| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Allow noauth for Neutron https://review.openstack.org/98922 | 00:01 |
|---|---|---|
| JoshNang | rloo_: updated that patch :) sorry for the lag, was in a meeting. i think i did lose things in that spec (local got out of sync it seems). i'll fix it tonight/tmr morning. thanks for the reviews! | 00:01 |
| faizan | Devananda: Its regarding nova ironic virt driver changes, where we need to pass the boot options to the ironic node via instance_info field. | 00:05 |
| faizan | There is one more change where nova scheduler should pick the ironic node which support a specified boot mode. | 00:07 |
| faizan | Devananda: wanted to know how best we can make these changes. | 00:09 |
| faizan | Devananda: I can share you the code changes required in nova virt dirver. I have described them in the uefi spec patch | 00:13 |
| NobodyCam | devananda: you want to land that | 00:15 |
| NobodyCam | ?? | 00:15 |
| NobodyCam | oh brb | 00:16 |
| devananda | NobodyCam: it's a bug fix for a High impact bug (rebuild unable to update image, tripleo can't upgrade nodes) | 00:17 |
| devananda | faizan: if you have something I can look at later, that'd be great -- perhaps post a link to code samples on the spec review | 00:17 |
| devananda | faizan: and I'll try to get to it again later this week | 00:18 |
| faizan | devananda: thanks I will post a link to the code samples later today. | 00:19 |
| *** chuckC has quit IRC | 00:19 | |
| faizan | Devananda: Is it ok if I bug you for one more review comment? | 00:21 |
| *** ramineni has joined #openstack-ironic | 00:21 | |
| faizan | Devananda: Its regarding - "IPMI does not support changing/querying the boot type" | 00:24 |
| NobodyCam | devananda: +2 but not +a'd | 00:24 |
| NobodyCam | now off to Dairy Queen to pick up dinner | 00:27 |
| devananda | faizan: what's the question? I'll answer when possible | 00:28 |
| *** overlayer has quit IRC | 00:32 | |
| ramineni | devananda: Hi | 00:33 |
| ramineni | devananda: did you get a chance to look at firmware update design spec - https://review.openstack.org/#/c/100842/ | 00:34 |
| devananda | ramineni: if I did, I would have posted a review on it. otherwise, no, I haven't had time to review everything | 00:35 |
| ramineni | devananda: any comments would be helpful. | 00:35 |
| ramineni | devananda: ok | 00:37 |
| faizan | devananda: ipmitool dont have the capability to change the boot mode, but other vendors can do this with their tools like iLO, Thats why I have suggested to add dummy vendor pass through in ipmitool vendor passthrou, which will return failure if current and next boot modes are different. | 00:39 |
| devananda | faizan: can ipmitool *check* the current boot mode? | 00:40 |
| devananda | faizan: if not, then it seems like that won't work, and you should clearly state that this isn't supported by ipmitool | 00:40 |
| faizan | This way we can have different class of drivers, like pxe_ilo, where we can use ilo capability to change boot mode and use pxe deploy | 00:40 |
| jroll | what do we mean by 'boot mode' here? | 00:41 |
| devananda | jroll: BIOS // UEFI | 00:41 |
| jroll | ah | 00:41 |
| * jroll hides again | 00:41 | |
| devananda | heh | 00:41 |
| faizan | ipmitool does not have the capability to query the bootmode either, but we can add the current boot mode and the next boot mode of the node as the node property | 00:43 |
| devananda | faizan: that doesn't make sense | 00:43 |
| devananda | faizan: if the driver doesn't support it, a request to get the current boot mode should error | 00:44 |
| faizan | and pxe driver can check this node property before sending a request to vendor pass through for making the switch | 00:44 |
| devananda | faizan: and similarly a request to change the boot mode should error | 00:44 |
| faizan | I agree with that, | 00:50 |
| faizan | devananda: Instead of doing get/set of bootmode, why can't we make these node properties static? and let admin fill those fields while creating the nodes | 00:56 |
| *** ccrouch has joined #openstack-ironic | 00:59 | |
| faizan | devananda: and also leave it to the respective vendor driver, if they have the capability to get/set the boot mode let them implement and update the node properties on the fly | 00:59 |
| *** rloo_ has quit IRC | 01:05 | |
| *** chuckC has joined #openstack-ironic | 01:07 | |
| *** eghobo has quit IRC | 01:13 | |
| *** nosnos has joined #openstack-ironic | 01:13 | |
| *** ramineni has quit IRC | 01:19 | |
| *** zdiN0bot has joined #openstack-ironic | 01:20 | |
| *** Haomeng has joined #openstack-ironic | 01:20 | |
| *** nosnos has quit IRC | 01:20 | |
| *** Haomeng|2 has quit IRC | 01:21 | |
| *** faizan has quit IRC | 01:26 | |
| *** zdiN0bot has quit IRC | 01:57 | |
| *** radsy has quit IRC | 01:59 | |
| *** zdiN0bot has joined #openstack-ironic | 01:59 | |
| *** blamar has joined #openstack-ironic | 02:01 | |
| *** killer_prince has quit IRC | 02:26 | |
| *** Sam_S has left #openstack-ironic | 02:36 | |
| *** ramineni has joined #openstack-ironic | 02:52 | |
| *** eghobo has joined #openstack-ironic | 03:01 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Allow noauth for Neutron https://review.openstack.org/98922 | 03:02 |
| *** killer_prince has joined #openstack-ironic | 03:10 | |
| *** killer_prince is now known as lazy_prince | 03:10 | |
| *** dwalleck has joined #openstack-ironic | 03:11 | |
| *** zdiN0bot has quit IRC | 03:25 | |
| *** Poornima has joined #openstack-ironic | 03:35 | |
| *** Nisha has joined #openstack-ironic | 03:52 | |
| *** nosnos has joined #openstack-ironic | 03:52 | |
| *** max_lobur has joined #openstack-ironic | 03:54 | |
| *** max_lobur has quit IRC | 04:05 | |
| *** eghobo has quit IRC | 04:05 | |
| *** ajc_ has joined #openstack-ironic | 04:17 | |
| *** yongli has quit IRC | 04:21 | |
| *** blamar has quit IRC | 04:28 | |
| *** k4n0 has joined #openstack-ironic | 04:39 | |
| *** pcrews has quit IRC | 04:42 | |
| *** bvivek has joined #openstack-ironic | 04:43 | |
| *** rakesh_hs has joined #openstack-ironic | 04:46 | |
| *** blamar has joined #openstack-ironic | 04:50 | |
| *** amitpp has joined #openstack-ironic | 04:55 | |
| *** sabah has joined #openstack-ironic | 05:02 | |
| *** blamar has quit IRC | 05:05 | |
| *** rameshg87_afk is now known as rameshg87 | 05:05 | |
| *** rameshg87 has quit IRC | 05:05 | |
| *** rameshg87 has joined #openstack-ironic | 05:06 | |
| *** pcrews has joined #openstack-ironic | 05:11 | |
| *** loki184 has joined #openstack-ironic | 05:14 | |
| *** stendulker has joined #openstack-ironic | 05:18 | |
| *** pcrews has quit IRC | 05:28 | |
| *** Penick has joined #openstack-ironic | 05:45 | |
| *** harlowja is now known as harlowja_away | 05:47 | |
| *** dwalleck has quit IRC | 05:58 | |
| *** dwalleck has joined #openstack-ironic | 05:58 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/103291 | 06:01 |
| *** wanyen has quit IRC | 06:02 | |
| *** max_lobur has joined #openstack-ironic | 06:02 | |
| *** coolsvap|afk is now known as coolsvap | 06:09 | |
| *** dwalleck_ has joined #openstack-ironic | 06:22 | |
| *** dwalleck has quit IRC | 06:26 | |
| *** lazy_prince is now known as killer_prince | 06:28 | |
| *** jcoufal has joined #openstack-ironic | 06:31 | |
| *** pradipta_away is now known as pradipta | 06:33 | |
| *** killer_prince is now known as lazy_prince | 06:39 | |
| *** Nisha has quit IRC | 06:40 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 06:53 | |
| *** takadayuiko has joined #openstack-ironic | 07:20 | |
| *** Penick has quit IRC | 07:23 | |
| *** lsmola__ has joined #openstack-ironic | 07:24 | |
| *** ifarkas has joined #openstack-ironic | 07:26 | |
| openstackgerrit | Michael Davies proposed a change to openstack/python-ironicclient: Expose auth_ref in ironicclient client object https://review.openstack.org/104467 | 07:28 |
| *** max_lobur has quit IRC | 07:32 | |
| *** lucasagomes has joined #openstack-ironic | 07:36 | |
| *** Mikhail_D_ltp has quit IRC | 07:37 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 07:39 | |
| *** overlayer has joined #openstack-ironic | 07:41 | |
| *** dwalleck_ has quit IRC | 07:41 | |
| *** dwalleck has joined #openstack-ironic | 07:42 | |
| *** ndipanov has joined #openstack-ironic | 07:47 | |
| *** bvivek has quit IRC | 07:47 | |
| *** dwalleck has quit IRC | 07:52 | |
| openstackgerrit | Daryl Walleck proposed a change to openstack/ironic: Updates the Ironic on Devstack dev documentation https://review.openstack.org/104469 | 07:53 |
| *** nosnos has quit IRC | 08:05 | |
| *** nosnos has joined #openstack-ironic | 08:05 | |
| *** derekh_ has joined #openstack-ironic | 08:05 | |
| *** martyntaylor has joined #openstack-ironic | 08:12 | |
| *** athomas has joined #openstack-ironic | 08:14 | |
| *** bvivek has joined #openstack-ironic | 08:19 | |
| *** foexle has joined #openstack-ironic | 08:24 | |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Ironic nova driver to cache ironic client calls https://review.openstack.org/102695 | 08:31 |
| *** loki184 has quit IRC | 08:32 | |
| mrda | devananda: Just FYI - new version of the Ironic Nova driver auth patch up, along with a dependent review on ironicclient. Chat to you in 13 hours :) | 08:38 |
| mrda | Night Ironic! | 08:39 |
| *** mrda is now known as mrda-away | 08:39 | |
| *** jcoufal has quit IRC | 08:41 | |
| *** jcoufal has joined #openstack-ironic | 08:44 | |
| *** pelix has joined #openstack-ironic | 08:55 | |
| *** bvivek has quit IRC | 09:01 | |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Ironic nova driver to cache ironic client calls https://review.openstack.org/102695 | 09:02 |
| *** loki184 has joined #openstack-ironic | 09:03 | |
| *** loki184 has quit IRC | 09:04 | |
| *** romcheg has joined #openstack-ironic | 09:11 | |
| *** loki184 has joined #openstack-ironic | 09:17 | |
| *** loki184 has quit IRC | 09:22 | |
| *** mitz has quit IRC | 09:22 | |
| *** mitz has joined #openstack-ironic | 09:26 | |
| *** Alexei_9871 has joined #openstack-ironic | 09:34 | |
| *** rakesh_hs has quit IRC | 09:38 | |
| *** rakesh_hs has joined #openstack-ironic | 09:39 | |
| *** lazy_prince is now known as killer_prince | 09:40 | |
