| *** lucas-afk has quit IRC | 00:05 | |
| * NobodyCam wanders away | 00:05 | |
| GheRivero | jroll: ping | 00:11 |
|---|---|---|
| jroll | GheRivero: hey | 00:12 |
| GheRivero | on review 103685, how it will fail with old clients? | 00:14 |
| jroll | GheRivero: service.client.http_client.endpoint is the new thing, right? | 00:15 |
| jroll | GheRivero: does service.client.http_client.endpoint exist in older clients? | 00:15 |
| GheRivero | getattr(service.client, 'endpoint', service.client.http_client.endpoint) | 00:16 |
| GheRivero | no it doesn't | 00:16 |
| jroll | right | 00:16 |
| jroll | so if you call that | 00:17 |
| jroll | service.client.http_client.endpoint will be resolved before calling getattr | 00:17 |
| jroll | and fail | 00:17 |
| jroll | I believe | 00:17 |
| *** zigo has quit IRC | 00:17 | |
| jroll | yeah, just tested | 00:18 |
| GheRivero | ah. Ok. I got it now. | 00:19 |
| jroll | :) | 00:19 |
| jroll | sorry, I should have explained more in the comments | 00:19 |
| GheRivero | nah. It's ok | 00:19 |
| *** zigo has joined #openstack-ironic | 00:21 | |
| GheRivero | getattr(service.client, 'endpoint', getattr(service.client, 'http_client.endpoint', None) will work but looks crappy :/ | 00:23 |
| devananda | GheRivero: split on two lines witha coment explaining compat issue? | 00:23 |
| jroll | yeah :/ | 00:24 |
| jroll | GheRivero: I would split it up, like: if getattr(service.client, 'endpoint'): x = service.client.endpoint; else:... | 00:25 |
| *** zigo has quit IRC | 00:35 | |
| *** max_lobur has quit IRC | 00:39 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 00:40 |
| *** GheRivero is now known as GheRivero_afk | 00:40 | |
| jroll | GheRivero_afk: sorry for the -1 again :/ | 00:44 |
| * GheRivero_afk grabs some coffe | 00:45 | |
| *** zigo has joined #openstack-ironic | 00:47 | |
| *** rakesh_hs has joined #openstack-ironic | 00:50 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 00:54 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 00:54 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 00:55 |
| jroll | JoshNang: your dependency is out of date | 00:55 |
| jroll | ha ok | 00:55 |
| jroll | never mind | 00:55 |
| JoshNang | jroll: ha just saw that | 00:55 |
| JayF | FWIW the Rackspace team is booked up for the mid-cycle meetup | 00:57 |
| jroll | speak for yourself :) | 00:57 |
| mrda | \o/ | 01:00 |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 01:12 |
| devananda | jroll: any chance someone on your team would write operational docs upstream? | 01:20 |
| *** nosnos has joined #openstack-ironic | 01:24 | |
| *** chuckC has quit IRC | 01:29 | |
| * devananda runs out for dinner | 01:36 | |
| *** rakesh_hs has quit IRC | 01:44 | |
| *** mdorman has quit IRC | 01:57 | |
| *** mdorman has joined #openstack-ironic | 01:58 | |
| *** mdorman has quit IRC | 02:05 | |
| *** Haomeng|2 has joined #openstack-ironic | 02:06 | |
| *** Haomeng has quit IRC | 02:06 | |
| *** Haomeng|2 has quit IRC | 02:11 | |
| *** dwalleck has joined #openstack-ironic | 02:52 | |
| *** harlowja is now known as harlowja_away | 03:02 | |
| *** harlowja_away is now known as harlowja | 03:09 | |
| *** harlowja is now known as harlowja_away | 03:10 | |
| *** harlowja_away is now known as harlowja | 03:11 | |
| *** nosnos has quit IRC | 03:38 | |
| *** max_lobur has joined #openstack-ironic | 03:40 | |
| *** annegentle has joined #openstack-ironic | 03:42 | |
| *** sseago__ has quit IRC | 04:07 | |
| *** blamar has quit IRC | 04:11 | |
| *** annegentle has quit IRC | 04:11 | |
| *** notq has quit IRC | 04:11 | |
| *** ekarlso has quit IRC | 04:11 | |
| *** devananda has quit IRC | 04:11 | |
| *** kylestev has quit IRC | 04:11 | |
| *** russell_h has quit IRC | 04:11 | |
| *** lynxman has quit IRC | 04:11 | |
| *** matty_dubs|gone has quit IRC | 04:11 | |
| *** pquerna has quit IRC | 04:11 | |
| *** Madasi has quit IRC | 04:11 | |
| *** mrda has quit IRC | 04:12 | |
| *** mrda has joined #openstack-ironic | 04:21 | |
| *** annegentle has joined #openstack-ironic | 04:22 | |
| *** blamar has joined #openstack-ironic | 04:22 | |
| *** notq has joined #openstack-ironic | 04:22 | |
| *** devananda has joined #openstack-ironic | 04:22 | |
| *** kylestev has joined #openstack-ironic | 04:22 | |
| *** russell_h has joined #openstack-ironic | 04:22 | |
| *** lynxman has joined #openstack-ironic | 04:22 | |
| *** ekarlso has joined #openstack-ironic | 04:22 | |
| *** matty_dubs|gone has joined #openstack-ironic | 04:22 | |
| *** pquerna has joined #openstack-ironic | 04:22 | |
| *** Madasi has joined #openstack-ironic | 04:22 | |
| *** sseago has joined #openstack-ironic | 04:25 | |
| *** rameshg87 has joined #openstack-ironic | 04:26 | |
