org/139229 | 22:37 |
| *** penick has joined #openstack-ironic | 22:37 | |
| *** alexpilotti has quit IRC | 22:45 | |
| *** jcoufal has quit IRC | 22:48 | |
| clif_h | jroll: the same issue exists under opensuse? | 22:49 |
| *** romcheg has quit IRC | 22:57 | |
| *** romcheg has joined #openstack-ironic | 22:57 | |
| *** romcheg has quit IRC | 22:58 | |
| NobodyCam | brb | 22:59 |
| jroll | clif_h: idk, but no point in using distro packages for that IMO | 22:59 |
| clif_h | I suppose pip install should *always* do the right thing | 23:00 |
| clif_h | one would hope | 23:00 |
| *** andreykurilin__ has quit IRC | 23:09 | |
| *** vdrok has quit IRC | 23:09 | |
| *** vdrok has joined #openstack-ironic | 23:10 | |
| *** andreykurilin_ has joined #openstack-ironic | 23:10 | |
| *** penick has quit IRC | 23:18 | |
| *** shakamunyi has quit IRC | 23:19 | |
| *** shakamunyi has joined #openstack-ironic | 23:20 | |
| *** naohirot has joined #openstack-ironic | 23:21 | |
| *** ryanpetrello has joined #openstack-ironic | 23:32 | |
| * NobodyCam steps away to make some food... bbiab | 23:42 | |
| *** krtaylor has joined #openstack-ironic | 23:49 | |
| openstackgerrit | Clif Houck proposed openstack/ironic-python-agent: Fix badly named function _get_kernel_params() https://review.openstack.org/139269 | 23:55 |
| *** smoriya has joined #openstack-ironic | 23:56 | |
| smoriya | 3 | 23:56 |
| *** Haomeng|2 has joined #openstack-ironic | 23:58 | |
| *** Haomeng has quit IRC | 23:59 | |
| *** smoriya has quit IRC | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!