| *** killer_prince is now known as lazy_prince | 09:48 | |
| *** bvivek has joined #openstack-ironic | 09:52 | |
| *** overlayer has quit IRC | 09:55 | |
| *** martyntaylor has quit IRC | 09:55 | |
| *** Nisha has joined #openstack-ironic | 09:59 | |
| openstackgerrit | Imre Farkas proposed a change to openstack/ironic-specs: DRAC power driver https://review.openstack.org/99352 | 10:01 |
| Nisha | lucasagomes: hi | 10:12 |
| romcheg | Morning all! | 10:13 |
| *** rakesh_hs has quit IRC | 10:14 | |
| *** rakesh_hs has joined #openstack-ironic | 10:15 | |
| *** pradipta is now known as pradipta_away | 10:22 | |
| *** lazy_prince has quit IRC | 10:55 | |
| *** lazy_prince has joined #openstack-ironic | 10:56 | |
| *** geekyogi1 has quit IRC | 10:57 | |
| *** Haomeng|2 has joined #openstack-ironic | 10:57 | |
| *** geekyogi has joined #openstack-ironic | 10:58 | |
| *** Haomeng has quit IRC | 10:58 | |
| *** derekh_ has quit IRC | 10:59 | |
| *** ramineni has quit IRC | 11:00 | |
| *** derekh_ has joined #openstack-ironic | 11:00 | |
| *** geekyogi has quit IRC | 11:10 | |
| *** sabah has quit IRC | 11:22 | |
| *** Poornima has quit IRC | 11:26 | |
| *** stendulker has quit IRC | 11:27 | |
| *** jcoufal has quit IRC | 11:30 | |
| *** jcoufal has joined #openstack-ironic | 11:30 | |
| *** aswadr has joined #openstack-ironic | 11:47 | |
| lazy_prince | lifeless: hi, there..? | 11:51 |
| *** mkerrin has quit IRC | 11:51 | |
| *** geekyogi has joined #openstack-ironic | 11:53 | |
| *** nosnos has quit IRC | 11:55 | |
| *** nosnos_ has joined #openstack-ironic | 11:58 | |
| *** mkerrin has joined #openstack-ironic | 11:59 | |
| *** jcoufal has quit IRC | 12:01 | |
| *** jcoufal has joined #openstack-ironic | 12:04 | |
| *** proffalken has joined #openstack-ironic | 12:16 | |
| proffalken | hi all, is https://review.openstack.org/#/c/98506/3/specs/juno/agent-driver.rst the best/most relevant documentation to read when considering how to auto-discover a rack of equipment and get it into the correct state for deployment via TripleO/HEAT? | 12:18 |
| *** jcoufal has quit IRC | 12:24 | |
| *** jcoufal has joined #openstack-ironic | 12:24 | |
| *** Haomeng has joined #openstack-ironic | 12:24 | |
| *** dtantsur has quit IRC | 12:24 | |
| *** dtantsur has joined #openstack-ironic | 12:24 | |
| *** Haomeng|2 has quit IRC | 12:25 | |
| *** linggao has joined #openstack-ironic | 12:29 | |
| *** ajc_ has quit IRC | 12:30 | |
| *** coolsvap is now known as coolsvap|afk | 12:31 | |
| *** athomas_ has joined #openstack-ironic | 12:34 | |
| *** athomas has quit IRC | 12:34 | |
| *** bvivek has quit IRC | 12:35 | |
| *** zigo has quit IRC | 12:37 | |
| *** zigo has joined #openstack-ironic | 12:37 | |
| *** jbjohnso has joined #openstack-ironic | 12:38 | |
| *** zigo has quit IRC | 12:41 | |
| *** zigo has joined #openstack-ironic | 12:42 | |
| *** k4n0 has quit IRC | 12:44 | |
| *** zigo has quit IRC | 12:45 | |
| *** avozza has joined #openstack-ironic | 12:50 | |
| *** Nisha has quit IRC | 12:56 | |
| *** martyntaylor has joined #openstack-ironic | 12:58 | |
| *** martyntaylor has left #openstack-ironic | 13:02 | |
| *** ramineni has joined #openstack-ironic | 13:04 | |
| *** amitpp has quit IRC | 13:17 | |
| *** pcrews has joined #openstack-ironic | 13:18 | |
| *** overlayer has joined #openstack-ironic | 13:26 | |
| *** matty_dubs|gone is now known as matty_dubs | 13:28 | |
| *** Haomeng|2 has joined #openstack-ironic | 13:36 | |
| *** jdob has joined #openstack-ironic | 13:36 | |
| jroll | proffalken: discovery doesn't really exist yet... that will help, this will also help: https://review.openstack.org/#/c/102565/ | 13:37 |
| *** Haomeng has quit IRC | 13:37 | |
| jroll | proffalken: but the code doesn't exist | 13:37 |
| *** nosnos_ has quit IRC | 13:37 | |
| *** Haomeng has joined #openstack-ironic | 13:40 | |
| NobodyCam | good morning Ironic | 13:41 |
| matty_dubs | Morning | 13:41 |
| jroll | mornin | 13:42 |
| *** Haomeng|2 has quit IRC | 13:42 | |
| * jroll passes around coffee | 13:42 | |
| NobodyCam | morning matty_dubs jroll :) | 13:42 |
| *** rakesh_hs has quit IRC | 13:42 | |
| * NobodyCam takes the coffee and hands out bagels | 13:42 | |
| jroll | \o/ | 13:42 |
| NobodyCam | :) | 13:43 |
| NobodyCam | is anyone working tomorrow? | 13:43 |
| jroll | nope | 13:44 |
| matty_dubs | Not I. | 13:44 |
| *** avozza has quit IRC | 13:48 | |
| NobodyCam | :-p was just checking :-p | 13:49 |
| *** nosnos has joined #openstack-ironic | 13:49 | |
| proffalken | jroll: thanks. It's something I'd be interested in helping develop so I'll give them both a good read. What's the approach on implementation? Starting writing more specs and get them reviewed? | 13:53 |
| jroll | proffalken: so 102565 is for the generic bits | 13:54 |
| jroll | so like, implement that. then implement it for a certain driver :) | 13:54 |
| jroll | perhaps with another spec; idk | 13:54 |
| proffalken | fairy nuff | 13:54 |
| jroll | that one is a bit in flux... might want to wait until that lands or whatever | 13:54 |
| jroll | or chat with dtantsur (he's currently on PTO) | 13:55 |
| proffalken | kk, that's cool. Just want to know current state and who to talk to before I roll up my sleaves and break out the trance music... ;) | 13:55 |
| jroll | heh, sounds good | 13:56 |
| *** zigo has joined #openstack-ironic | 13:58 | |
| *** nosnos has quit IRC | 13:59 | |
| *** Mikhail_D_ltp has quit IRC | 14:00 | |
| openstackgerrit | Andrey Kurilin proposed a change to openstack/python-ironicclient: Remove aliases `arg` and `env` from utils https://review.openstack.org/104553 | 14:01 |
| *** dkehnx has quit IRC | 14:02 | |
| *** andreykurilin has joined #openstack-ironic | 14:10 | |
| *** ndipanov has quit IRC | 14:11 | |
| *** ndipanov has joined #openstack-ironic | 14:11 | |
| *** rloo has joined #openstack-ironic | 14:18 | |
| *** max_lobur has joined #openstack-ironic | 14:23 | |
| *** dkehnx has joined #openstack-ironic | 14:25 | |
| *** jgrimm has joined #openstack-ironic | 14:27 | |
| *** overlayer has quit IRC | 14:29 | |
| lucasagomes | hey all | 14:33 |
| lucasagomes | someone knows Nisha? he pinged me but I missed it | 14:33 |
| rloo | hey lucasagomes | 14:33 |
| NobodyCam | morning lucasagomes | 14:33 |
| matty_dubs | Heya lucasagomes | 14:33 |
| lucasagomes | (I'm in the mid-cycle) | 14:33 |
| lucasagomes | he pinged me yesterday as well | 14:33 |
| ifarkas | morning NobodyCam | 14:33 |
| lucasagomes | and I missed too | 14:33 |
| lucasagomes | :( | 14:33 |
| ifarkas | NobodyCam, I have a question regarding https://review.openstack.org/#/c/99352/8/specs/juno/drac-power-driver.rst | 14:33 |
| NobodyCam | morning ifarkas | 14:33 |
| NobodyCam | ifarkas: sure | 14:34 |
| *** jbjohnso has quit IRC | 14:34 | |
| lucasagomes | NobodyCam, he works for HP seems | 14:34 |
| ifarkas | NobodyCam, about the default port and protocol, I think Ironic shouldn't set any default value for the driver_info field. If the user doesn't set it, it should not connect. Default values makes sense in case of config options but it's a bit strange for the driver_info field. What do you think? | 14:35 |
| NobodyCam | lucasagomes: I chattted with nisha the other day | 14:37 |
| linggao | morning | 14:37 |
| NobodyCam | ifarkas: ws-man has default ports no? | 14:38 |
| linggao | morning NobodyCam lucasagomes rloo ifarkas matty_dubs proffalken jroll | 14:39 |
| ifarkas | NobodyCam, yeah, I guess that's a safe assumption. I haven't set up any, just using one. | 14:39 |
| *** dkehn_ has joined #openstack-ironic | 14:39 | |
| ifarkas | morning linggao | 14:39 |
| rloo | morning linggao | 14:39 |
| lucasagomes | morning | 14:39 |
| lucasagomes | rloo, I'm currently travelling, I will put another spec for the mgmt interface soon | 14:40 |
| linggao | morning devananda | 14:40 |
| lucasagomes | rloo, probably next monday | 14:40 |
| rloo | lucasagomes: no worries. I'm not in a hurry ;) | 14:40 |
| lucasagomes | aight, sorry for that | 14:40 |
| NobodyCam | ifarkas: port = 5985 / #ssl_port = 5986 <- from https://github.com/Openwsman/openwsman | 14:40 |
| linggao | just like to let you guys/gilrs know that I am going to have a 4-week vacation. | 14:40 |
| NobodyCam | morning linggao | 14:40 |
| NobodyCam | morning rloo | 14:41 |
| rloo | yay linggao. Enjoy! | 14:41 |
| NobodyCam | linggao: wow I want one too | 14:41 |
| rloo | morning NobodyCam. | 14:41 |
| ifarkas | NobodyCam, that's the client library but not only Dell products are using openwsman and it depends on how you set it up | 14:41 |
| ifarkas | NobodyCam, eg. I am using it with https on port 443 | 14:41 |
| *** dkehn_ has quit IRC | 14:42 | |
| * proffalken waves to everyone who's just turned up | 14:42 | |
| *** dkehn_ has joined #openstack-ironic | 14:42 | |
| linggao | NobodyCam, rloo, devananda, lucasagomes I'll be back the Auguest 4th. | 14:42 |
| *** dkehnx has quit IRC | 14:43 | |
| linggao | So you guys have good time for mid-cycle meeting. | 14:43 |
| *** dkehn_ is now known as dkehnx | 14:43 | |
| NobodyCam | ifarkas: so the hardware had no default config? | 14:43 |
| NobodyCam | s/had/has/ | 14:44 |
| ifarkas | NobodyCam, btw, are you suggesting to use a default value when the user didn't provide the proper driver_info fields? | 14:44 |
| * NobodyCam does not have dell HW | 14:44 | |
| *** jbjohnso has joined #openstack-ironic | 14:44 | |
| ifarkas | NobodyCam, I guess but I have never set up any. I just have an remote access to one | 14:44 |
| linggao | I still have a few patches open https://review.openstack.org/#/q/owner:linggao+status:open,n,z Please review them while I am on vacation :). I'll take care of them after I am back. | 14:45 |
| NobodyCam | ifarkas: I was thinking the hardware would have a default value | 14:45 |
| NobodyCam | ie with iLo There is a default port, I can change it, but out of the box 443 works | 14:46 |
| ifarkas | NobodyCam, yeah. should Ironic use that driver_info value even the user didn't provide it? | 14:46 |
| NobodyCam | ifarkas: the user had to select the drac driver, | 14:47 |
| NobodyCam | why make them enter the default vault id they did change it | 14:47 |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Ironic nova driver to cache ironic client calls https://review.openstack.org/102695 | 14:47 |
| NobodyCam | *value if * | 14:47 |
| ifarkas | NobodyCam, okay, just wasn't sure | 14:47 |
| ifarkas | if I understand it correctly | 14:48 |
| NobodyCam | ya I was asking I do not have drac hw to test with | 14:48 |
| NobodyCam | my thought was the mfg sets a default (or I think they should) and we too can use that default unless the user has changed it for their env | 14:49 |
| ifarkas | NobodyCam, ok, makes sense | 14:49 |
| ifarkas | NobodyCam, according to http://en.community.dell.com/techcenter/systems-management/f/4494/t/19499050.aspx the default values depend on what the user selected on the OME disc wizard | 14:50 |
| NobodyCam | lol | 14:50 |
| NobodyCam | ofc | 14:50 |
| ifarkas | NobodyCam, if they select secure the default is 443, if not, the default is 623 | 14:50 |
| *** max_lobur has quit IRC | 14:50 | |
| NobodyCam | so we could get have a default port value but not protocol | 14:50 |
| *** foexle has quit IRC | 14:51 | |
| NobodyCam | s/get// | 14:51 |
| ifarkas | NobodyCam, I think if secure is selected, then the default port is 443 and the default protocol is https. Otherwise the port is 623 and the port is http | 14:52 |
| ifarkas | NobodyCam, maybe we can use the secure one as a default | 14:52 |
| NobodyCam | :) ok ya :) | 14:53 |
| ifarkas | cool ;-) | 14:53 |
| NobodyCam | i really cann;t see dell forcing some who ordered 1000 servers to run the wizard on all of them | 14:53 |
| NobodyCam | thou they may | 14:54 |
| NobodyCam | :-p | 14:54 |
| lucasagomes | linggao, oh great! enjoy ur vacations | 14:54 |
| linggao | lucasagomes, thanks. | 14:54 |
| *** coolsvap|afk is now known as coolsvap | 14:55 | |
| NobodyCam | linggao: is there someone who is going to take over your on going work while your out enjoying yourself? | 14:55 |
| *** lazy_prince is now known as killer_prince | 14:59 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Rename/update common/tftp to drivers/modules/pxe_utils https://review.openstack.org/103595 | 15:01 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add iPXE support for Ironic https://review.openstack.org/99318 | 15:01 |
| ifarkas | NobodyCam, iiuc, you just need to run it on one machine and it will discover all dracs on the network and set them up | 15:03 |
| ifarkas | NobodyCam, but yeah, running it one by one would be terrible | 15:03 |
| linggao | NobodyCam, no. I need to take care of them when I am back, if I can remeber what I did. :) | 15:03 |
| NobodyCam | linggao: as your gone for a month. would you mind if we (the team) made changes / updated to your work if it becomes a blocker for others? | 15:05 |
| *** max_lobur has joined #openstack-ironic | 15:06 | |
| NobodyCam | s/updated/updates/ | 15:06 |
| *** mmitchell_ has quit IRC | 15:06 | |
| linggao | NobodyCam, feel free to modify them. | 15:09 |
| NobodyCam | linggao: TY :) | 15:10 |
| linggao | NobodyCam, have a nice summer! | 15:11 |
| devananda | morning, all | 15:11 |
| NobodyCam | good morning devananda :) | 15:11 |
| linggao | morning devananda | 15:11 |
| *** mmitchell has joined #openstack-ironic | 15:11 | |
| linggao | devananda, I just told the team that I will have a 4-week vacation. Will be back on Auguest 4th. | 15:12 |
| *** hemna_ has joined #openstack-ironic | 15:20 | |
| *** ifarkas has quit IRC | 15:22 | |
| *** lynxman has quit IRC | 15:24 | |
| *** ifarkas has joined #openstack-ironic | 15:24 | |
| *** lynxman has joined #openstack-ironic | 15:25 | |
| devananda | linggao: nice! have a great time :) | 15:25 |
| linggao | Deva, thanks. enjoy the summer. | 15:27 |
| openstackgerrit | Imre Farkas proposed a change to openstack/ironic-specs: DRAC power driver https://review.openstack.org/99352 | 15:29 |
| *** hemna_ has quit IRC | 15:30 | |
| lucasagomes | morning devananda | 15:30 |
| *** jcoufal has quit IRC | 15:31 | |
| ifarkas | morning devananda | 15:33 |
| ifarkas | devananda, a question regarding https://review.openstack.org/#/c/90126/7..12/ironic/common/disk_partitioner.py | 15:34 |
| ifarkas | devananda, if you check the 7th rev, the _wait_for_disk_to_become_available method was inline and you suggested to make it an instance method on the class because it's easier to test. in the 12th revision, you suggested to convert it back to inline. | 15:34 |
| ifarkas | devananda, I am a bit confused, maybe I did not understand something correctly? | 15:35 |
| *** dwalleck has joined #openstack-ironic | 15:36 | |
| devananda | ifarkas: .... oops. you're aboslutely correct. sorry about that | 15:37 |
| ifarkas | devananda, ok. so it should be inline, right? | 15:38 |
| devananda | ifarkas: no - leave it as a separate method. it is better to test it that way. | 15:39 |
| ifarkas | devananda, ack, thanks ;-) | 15:39 |
| *** hemna_ has joined #openstack-ironic | 15:42 | |
| *** jbjohnso has quit IRC | 15:42 | |
| *** jbjohnso has joined #openstack-ironic | 15:43 | |
| *** jbjohnso_ has joined #openstack-ironic | 15:43 | |
| Shrews | devananda: So, wrt disk_available_least, it looks like this value is used only for live migrations doing block migration. AFAIK, ironic doesn't support that, so it may be enough to just set this value to 0. | 15:44 |
| Shrews | or set it to the same value as free_disk_gb | 15:44 |
| Shrews | the last may be more correct | 15:45 |
| Shrews | or not... confusing option :/ | 15:45 |
| *** jbjohnso has quit IRC | 15:48 | |
| *** geekyogi has quit IRC | 15:48 | |
| *** stendulker has joined #openstack-ironic | 15:48 | |
| devananda | Shrews: right. but I think it has more to do with the *host* that's running n-cpu | 15:49 |
| Shrews | devananda: It's the disk space available to the compute host FOR instances. In our case, that is the disk space on the instance itself | 15:50 |
| devananda | Shrews: AIUI, the entire use of that option makes no sense for ironic, since there isn't a 1:1 relationship between n-cpu:node | 15:50 |
| devananda | Shrews: yes. but it's used to identify an available n-cpu host to which a migration could happen, based on disk space, right? | 15:51 |
| devananda | Shrews: if we run ironic with 1 or >1 n-cpu, in either case, that function makes no sense | 15:51 |
| Shrews | devananda: right, doesn't really relate to ironic. which is why i suggested 0 | 15:52 |
| devananda | Shrews: even though a cold migration of an instance from one node to another is, strictly speaking, possible | 15:52 |
| devananda | Shrews: yea... but how's that going to affect Nova? | 15:52 |
| stendulker | JayF: Hello | 15:52 |
| Shrews | devananda: should only affect migrations. we have no official stance on migrations, do we? | 15:52 |
| devananda | Shrews: I guess my point is, what ever test is exercising that option probably should not be run if ironic is the only hypervisor available | 15:53 |
| JayF | stendulker: hey, I saw your comments and will try to reply today. I am busy this morning though | 15:53 |
| NobodyCam | JoshNang: around? | 15:53 |
| devananda | Shrews: official stance: we don't support it now. we *may* eventually support cold-migration. | 15:53 |
| Shrews | devananda: the tempest test is a generic test. deactivating it for ironic would be bad, imo | 15:54 |
| stendulker | JayF: Thank you. Had pinged you to let you know. :) | 15:54 |
| *** eghobo has joined #openstack-ironic | 15:54 | |
| devananda | Shrews: remind me - which test? | 15:54 |
| Shrews | devananda: tempest.api.compute.admin.test_hypervisor.HypervisorAdminTestJSON.test_get_hypervisor_show_details | 15:54 |
| Shrews | devananda: my suggestion is, set it to 0 for now. when we DO need to start supporting migrations in some form, revisit it. | 15:55 |
| Shrews | devananda: https://github.com/openstack/nova/blob/master/nova/scheduler/host_manager.py#L183-L192 | 15:57 |
| Shrews | if we set it to free_disk_gb, that block of code will run as-is today i think | 15:58 |
| Shrews | which may be the better choice | 15:59 |
| *** eghobo has quit IRC | 15:59 | |
| devananda | Shrews: and if disk_available_least is NONE, then we still get free_gb = compute['free_disk_gb'] | 16:00 |
| Shrews | right, which should be what's happening today | 16:00 |
| devananda | Shrews: i'm not sure I see how this relates to the tempest test, which is merely asserting that /sometehing/ was returned | 16:00 |
| devananda | Shrews: not that it has a value for disk_available_least | 16:00 |
| Shrews | devananda: the tempest test verifies that disk_available_least is an integer value, which it is not (it's null) | 16:00 |
| Shrews | devananda: it runs the returned json against a schema validator | 16:01 |
| Shrews | which is the failure | 16:01 |
| Shrews | devananda: https://github.com/openstack/tempest/blob/master/tempest/api_schema/compute/hypervisors.py#L27 | 16:03 |
| *** ndipanov has quit IRC | 16:03 | |
| *** dhellmann is now known as dhellmann_ | 16:04 | |
| *** eghobo has joined #openstack-ironic | 16:04 | |
| Shrews | devananda: oh! this is interesting... this schema def allows null: https://github.com/openstack/tempest/blob/master/tempest/api_schema/compute/hypervisors.py#L61 | 16:06 |
| Shrews | we could change tempest to allow nulls in all of the schemas for disk_available_least | 16:06 |
| * Shrews tries the change | 16:07 | |
| devananda | Shrews: :) | 16:07 |
| NobodyCam | bbt...brb | 16:15 |
| *** eguz has joined #openstack-ironic | 16:15 | |
| *** eguz has quit IRC | 16:17 | |
| *** eghobo has quit IRC | 16:19 | |
| *** lucasagomes_ has joined #openstack-ironic | 16:21 | |
| *** lucasagomes_ has quit IRC | 16:22 | |
| *** lucasagomes has quit IRC | 16:22 | |
| *** romcheg has quit IRC | 16:25 | |
| *** ifarkas has quit IRC | 16:27 | |
| *** athomas_ has quit IRC | 16:27 | |
| *** lsmola__ has quit IRC | 16:30 | |
| *** Nisha has joined #openstack-ironic | 16:34 | |
| Shrews | devananda: submitted a patch to tempest for that. maybe we don't have to think about the whole migration thing for a while longer yet :) | 16:35 |
| * Shrews in dire need of lunch. bbiab | 16:35 | |
| NobodyCam | enjoy Shrews :) | 16:35 |
| JoshNang | NobodyCam: yup! | 16:36 |
| NobodyCam | hey JoshNang wanted to see if you saw rloo's comment on the swift spec. looks like we lost a few things | 16:37 |
| NobodyCam | :-p | 16:37 |
| JoshNang | NobodyCam: yeah :/ must have gotten my local version out of sync somehow. going to fix it in about 5 minutes. | 16:38 |
| NobodyCam | :) woot :) ty | 16:38 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Allow noauth for Neutron https://review.openstack.org/98922 | 16:39 |
| *** amitpp has joined #openstack-ironic | 16:39 | |
| *** stendulker has quit IRC | 16:40 | |
| *** linggao has quit IRC | 16:40 | |
| *** ramineni has quit IRC | 16:41 | |
| *** romcheg has joined #openstack-ironic | 16:43 | |
| *** dhellmann_ is now known as dhellmann | 16:44 | |
| *** dwalleck has quit IRC | 16:49 | |
| *** Alexei_9871 has quit IRC | 16:52 | |
| *** derekh_ has quit IRC | 16:58 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-specs: Swift Temporary URLs Spec https://review.openstack.org/98904 | 17:03 |
| JayF | JoshNang: did you patch the holes in your spec with that patch? :P | 17:03 |
| JoshNang | NobodyCam: ^ that should be good. FWIW, when setting up Glance to use Swift, you have to specify a single user and container for the images to go in, so it doesn't seem like it would limit you to use the config options | 17:04 |
| devananda | Shrews: awesome, ty | 17:04 |
| JoshNang | JayF: :D | 17:04 |
| *** Nisha has quit IRC | 17:11 | |
| NobodyCam | quick walkies... brb | 17:12 |
| comstud | devananda: Have a few minutes to talk about a bug/race condition? :) | 17:13 |
| comstud | affects the virt driver | 17:14 |
| devananda | comstud: shoot | 17:16 |
| comstud | devananda: https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L626-L630 | 17:16 |
| *** overlayer has joined #openstack-ironic | 17:16 | |
| comstud | devananda: So, remember this conversation about multiple nova-computes and same host... and the synchronization on instance_uuid ? :) | 17:16 |
| comstud | Imagine this case... a build on one compute... a delete right after on another. | 17:17 |
| comstud | the delete skips the unprovision because the state is like 'building' or 'deploying' or whatever due to the code above | 17:17 |
| comstud | 1) These checks are racey no matter what | 17:18 |
| comstud | 2) ironic seems to allow you to unset instance_uuid on a node... while there's a target_provision_state | 17:18 |
| comstud | So, what happens here is that instance ends up building and going active in ironic | 17:18 |
| comstud | but it has no instance_uuid associated... | 17:18 |
| comstud | There's somewhat of a separate bug here... in that the scheduler will happily try to schedule to this node again :) | 17:19 |
| comstud | because of: https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L200 | 17:19 |
| comstud | doesn't also check that provision_state is NOSTATE | 17:19 |
| *** blamar has joined #openstack-ironic | 17:20 | |
| comstud | So, it's easy to fix the scheduler to not schedule to these "broken" nodes... however, we need to solve this race properly, also | 17:21 |
| devananda | huh | 17:21 |
| comstud | yeah, ask questions... that was a lot at once | 17:21 |
| comstud | :) | 17:21 |
| devananda | I thought there was something preventing deletion when node.provision_state NOT IN that list | 17:21 |
| devananda | but dont see it now | 17:21 |
| comstud | yeah, so, i'm seeing | 17:22 |
| comstud | https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L626-L630 | 17:22 |
| comstud | unprovision not called because state is 'deploying' or whatever | 17:22 |
| devananda | right | 17:22 |
| comstud | and _cleanup_deploy() happily removes the instance_uuid | 17:22 |
| comstud | with no error. | 17:22 |
| comstud | at least, that seems to be what I'm seeing :) | 17:23 |
| comstud | we certainly end up with a node in 'active' with no instance_uuid associated. | 17:23 |
| comstud | I almost feel like because nova's state checks are always going to be racey... that it always tries to unprovision. | 17:23 |
| devananda | comstud: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L685-L689 | 17:23 |
| devananda | so it doesn't remove the instance_uuid | 17:23 |
| devananda | that errors | 17:23 |
| devananda | that should prevent nova from unprovisioning | 17:24 |
| comstud | hrm. | 17:24 |
| jroll | well, that could race... if it goes to active between the state check and the api call | 17:24 |
| devananda | however it might be doing rather poor job and just breaking nova's instance at that point | 17:24 |
| jroll | but that's a much less likely situation | 17:24 |
| comstud | actually.... yeah | 17:24 |
| devananda | jroll: true | 17:24 |
| comstud | nevermind, let me back up | 17:24 |
| comstud | I forgot | 17:24 |
| devananda | comstud: :) | 17:24 |
| comstud | I do see nova retrying a lot during the delete :) | 17:25 |
| comstud | And then it succeeds | 17:25 |
| *** Nisha has joined #openstack-ironic | 17:25 | |
| jroll | a ha | 17:25 |
| *** dwalleck has joined #openstack-ironic | 17:25 | |
| *** dwalleck has quit IRC | 17:25 | |
| devananda | comstud: question i've had for a long time on this topic, actually | 17:25 |
| comstud | And it's the ironicclient code itself that is retrying | 17:25 |
| devananda | if a nova delete fails to remove the instance from the hypervisor | 17:25 |
| devananda | it leaves the instance in state DELETING, ERROR | 17:26 |
| *** dwalleck has joined #openstack-ironic | 17:26 | |
| devananda | does nova ever retry, aside from during _init_host() ? | 17:26 |
| devananda | evacuate-deleted_instances is the only place I've seen that action happen from | 17:26 |
| comstud | It now leaves it in ERROR, None (vm_state, task_state) | 17:26 |
| devananda | but i may have missed something | 17:26 |
| comstud | ie, it reversts DELETING to None | 17:26 |
| devananda | ok | 17:26 |
| comstud | And no, nova will not retry... outside of _init_host() | 17:26 |
| devananda | but it doesn't allow retrying the delete | 17:26 |
| devananda | ugh | 17:26 |
| comstud | It allows it now, yes | 17:26 |
| jroll | the user can retry | 17:26 |
| comstud | if you manually delete again | 17:26 |
| comstud | there was a bug there... | 17:26 |
| comstud | where task_state was being left | 17:27 |
| devananda | roughly when did that change? | 17:27 |
| comstud | and that prevented a user from retrying | 17:27 |
| comstud | last week or 2 | 17:27 |
| devananda | cause that's awesome. been bugging me for about a year now | 17:27 |
| jroll | heh | 17:27 |
| comstud | yes, it's been a bug since July/Aug 2013 | 17:27 |
| comstud | er, maybe i'm confusing bugs | 17:27 |
| comstud | but yes, it's been broken for a while :) | 17:27 |
| devananda | seems like that will help tripleo a lot, too | 17:27 |
| comstud | and infra... getting stuck instances | 17:28 |
| devananda | yea | 17:28 |
| devananda | sweet | 17:28 |
| comstud | that's where it came up recently | 17:28 |
| comstud | although I'd found this while debugging OnMetal a couple months ago | 17:28 |
| comstud | heh. | 17:28 |
| comstud | anyway :) | 17:28 |
| devananda | so. nova tries to delete the instance. ironic-client retries a bunch | 17:28 |
| comstud | yeah, then it succeeds | 17:28 |
| devananda | if that succeeds - great. things work then, right? | 17:28 |
| jroll | no | 17:28 |
| comstud | no | 17:28 |
| devananda | oh | 17:28 |
| comstud | somehow ironic still ends up with 'active' with no instance_uuid | 17:29 |
| jroll | unprovision_instance gets skipped | 17:29 |
| comstud | but it's gone from nova | 17:29 |
| *** bandicot has joined #openstack-ironic | 17:29 | |
| comstud | yeah | 17:29 |
| devananda | ooh right | 17:29 |
| jroll | the retry is on _cleanup_whateveritscalled | 17:29 |
| comstud | yes | 17:29 |
| comstud | So | 17:29 |
| devananda | the client retry is essentially increasing the window between those checks | 17:29 |
| comstud | I kinda think nova should _always_ call unprovision. | 17:29 |
| devananda | ugh | 17:29 |
| comstud | and then... we deal with the races with that | 17:29 |
| jroll | +1 | 17:29 |
| jroll | ironic should have that logic | 17:29 |
| comstud | Like... if it's already unprovisioned... we just say 'okay cool' | 17:29 |
| comstud | Otherwise perhaps nova just retries... on *that* call. | 17:30 |
| devananda | hmm | 17:30 |
| comstud | which will finally succeed... when the build is done on ironic side | 17:30 |
| comstud | These state checks in nova are racey | 17:31 |
| comstud | because once we call ironic api and get them back, they are no longer reality. | 17:31 |
| comstud | we can assume a certain order, perhaps... and i think that's what this tries to do | 17:31 |
| comstud | except it's missing that an instance can actually be in process of being built. | 17:31 |
| devananda | https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L225-L241 | 17:32 |
| comstud | another interesting solution... | 17:32 |
| comstud | is that we could make ironic check to see if an instance_uuid is still assigned | 17:32 |
| comstud | at the end of provisioning | 17:32 |
| comstud | and automatically teardown if it's gone | 17:32 |
| devananda | comstud: well. that's not possible today. | 17:33 |
| comstud | But there's still room for race there without making nova always call unprovision | 17:33 |
| comstud | nod | 17:33 |
| devananda | comstud: the lock is released when provisioning is completed. So it's not possible for the instance_uuid to change *during* provisioning | 17:33 |
| devananda | it could happen immediately after prov finishes (which is what you're seeing, i bet) | 17:33 |
| comstud | I don't particularly like that option either... was just throwing it out there. | 17:33 |
| comstud | yeah, net result is that i think we're seeing nova skip unprovision | 17:34 |
| devananda | it's an interesting idea, but the way we use locks prevents it | 17:34 |
| comstud | tries to unset instnace_uuid | 17:34 |
| devananda | right | 17:34 |
| comstud | retries for a long time | 17:34 |
| comstud | and then it succeeds | 17:34 |
| devananda | i can see how this code can easily result in that | 17:34 |
| comstud | maybe the non-atomic operation of unprovision and unsetting instance_uuid may be the real problem | 17:34 |
| comstud | maybe that should be atomic | 17:34 |
| comstud | from nova's perspective | 17:35 |
| devananda | not skipping unprovision and retrying earlier seems much better | 17:35 |
| comstud | or like... when we unset instance_uuid we make sure to unprovision | 17:35 |
| devananda | right | 17:35 |
| comstud | yeah | 17:35 |
| devananda | from ironic's side, they aren't atomic | 17:35 |
| devananda | one is state, one is metadata | 17:35 |
| comstud | yeah | 17:35 |
| comstud | So anyway, I need to file a bug here... but before I do, wanted to walk through the issue | 17:35 |
| devananda | you could change instance_info and instance_uuid in one PATCH call | 17:35 |
| devananda | but need to use PUT for the /state/provision resource to initiate that action | 17:36 |
| devananda | so here's another solution | 17:36 |
| devananda | https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L563 | 17:37 |
| devananda | retry that line if it gets a 409 CONFLICT | 17:37 |
| devananda | which would be coming only from here: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L238 | 17:37 |
| comstud | if we make destroy() always call _unprovision(), yes | 17:38 |
| devananda | right | 17:38 |
| devananda | that's what i mean | 17:38 |
| *** bandicot is now known as bensonmiller | 17:38 | |
| comstud | I think the client itself will actually retry | 17:38 |
| comstud | and we may not need to do anything here | 17:38 |
| comstud | the client itself is what is retrying for unsetting instance_uuid, at least | 17:39 |
| *** blamar has quit IRC | 17:39 | |
| devananda | ah, so it does -- https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/client_wrapper.py#L83 | 17:39 |
| comstud | so I think it's retrying on 409s ? | 17:39 |
| comstud | yeah | 17:39 |
| devananda | ugh. we use Conflict too much :( | 17:39 |
| * devananda is increasingly frustrated with the API | 17:40 | |
| comstud | hehe | 17:40 |
| comstud | I have another API thing to talk to you about :) | 17:40 |
| comstud | after this | 17:40 |
| devananda | :) | 17:40 |
| comstud | so.. we think just making nova always call unprovision is the fix? | 17:40 |
| comstud | i mean, destroy() always call _unprovision() | 17:41 |
| devananda | comstud: ++ | 17:41 |
| comstud | Ok | 17:41 |
| comstud | Related... | 17:41 |
| comstud | In case something else like this comes up.. | 17:41 |
| comstud | https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L200 | 17:41 |
| comstud | I think this should be: | 17:41 |
| comstud | if node.instance_uuid or node.provision_state != ironic_states.NOSTATE: | 17:42 |
| comstud | just as a safety net so the scheduler doesn't try to pick nodes that are in some state but don't have an instance_uuid | 17:42 |
| comstud | (it really should only pick nodes that are in NOSTATE and have no instance_uuid) | 17:42 |
| comstud | This is probably not needed with the "real fix", however I think that makes this check technically more accurate | 17:43 |
| devananda | comstud: see _node_resources_unavailable above | 17:43 |
| devananda | perhaps just adding another condition there is sufficient? | 17:43 |
| comstud | Yeah, I was going to put it there, but... | 17:44 |
| comstud | that makes the node show that it has 0 resources | 17:44 |
| devananda | oryou think the difference between | 17:44 |
| comstud | vs showing that the node is used | 17:44 |
| devananda | "node has 0 resources" and "all resources used" is important here | 17:44 |
| comstud | I'm not sure it's *that* important. | 17:44 |
| devananda | how will that affect or be exposed by nova? | 17:44 |
| comstud | The node having 0 resources might better indiciate there's a problem | 17:44 |
| devananda | right | 17:44 |
| comstud | I think it has the same net effect | 17:44 |
| devananda | rather than showing a node fully utilized with no instance | 17:44 |
| *** amitpp has quit IRC | 17:45 | |
| comstud | 0 resources should also translate to 0 free | 17:45 |
| comstud | although | 17:45 |
| comstud | We do have a node that is reporting -32768M of ram free :) | 17:45 |
| devananda | hah | 17:45 |
| comstud | which is odd.. | 17:45 |
| devananda | is it still getting instances scheduled on it? | 17:45 |
| comstud | i haven't looked into it | 17:46 |
| devananda | there's definitely some races in nova that lead to negative resources | 17:46 |
| comstud | just an oddity i noticed this morning | 17:46 |
| comstud | or someone noticed | 17:46 |
| devananda | been that way for a few years | 17:46 |
| comstud | okay, so I think I'll do what you suggest and put it in _node_resources_unavailable() | 17:47 |
| comstud | i'ma file a bug and push up a patch | 17:47 |
| devananda | cool | 17:47 |
| comstud | I assume you want this into the nova review before it lands, also | 17:48 |
| devananda | comstud: w.r.t. the nova patch sets, a few questions | 17:48 |
| comstud | ok | 17:48 |
| comstud | shoot. | 17:48 |
| devananda | IIRC mikal's plan was to -2 the head of the ironic patch set | 17:48 |
| devananda | but he hasn't, and alaski +A'd it | 17:48 |
| devananda | so ... ?? | 17:48 |
| comstud | hah | 17:48 |
| jroll | are you complaining? :) | 17:48 |
| comstud | I can add a -2 real quick if it's not landed yet | 17:48 |
| devananda | also, which is better from the nova review POV - continue to carry/merge fixes as I did this week, or wait until it lands in nova, or ? | 17:49 |
| devananda | jroll: there were reasons we had agreed on that process ... | 17:49 |
| comstud | devananda: well, jenkins failed on it anyway :) | 17:49 |
| jroll | devananda: mostly kidding | 17:50 |
| comstud | so it didn't merge | 17:50 |
| devananda | jroll: namely, being able to split the driver into patch sets that don't pass tests | 17:50 |
| devananda | comstud: hah, cool | 17:50 |
| devananda | comstud: if the plan changed, that's fine | 17:50 |
| comstud | i find it funny when jenkins fails on adding an empty directory | 17:50 |
| comstud | or file | 17:50 |
| devananda | failed five times | 17:50 |
| devananda | on adding an empty file | 17:50 |
| devananda | :) | 17:50 |
| comstud | even better | 17:50 |
| devananda | look at the retries! | 17:50 |
| devananda | <sadface> | 17:50 |
| comstud | I honestly didn't pay attention to what was said about merging these | 17:51 |
| comstud | I mean, the spec is approved | 17:51 |
| comstud | so that's all I looked at | 17:51 |
| devananda | comstud: I suspect that's generally the case | 17:51 |
| devananda | i'll poke mikal when he's online later about it | 17:51 |
| devananda | comstud: the rationale was: split the driver code into small bits which individually don't pass any tests, but are easy to review | 17:52 |
| devananda | comstud: -2 the tip, and watch the tail for test results | 17:52 |
| comstud | ok, i set this to WF -1 | 17:52 |
| comstud | for now | 17:52 |
| *** aswadr has quit IRC | 17:52 | |
| comstud | Anyone should be able to un-WF -1 it | 17:52 |
| devananda | comstud: then have someone in nova do the squash-and-resubmit when everything's approved | 17:52 |
| comstud | ah | 17:52 |
| *** dkehn_ has joined #openstack-ironic | 17:52 | |
| Nisha | jroll: JayF hi | 17:53 |
| devananda | because https://review.openstack.org/#/c/103167/ is 2500+ lines | 17:53 |
| jroll | hi, Nisha | 17:53 |
| Nisha | jroll, one query | 17:53 |
| * jroll listens | 17:54 | |
| Nisha | jroll: When IPA detects the link state of NIC, the OS is already there on the node (Atleast mini kernel)? | 17:54 |
| Nisha | jroll: Can IPA detect the link state of the NIC without any OS on teh BM? | 17:54 |
| Nisha | jroll: if yes, how? | 17:55 |
| jroll | Nisha: not sure what you mean. why would IPA detect the link state? | 17:55 |
| *** dkehnx has quit IRC | 17:56 | |
| Nisha | jroll: You mean IPA doesnt need to know whether the NIC is connected or not? | 17:56 |
| jroll | Nisha: why would it? | 17:56 |
| jroll | Nisha: other than to make sure it can talk to ironic... | 17:56 |
| jroll | Nisha: but that can be detected by sending an HTTP request | 17:56 |
| jroll | Nisha: why would anything need to detect NIC link state? | 17:56 |
| Nisha | jroll: by broadcasting, correct? | 17:56 |
| jroll | what? | 17:57 |
| * jroll is confused | 17:57 | |
| jroll | Nisha: what are you trying to do? what is the end goal? | 17:57 |
| Nisha | jroll: This is regarding https://review.openstack.org/#/c/100951/16 | 17:57 |
| devananda | Nisha: this question "Can IPA detect the link state of the NIC without any OS on teh BM?" does not make sense | 17:57 |
| devananda | Nisha: ironic-python-agent is a set of python modules that runs inside an operating system on the bare metal node | 17:58 |
| jroll | none of this makes sense | 17:58 |
| comstud | devananda: https://bugs.launchpad.net/ironic/+bug/1337461 assign importance to that if you want | 17:58 |
| comstud | but it's in progress here | 17:58 |
| jroll | why would ironic care if a NIC has a cable attached? | 17:58 |
| devananda | Nisha: IPA will implement a deploy interface, not power or management interfaces. | 17:58 |
| jroll | Nisha: what line should I be looking at? | 17:58 |
| devananda | Nisha: so it will be able to use eg. IPMITool or iLO interfaces for those | 17:58 |
| Nisha | jroll: see --create-ports option | 17:59 |
| Nisha | it has options [connected | all] | 17:59 |
| devananda | jroll: so the context from Nisha's side is autodiscovery | 17:59 |
| jroll | I don't see why 'connected' si useful | 17:59 |
| jroll | s/si/is/ | 18:00 |
| NobodyCam | jroll: for hw discovery | 18:00 |
| devananda | jroll: could we use IPA deploy + PDU power to do inventory discovery, with absolutely no out-of-band mgmt cards? -- I think so. | 18:00 |
| jroll | Nisha: IPA *could* detect link state, if we wrote code to do so. does that help? | 18:00 |
| jroll | devananda: I agree | 18:00 |
| comstud | Hm | 18:00 |
| jroll | I still don't see why only adding ports in ironic for connected NICs is useful | 18:01 |
| devananda | jroll: knowing link state of a NIC in general will be important as we add more interaction with SDN controllers (eg, neutron OVS) | 18:01 |
| jroll | but that might just be me | 18:01 |
| comstud | What error does ironic return if we try to unprovision a node that's already unprovisioned? | 18:01 |
| devananda | jroll: you mean "unconnected NICs" | 18:01 |
| * comstud looks | 18:01 | |
| jroll | devananda: I mean "only adding ports in ironic for connected NICs" | 18:01 |
| devananda | comstud: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L228 | 18:01 |
| jroll | devananda: or "not adding ports for unconnected NICs" | 18:01 |
| devananda | jroll: oh. gotcha. I agree. we should add all of htem | 18:01 |
| jroll | right | 18:01 |
| comstud | thanks | 18:02 |
| devananda | jroll: but stashing metadata saying "this nic is not cabled" could be helpful. I think that was the question? | 18:02 |
| devananda | NobodyCam: ^ did I get that right? | 18:02 |
| NobodyCam | reading now | 18:02 |
| jroll | devananda: that's fine. the question was about the --create-ports=[connected | all] in the spec | 18:02 |
| devananda | oh | 18:02 |
| NobodyCam | ya | 18:02 |
| jroll | which I don't think should be an option unless there's a great use case | 18:03 |
| jroll | anyway | 18:03 |
| Nisha | devananda: yes | 18:03 |
| jroll | Nisha: did I answer your question? | 18:03 |
| devananda | why would we NOT create a port for a NIC that isn't cabled, but is in the machine and might be cabled later? | 18:03 |
| Nisha | jroll: yes | 18:03 |
| jroll | Nisha: :) | 18:03 |
| devananda | plugging in a cable later shouldn't require RE-discovering the hardware | 18:03 |
| devananda | the hardware hasnt changed | 18:03 |
| devananda | only link state | 18:04 |
| jroll | agree | 18:04 |
| devananda | exposing a means (and perhaps then automating it) to disable NICs would be good | 18:04 |
| Nisha | devananda: ok , then we can have an option of link state which can be updated | 18:04 |
| devananda | eg, tell Neutron "this NIC is good, give it an IP, that NIC should be offline, make sure it doesn't get on the network" | 18:04 |
| Nisha | devananda: not option | 18:04 |
| Nisha | a sub-field in inventory field | 18:05 |
| devananda | that'd be a good feature IMO | 18:05 |
| devananda | (just not in Juno... we're already swamped :) ) | 18:05 |
| jroll | all this discovery stuff is sooo close to making ironic an inventory system | 18:05 |
| devananda | jroll: i know :( | 18:05 |
| jroll | burn it all | 18:05 |
| devananda | it's a slippery slope and i've been trying to find the middle ground | 18:06 |
| jroll | auto-registration, don't store any info :) | 18:06 |
| jroll | (way easier said then done) | 18:06 |
| jroll | s/then/than/ | 18:06 |
| devananda | auto-registration, store opaque blob that an external inventory system can pull at a later date | 18:06 |
| Nisha | NobodyCam: so is it fine if i remove the options from --create-ports | 18:06 |
| NobodyCam | Nisha: yes | 18:06 |
| jroll | yeah, but you know that will get abused | 18:07 |
| jroll | Nisha: yes, please do | 18:07 |
| devananda | because real-time request to gather invetory data is impractical/impossible | 18:07 |
| devananda | jroll: it already is | 18:07 |
| jroll | right | 18:07 |
| devananda | jroll: the requests to search inside that blob are just the beginning -- and i'm being clear in saing "no" even to that | 18:07 |
| devananda | well, i think i'm being clear | 18:07 |
| devananda | :( | 18:07 |
| jroll | you've been clear, just ignored | 18:08 |
| devananda | awesome | 18:08 |
| *** coolsvap is now known as coolsvap|afk | 18:08 | |
| Nisha | devananda: since we are discussing this, could i request you to review the spec https://review.openstack.org/#/c/100951 again ? :) | 18:08 |
| jroll | devananda: (or so it seems) | 18:08 |
| *** rloo has quit IRC | 18:10 | |
| comstud | devananda: https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L570 | 18:10 |
| comstud | ^ does that actually work? It seems like a BadRequest is what will be returned | 18:10 |
| Nisha | comstud: just remove #L570 from the URL, it will work | 18:11 |
| devananda | comstud: i was wondering about that ... I suspect the #TODO has been done and that can be removed now? | 18:11 |
| *** Nisha has left #openstack-ironic | 18:11 | |
| comstud | I see no reference to InstanceDeployFailure in the API at all | 18:11 |
| comstud | https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L228 | 18:12 |
| NobodyCam | do we call update_port_dhcp_opts for every port attached to a node? | 18:12 |
| comstud | devananda: I'm thinking: http://paste.openstack.org/show/85443/ | 18:14 |
| devananda | NobodyCam: no - https://github.com/openstack/ironic/blob/master/ironic/common/neutron.py#L149 | 18:14 |
| devananda | NobodyCam: only for ports with a VIF | 18:15 |
| devananda | NobodyCam: which mayu or may not be all of them | 18:15 |
| devananda | comstud: i hate the string matching on e | 18:16 |
| NobodyCam | my concern with the connected status was that dhcp would be attached to a mac address that had no cable on it | 18:17 |
| devananda | comstud: what happens if that exception is translated? | 18:17 |
| comstud | FWIW, I also hate the string matching :) | 18:17 |
| comstud | I'm wondering if just BadRequest will suffice on its own | 18:17 |
| *** foexle has joined #openstack-ironic | 18:17 | |
| comstud | I don't think 400 is returned for anything else | 18:17 |
| comstud | in here | 18:17 |
| comstud | worst case... | 18:17 |
| comstud | If we fall through for all BadRequest.. | 18:18 |
| comstud | nova is going to poll for target_provision_state=None anyway | 18:18 |
| comstud | and error there | 18:18 |
| comstud | so I think maybe just a 'pass' here will be fine | 18:18 |
| devananda | k | 18:19 |
| * devananda moves to -qa to talk tempest | 18:19 | |
| comstud | devananda: thnx, patch shortly. | 18:20 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Adding support for decommissioning https://review.openstack.org/104379 | 18:22 |
| *** Mikhail_D_ltp has joined #openstack-ironic | 18:22 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 18:23 |
| *** Nisha has joined #openstack-ironic | 18:23 | |
| *** Mikhail_D_ltp has left #openstack-ironic | 18:31 | |
| *** dwalleck has quit IRC | 18:33 | |
| *** romcheg has quit IRC | 18:35 | |
| *** romcheg has joined #openstack-ironic | 18:35 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 18:37 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Adding support for decommissioning https://review.openstack.org/104379 | 18:39 |
| *** eghobo has joined #openstack-ironic | 18:39 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 18:40 |
| *** romcheg1 has joined #openstack-ironic | 18:41 | |
| *** overlayer has quit IRC | 18:42 | |
| *** romcheg has quit IRC | 18:43 | |
| NobodyCam | brb | 18:47 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 18:48 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Adding support for decommissioning https://review.openstack.org/104379 | 18:52 |
| *** Nisha has quit IRC | 18:54 | |
| NobodyCam | devananda: I've miss places the link for the spec's review team, happen to have that off hand? and would you mind if I added it to https://wiki.openstack.org/wiki/Ironic/Specs_Process | 18:56 |
| devananda | NobodyCam: it's a gerrit group | 18:57 |
| *** foexle has quit IRC | 18:57 | |
| *** pelix has quit IRC | 19:02 | |
| *** ccrouch has quit IRC | 19:03 | |
| openstackgerrit | Chris Behrens proposed a change to openstack/ironic: Fix build + delete races https://review.openstack.org/104649 | 19:03 |
| comstud | devananda: ^ hopefully that covers that | 19:04 |
| *** rloo has joined #openstack-ironic | 19:16 | |
| NobodyCam | devananda: found it | 19:16 |
| NobodyCam | :) | 19:16 |
| * NobodyCam switches to internal mode to take Standards of Business training | 19:18 | |
| *** rloo has quit IRC | 19:21 | |
| *** rloo has joined #openstack-ironic | 19:23 | |
| matty_dubs | Is that like the ethics training we have to take? A once-a-year training video that mostly just tells you stuff you already knew, like that bribing people or trading on insider information is wrong? | 19:24 |
| NobodyCam | yep | 19:25 |
| NobodyCam | :-p | 19:25 |
| devananda | jroll: who is working on IPA support in devstack? | 19:30 |
| jroll | devananda: literally trying to get devstack to run right now | 19:30 |
| devananda | ah | 19:30 |
| jroll | just came back from lunch to "Cannot open network namespace "qdhcp-e23ca5f6-d06e-4015-84f4-d1d4a164629f": No such file or directory" | 19:31 |
| jroll | and " ovs-vsctl: tag=: argument does not end in "=" followed by a value." | 19:31 |
| jroll | (I hate devstack) | 19:32 |
| jroll | oh, maybe those are non-critical errors | 19:34 |
| jroll | got a node booting | 19:35 |
| * jroll adds IPA to this | 19:35 | |
| jroll | how do I grab a console with virsh? | 19:37 |
| NobodyCam | virsh console <name> | 19:38 |
| jroll | thought so... error: internal error: character device <null> is not using a PTY | 19:38 |
| jroll | :( | 19:39 |
| NobodyCam | thats a vm config issues | 19:39 |
| jroll | I... didn't configure a vm | 19:39 |
| jroll | afaik | 19:39 |
| NobodyCam | need to setup pty devicw | 19:39 |
| jroll | I'm just following http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html | 19:39 |
| * jroll has no idea what he's doing | 19:39 | |
| jroll | brb | 19:39 |
| *** enikanorov_ has joined #openstack-ironic | 19:43 | |
| matty_dubs | devananda: BTW, thanks for the review on the NetApp spec. Very helpful in figuring out the right way to proceed with that code. | 19:45 |
| matty_dubs | I haven't replied to it since I'm bugging Lucas about whether it makes sense to pursue any of that stuff through Ironic, or if it should be an entirely separate thing. | 19:46 |
| *** enikanorov__ has quit IRC | 19:47 | |
| matty_dubs | (But the latter seems likely.) | 19:47 |
| devananda | matty_dubs: if you start provisioning OSs on the netapp and treat them as disposable "cloud instances" -- sure | 19:49 |
| devananda | matty_dubs: part of this is also the paradigm shift from "it's physical so it's a snowflake" to "it's just an instance in a cloud, regardless of whether it's a VM or not" | 19:49 |
| matty_dubs | Yeah. What I struggled with is that you can do some OS management, though mostly updates. | 19:50 |
| matty_dubs | So it's conceivable that this would be very useful if you wanted to, say, reprovision one of these for a different client | 19:50 |
| matty_dubs | But I don't think that's a common operation. | 19:50 |
| devananda | matty_dubs: if you dont want to manage (deploy, delete) the instance (of an operating system on your netapp filer) through Nova, perhaps it's not a good fit | 19:51 |
| *** blamar has joined #openstack-ironic | 19:51 | |
| *** Penick has joined #openstack-ironic | 19:51 | |
| matty_dubs | That's a good way to frame it. We almost certainly don't want to do that. | 19:52 |
| devananda | matty_dubs: :) | 19:53 |
| devananda | so how can we improve the messaging that | 19:53 |
| devananda | ironic is not a general tool to install an operating system | 19:53 |
| devananda | it's a tool for managing hardware *just like it's in a cloud* | 19:53 |
| devananda | cause, well, it is | 19:53 |
| matty_dubs | Hehe | 19:53 |
| matty_dubs | Well, I think the problem is that I started more bottom-up here. We do power control + deploys + console in the code, and hey, we need power control + console and maybe a little deploying, so Ironic! | 19:54 |
| jroll | NobodyCam: hey, looks like this vm was started with: -chardev file,id=charserial0,path=/opt/stack/ironic-bm-logs/baremetalbrbm_2_console.log -chardev pty,id=charserial1 | 19:54 |
| jroll | NobodyCam: so uh, even if pty isn't working, there should be logs in those files, no? | 19:55 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Adding support for decommissioning https://review.openstack.org/104379 | 19:55 |
| *** Penick has quit IRC | 19:57 | |
| NobodyCam | ??? I would assume so? | 20:00 |
| jroll | yeah :/ | 20:01 |
| jroll | nothing there | 20:01 |
| jroll | hrmmmmmmm | 20:02 |
| *** foexle has joined #openstack-ironic | 20:03 | |
| *** blamar has quit IRC | 20:03 | |
| *** dkehn__ has joined #openstack-ironic | 20:04 | |
| *** dkehn_ has quit IRC | 20:07 | |
| *** foexle has quit IRC | 20:10 | |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Run make with sudo when building https://review.openstack.org/104361 | 20:10 |
| *** foexle has joined #openstack-ironic | 20:10 | |
| JayF | jroll: ^ | 20:10 |
| jroll | nice | 20:11 |
| JayF | so the question is if the old or new version of post job ran | 20:11 |
| JayF | should be new | 20:11 |
| jroll | it just runs the script | 20:11 |
| jroll | the latest version of the script | 20:11 |
| JayF | https://jenkins02.openstack.org/job/ironic-python-agent-buildimage-coreos/2/console | 20:12 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission interfaces https://review.openstack.org/102953 | 20:13 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission in conductor https://review.openstack.org/102984 | 20:14 |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Install requirements for coreos-oem-inject.py https://review.openstack.org/104672 | 20:17 |
| JayF | JoshNang: jroll ^^ | 20:17 |
| jroll | hmm | 20:18 |
| jroll | what was that --external flag we were using at some point? | 20:18 |
| jroll | was that for non-OS requirements | 20:18 |
| jroll | ? | 20:18 |
| JayF | it was required because '-e' in requirements.txt is not allowed without allow_external | 20:18 |
| jroll | ah | 20:19 |
| jroll | approved | 20:19 |
| *** dkehn__ is now known as dkehnx | 20:19 | |
| JayF | jroll: I'm going to file a bug (low-hanging-fruit) to migrate the script off plumbum | 20:19 |
| jroll | cool | 20:19 |
| jroll | kylestev: ^ :) | 20:19 |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Install requirements for coreos-oem-inject.py https://review.openstack.org/104672 | 20:21 |
| jroll | does everybody else's virsh logs just work? or is there configs I need to do | 20:23 |
| jroll | 'getting started with ironic' isn't very easy | 20:23 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission in conductor https://review.openstack.org/102984 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factor out TFTPImageCache https://review.openstack.org/100734 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factor out deploy info from PXE driver https://review.openstack.org/100735 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add methods to ipmitool driver https://review.openstack.org/100364 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission interfaces https://review.openstack.org/102953 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding swift temp url support https://review.openstack.org/81391 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add ironic-python-agent deploy driver https://review.openstack.org/101020 | 20:24 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 20:24 |
| jroll | whoa | 20:24 |
| *** ccrouch has joined #openstack-ironic | 20:25 | |
| JoshNang | rebase all the things! | 20:25 |
| JoshNang | (but seriously, didn't realize it would do that whole giant patch tree) | 20:26 |
| jroll | -R ftw | 20:26 |
| JayF | kylestev: https://bugs.launchpad.net/ironic/+bug/1337551 -- if you wanted a low hanging IPA bug :) | 20:26 |
| JoshNang | I was fixing a merge issue in the last patch on that chain | 20:27 |
| JayF | heh. | 20:27 |
| JayF | http://tarballs.openstack.org/ironic-python-agent/coreos/ipa-coreos.tar.gz/ipa-coreos.tar.gz | 20:27 |
| * devananda goes out in search of food | 20:27 | |
| JayF | looks like we have an infra fix, but it appears this should be our IPA CoreOS image, built fresh off openstack infra | 20:27 |
| devananda | jroll, JoshNang: i dont recall if we talked about this explicitly | 20:28 |
| *** jgrimm has quit IRC | 20:28 | |
| *** bensonmiller has quit IRC | 20:28 | |
| devananda | but I am going to (and should have a while ago) -2'd the IPA patch series | 20:28 |
| JayF | until the spec is approved? | 20:29 |
| jroll | yep | 20:29 |
| devananda | as a way to indicate which ones are ready for merging (like the ones at the start, which arent dependent on the spec | 20:29 |
| jroll | iirc | 20:29 |
| devananda | right | 20:29 |
| JayF | I assumed that it already was tbh :) | 20:29 |
| devananda | cool | 20:29 |
| devananda | i've been lax about actually doing that | 20:29 |
| devananda | well, busy | 20:29 |
| jroll | rloo: has been on it | 20:29 |
| jroll | but missed that one, I guess | 20:29 |
| rloo | jroll: which one did I miss? you guys have so many! | 20:30 |
| JayF | That uploaded ipa image looks golden | 20:30 |
| devananda | rloo: all of them :) | 20:30 |
| rloo | let me take a look. I thought I -2'd the 'base' patch. but maybe I missed or you snuck some new ones in ;) | 20:30 |
| jroll | rloo: :) | 20:30 |
| jroll | rloo: 101020 | 20:31 |
| devananda | right. i'm really running out of fuel now ... gotta find food | 20:31 |
| rloo | we should just approve the spec soon. (although i just added a bunch of comments to it.) | 20:31 |
| devananda | bbiah | 20:31 |
| devananda | rloo: or that :) | 20:31 |
| jroll | ^^ | 20:31 |
| * devananda is really away now | 20:31 | |
| JoshNang | rloo: yeah i saw those. on my todo list for today to fix up the spec | 20:31 |
| JoshNang | and thanks :) | 20:31 |
| *** bandicot has joined #openstack-ironic | 20:31 | |
| rloo | so 101020 depends on 81391 which I blocked. but 81391 is a diff blueprint so yeah, I should block 101020. thx for letting me know jroll! | 20:32 |
| jroll | no problem :) | 20:32 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic: Add get_host_uptime to nova driver https://review.openstack.org/104680 | 20:34 |
| jroll | Shrews: uhhh, how do we define 'host' in ironic? :) | 20:35 |
| Shrews | jroll: hrm, good question. i was assuming compute host, but that could be node i suppose. It's WIP anyway | 20:36 |
| jroll | yeah | 20:36 |
| jroll | just a thoughut | 20:37 |
| Shrews | jroll: i pretty much copied libvirt driver on it | 20:37 |
| jroll | Shrews: right... I get the idea. it's just a weird abstraction for ironic | 20:37 |
| jroll | also, shouldn't ever care how long an ironic compute node has been running | 20:38 |
| jroll | because it doesn't necessarily affect an ironic node | 20:38 |
| Shrews | "uptime of the hypervisor" .... | 20:38 |
| jroll | but like | 20:38 |
| jroll | there is no hypervisor | 20:39 |
| jroll | (maybe that's what you're trying to say) | 20:39 |
| Shrews | jroll: yeah, it's weird. i'll dig deeper later. i just wanted to get the code saved somewhere before i forget about it for the long weekend | 20:39 |
| jroll | ok :) | 20:39 |
| jroll | so, this doesn't work at all, as far I can tell: http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#deploying-ironic-with-devstack | 20:40 |
| jroll | boot fails | 20:40 |
| jroll | delete fails | 20:40 |
| jroll | everything is awesome | 20:40 |
| JoshNang | jroll: gah that rebase messed up the topics for the patches. | 20:40 |
| Shrews | jroll: sure it does | 20:40 |
| jroll | JoshNang: I've decided to stop caring about topics | 20:40 |
| Shrews | jroll: what os? | 20:40 |
| jroll | Shrews: not in my recent experience | 20:40 |
| JoshNang | ha k | 20:40 |
| jroll | 14.04 | 20:40 |
| Shrews | jroll: fail | 20:40 |
| jroll | which is what infra runs now | 20:40 |
| jroll | so it should work | 20:40 |
| Shrews | jroll: no problems on 12.04 | 20:41 |
| jroll | if this doesn't work on 14.04 I'm going to flip things | 20:41 |
| jroll | oh, wait. | 20:41 |
| Shrews | YES! | 20:41 |
| jroll | I just realized | 20:41 |
| jroll | ha | 20:41 |
| jroll | ignore me | 20:41 |
| jroll | this is debian testing | 20:41 |
| adam_g | jroll, IRONIC_VM_SPECS_RAM=512 thats outdated, set to 1024 | 20:41 |
| jroll | I switched boxes without thinking | 20:41 |
| jroll | adam_g: yeah, grabbed some things from dwalleck's patch | 20:41 |
| Shrews | /ignore jroll | 20:41 |
| jroll | including that | 20:41 |
| JayF | Ironic devs: http://tarballs.openstack.org/ironic-python-agent/coreos/ipa-coreos.tar.gz -- Infra is setup now to build a new tarball of the IPA CoreOS ramdisk on each commit. When conbined with http://git.openstack.org/cgit/openstack/ironic-python-agent/tree/imagebuild/coreos/README.rst that should get us closer to not-horribly-difficult testing of the agent and agent patches :) | 20:41 |
| jroll | thanks Shrews | 20:42 |
| JayF | jroll: ^ for great benefit devstack testing | 20:42 |
| jroll | JayF: (which I'm trying to do) | 20:42 |
| jroll | right | 20:42 |
| openstackgerrit | A change was merged to openstack/ironic: Allow noauth for Neutron https://review.openstack.org/98922 | 20:42 |
| Shrews | ok, after a 2+ hour IRC meeting, i'm going to officially start my long weekend by cracking open beers now. | 20:44 |
| Shrews | later gators | 20:44 |
| jroll | nice, have a good one Shrews | 20:46 |
| *** jdob has quit IRC | 20:55 | |
| ellenh | rloo: hi, are you around/available? | 21:00 |
| *** jbjohnso_ has quit IRC | 21:00 | |
| rloo | hi ellenh | 21:01 |
| ellenh | rloo: so, for the info logging patch, I just tried again to use the nicer import statement for _LI | 21:01 |
| ellenh | rloo: and it still breaks | 21:01 |
| rloo | ellenh: i wonder why it works for the non-ironic 'core' code but not for your changes. | 21:02 |
| rloo | ellenh: give me 15 minutes, and I'll look into it. | 21:02 |
| ellenh | rloo: cool, thanks. | 21:04 |
| *** foexle has quit IRC | 21:06 | |
| jroll | ellenh, rloo: see the last line of tox.ini | 21:12 |
| jroll | you could add an exception for ironic.openstack.common.gettextutils._LI, etc | 21:12 |
| jroll | the existing code works because it will skip the entire line, and the existing line is: from ironic.openstack.common.gettextutils import _, _LI | 21:13 |
| rloo | jroll: ahh. so ellenh just needs to add '_, ' to it too. | 21:13 |
| jroll | I mean, that's not the "right" way to do it, but that's the easy way :) | 21:14 |
| jroll | easy way being not adding exceptions for each _L* | 21:14 |
| ellenh | jroll: rloo that gives a “one import per line” complaint though | 21:14 |
| jroll | gah | 21:14 |
| jroll | I might be wrong then | 21:14 |
| jroll | idk | 21:14 |
| jroll | you could add to import_exceptions in tox.ini | 21:14 |
| rloo | ah, tox.ini excludes some directories like *openstack/common*, *ironic/nova*. that's why they worked there. | 21:16 |
| ellenh | jroll: rloo hmm. so, would it be better to block both info log patches and depend on a “change tox.ini” patch, change tox.ini in both patches, or just leave both as-is and deal with it later? | 21:17 |
| rloo | ellenh: it'd be better NOT to use the _L* stuff but I've been overruled here ;) | 21:18 |
| jroll | ellenh: I'd say the first | 21:18 |
| rloo | ellenh: I think updating tox.ini first would be better. | 21:18 |
| jroll | rloo: problem with making _LI happen in the log.info() calls would be that you can't translate it until you log it | 21:19 |
| jroll | which is not always ideal | 21:19 |
| jroll | because you want to translate before formatting | 21:19 |
| jroll | like, _LI('Node %(node)s blablabla') | 21:19 |
| jroll | does that make sense? I had this realization the other day | 21:20 |
| rloo | jroll: it just seems like we should be able to do better, maybe not as well but better. putting _LI, then if the log changes to warn, having to remember to change _LI etc. | 21:20 |
| jroll | rloo: sure, and I've always agreed with you until I noticed this | 21:20 |
| rloo | jroll: it just makes me sad. oh well. | 21:21 |
| jroll | indeed | 21:22 |
| ellenh | at least there’s a clear ruling, though | 21:23 |
| NobodyCam | bbiafm | 21:26 |
| jroll | wow, this totally works on ubuntu | 21:28 |
| jroll | :) | 21:28 |
| *** Haomeng|2 has joined #openstack-ironic | 21:41 | |
| *** Haomeng has quit IRC | 21:42 | |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic: Add gettextutils._L* to import_exceptions https://review.openstack.org/104696 | 21:48 |
| *** mrda-away is now known as mrda | 21:53 | |
| mrda | Morning Ironic! | 21:54 |
| *** ccrouch has quit IRC | 21:56 | |
| NobodyCam | morning mrda | 21:58 |
| mrda | \o | 21:59 |
| jroll | heya mrda | 22:00 |
| mrda | o/ | 22:03 |
| *** ccrouch has joined #openstack-ironic | 22:07 | |
| NobodyCam | devananda: when you get back could you ping me | 22:07 |
| *** hemna_ has quit IRC | 22:08 | |
| NobodyCam | any one played with barbican??? | 22:09 |
| JayF | NobodyCam: I haven't personally, but Jarret has demoed some of the bits to us internally before | 22:10 |
| JayF | NobodyCam: but honestly IDK how much of that is 'we want it to do this' vs 'it does this today' :) | 22:10 |
| NobodyCam | :) | 22:11 |
| NobodyCam | I was hopping we could off load "secure storage for IPMI creds" | 22:12 |
| JayF | ++++ | 22:13 |
| JayF | absolutely, and Barbican would be great for that | 22:13 |
| mrda | NobodyCam: Always a trade-off - adding a new dependency vs writing it yourself. | 22:13 |
| JayF | IDK if it's ready, but Jarret is a great guy and would probably like being asked at least | 22:13 |
| * mrda hasn't played with Barbican | 22:13 | |
| jroll | I think it's ready like ironic is ready | 22:15 |
| jroll | so, usable | 22:15 |
| mrda | so, awesome! :) | 22:15 |
| NobodyCam | we have a BP for it just need a spec.. | 22:15 |
| NobodyCam | oh and ofc code to do it:-p | 22:18 |
| NobodyCam | s/do it/use it/ | 22:19 |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic: Add more INFO logging to ironic/common https://review.openstack.org/99495 | 22:21 |
| * devananda is back | 22:27 | |
| devananda | NobodyCam: and someone to lead that code, the testing of it, write docs, etc... | 22:27 |
| NobodyCam | lol :) | 22:28 |
| NobodyCam | oh i'll brb | 22:28 |
| *** eghobo has quit IRC | 22:29 | |
| *** radsy has joined #openstack-ironic | 22:37 | |
| *** romcheg1 has quit IRC | 22:37 | |
| *** romcheg has joined #openstack-ironic | 22:39 | |
| *** Mikhail_D_ltp has quit IRC | 22:40 | |
| jroll | devananda: when putting up devstack patches for this stuff... should I add to the existing 3 nodes or take one of them for agent driver? | 22:47 |
| devananda | jroll: toggle | 22:47 |
| jroll | thanks | 22:47 |
| jroll | I might have something today | 22:47 |
| jroll | likely monday | 22:47 |
| devananda | jroll: so in the context of testing within devstack-gate, the direction we need to go is supporting the full set of tempest/api/compute tests | 22:48 |
| jroll | right, that's my eventual goal. first up is "random ironic dev can test" | 22:48 |
| devananda | jroll: which means having available min 3 (for serial) or 6 (for parallel) VMs for testing | 22:48 |
| jroll | which right now, is looking like 1 file changed, 11 insertions(+), 8 deletions(-) (and adding a couple of configs) | 22:49 |
| jroll | nod | 22:49 |
| devananda | jroll: we could have a jenkins job run on ironic's pipe that changes a flag (IRONIC_DRIVER=ipa_ssh) which triggers a bunch of other changes to set up the env for that | 22:49 |
| devananda | jroll: and then eg. pass that in from infra/config for that job | 22:50 |
| jroll | right | 22:50 |
| jroll | devananda: but I'd also like parallel testing | 22:50 |
| devananda | jroll: are you guys going to run this as third-party CI? or try to get it in d-g? | 22:50 |
| jroll | just d-g cause why not :) | 22:50 |
| devananda | jroll: ah, test both drivers at the same time? | 22:50 |
| jroll | both in parallel is the eventual goal, yes | 22:51 |
| jroll | because that means users can upgrade | 22:51 |
| jroll | etc | 22:51 |
| devananda | jroll: like enroll some nodes with each, then let nova pick them at random | 22:51 |
| jroll | (also, heterogeneous environments) | 22:51 |
| devananda | right | 22:51 |
| devananda | actually, that's neat | 22:51 |
| jroll | well, I'd say run each test twice in parallel | 22:51 |
| devananda | no | 22:51 |
| jroll | or... something | 22:51 |
| devananda | that's less neat | 22:51 |
| jroll | ha | 22:51 |
| jroll | random is fine | 22:51 |
| devananda | because it means twice as long to test, and doesn't really gain us much | 22:52 |
| jroll | well, I'd kind of like each driver gets each test at least once | 22:52 |
| devananda | so the tests aren't driver specific | 22:52 |
| jroll | to make sure e.g. we don't break agent tear_down and miss it | 22:52 |
| devananda | or even specific to Ironic | 22:52 |
| devananda | it's all the nova tests | 22:52 |
| jroll | right... | 22:52 |
| devananda | IOW, one run of tempest is ~ 70 deploys | 22:52 |
| jroll | wow | 22:53 |
| *** romcheg has quit IRC | 23:06 | |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic: Add gettextutils._L* to import_exceptions https://review.openstack.org/104696 | 23:07 |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic: Add more INFO logging to ironic/common https://review.openstack.org/99495 | 23:07 |
| *** matty_dubs is now known as matty_dubs|gone | 23:08 | |
| adam_g | devananda, would hard-coding the root=ramfs in the template possibly complicate things for other ramdisks? IPA? | 23:09 |
| jroll | adam_g: which template? | 23:09 |
| adam_g | actually, i guess its specific to the pxe driver | 23:10 |
| adam_g | jroll, https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe_config.template#L11 | 23:10 |
| jroll | ah yeah, we made our own template | 23:10 |
| devananda | that's what I thought | 23:10 |
| devananda | wtf? gertty is suddenly saying it has 750 changes to sync | 23:12 |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic: Add more INFO logging to ironic/conductor https://review.openstack.org/99744 | 23:13 |
| *** LiveOne has joined #openstack-ironic | 23:18 | |
| NobodyCam | devananda: you saw that https://review.openstack.org/#/c/104322 has two +2 but was not do it is location (our nova tree) | 23:19 |
| NobodyCam | gah | 23:19 |
| NobodyCam | s/was not do/was not landed do/ | 23:20 |
| devananda | NobodyCam: oh, hadn't seen. thanks | 23:22 |
| *** ccrouch has quit IRC | 23:25 | |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Ironic nova driver to cache ironic client calls https://review.openstack.org/102695 | 23:25 |
| *** max_lobur has quit IRC | 23:32 | |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/python-ironicclient: Make a few minor updates to node shell help strings https://review.openstack.org/102312 | 23:47 |
| devananda | rloo: if by chance you're still around, this spec https://review.openstack.org/#/c/102435/ now has 2 +2 and several +1, pending you taking another look at it, it could be approved | 23:48 |
| *** eghobo has joined #openstack-ironic | 23:52 | |
| *** derekh_ has joined #openstack-ironic | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!