| *** dwalleck has quit IRC | 04:33 | |
| *** rakesh_hs has joined #openstack-ironic | 04:43 | |
| *** rameshg87 has left #openstack-ironic | 04:44 | |
| *** rameshg87 has joined #openstack-ironic | 04:44 | |
| *** nosnos has joined #openstack-ironic | 04:45 | |
| *** nosnos has quit IRC | 04:51 | |
| *** killer_prince is now known as lazy_prince | 04:51 | |
| *** nosnos has joined #openstack-ironic | 04:51 | |
| *** ajc_ has joined #openstack-ironic | 04:59 | |
| *** coolsvap|afk is now known as coolsvap | 05:01 | |
| *** ramineni has joined #openstack-ironic | 05:01 | |
| *** harlowja is now known as harlowja_away | 05:06 | |
| *** rameshg87 has quit IRC | 05:06 | |
| *** amitpp has joined #openstack-ironic | 05:09 | |
| *** bvivek has joined #openstack-ironic | 05:21 | |
| *** pcrews has quit IRC | 05:21 | |
| *** geekyogi has joined #openstack-ironic | 05:28 | |
| *** rakesh_hs has quit IRC | 05:29 | |
| *** rakesh_hs has joined #openstack-ironic | 05:37 | |
| *** Nisha has joined #openstack-ironic | 05:44 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 05:57 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 05:58 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 05:59 |
| *** nosnos has quit IRC | 06:01 | |
| *** geekyogi has quit IRC | 06:01 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/103291 | 06:02 |
| *** nosnos has joined #openstack-ironic | 06:03 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 06:08 |
| *** rameshg87 has joined #openstack-ironic | 06:11 | |
| *** rameshg87 has left #openstack-ironic | 06:12 | |
| *** rameshg87 has joined #openstack-ironic | 06:12 | |
| *** zigo has quit IRC | 06:13 | |
| *** zigo has joined #openstack-ironic | 06:14 | |
| *** geekyogi has joined #openstack-ironic | 06:20 | |
| *** rameshg87 has left #openstack-ironic | 06:20 | |
| *** rameshg87 has joined #openstack-ironic | 06:20 | |
| *** rameshg87 has quit IRC | 06:20 | |
| *** rameshg87 has joined #openstack-ironic | 06:22 | |
| *** krtaylor has quit IRC | 06:23 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 06:24 | |
| *** nikunj2512 has joined #openstack-ironic | 06:33 | |
| *** krtaylor has joined #openstack-ironic | 06:36 | |
| openstackgerrit | Anusha Ramineni proposed a change to openstack/ironic-specs: Out-Of-Band firmware update for Baremetal Servers https://review.openstack.org/100842 | 06:39 |
| *** ndipanov has joined #openstack-ironic | 06:53 | |
| *** takadayuiko has joined #openstack-ironic | 06:56 | |
| takadayuiko | Hi, Ironic! | 07:00 |
| openstackgerrit | Anusha Ramineni proposed a change to openstack/ironic-specs: iLO Management Driver for firmware update https://review.openstack.org/103760 | 07:03 |
| *** romcheg has joined #openstack-ironic | 07:04 | |
| *** Mikhail_D_ltp has quit IRC | 07:12 | |
| *** sabah has joined #openstack-ironic | 07:13 | |
| *** jcoufal has joined #openstack-ironic | 07:14 | |
| *** coolsvap is now known as coolsvap|afk | 07:25 | |
| rameshg87 | hello takadayuiko | 07:26 |
| romcheg | Morning Ironic! | 07:32 |
| mrda | hey romcheg! | 07:32 |
| *** mkerrin has quit IRC | 07:33 | |
| openstackgerrit | Shivanand Tendulker proposed a change to openstack/ironic-specs: Firmware setting design spec. https://review.openstack.org/101122 | 07:34 |
| openstackgerrit | Shivanand Tendulker proposed a change to openstack/ironic-specs: Design spec for iLO driver for firmware settings https://review.openstack.org/103065 | 07:39 |
| mrda | 'night Ironic! See you tomorrow! | 07:49 |
| *** mrda is now known as mrda-away | 07:49 | |
| openstackgerrit | lawrancejing proposed a change to openstack/ironic: Fix the section name in CONTRIBUTING.rst https://review.openstack.org/103781 | 07:55 |
| takadayuiko | Hello, rameshg87 :) | 07:56 |
| openstackgerrit | lawrancejing proposed a change to openstack/python-ironicclient: Add CONTRIBUTING.rst https://review.openstack.org/103784 | 08:01 |
| *** max_lobur has quit IRC | 08:01 | |
| *** max_lobur has joined #openstack-ironic | 08:02 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 08:02 | |
| *** derekh_ has joined #openstack-ironic | 08:05 | |
| openstackgerrit | lawrancejing proposed a change to openstack/ironic: Use immutable arg rather mutable arg https://review.openstack.org/103789 | 08:14 |
| *** klindgren has quit IRC | 08:14 | |
| soren | I'm a bit confused why the docs say to run dnsmasq. Isn't that exactly why Ironic wants direct access to Neutron? To configure DHCP? | 08:14 |
| takadayuiko | soren: Could you please tell me which doc are you looking? | 08:17 |
| soren | takadayuiko: Right now I was looking at http://ma.ttwagner.com/bare-metal-deploys-with-devstack-and-ironic/ but that's far from the first time I've seen it mentioned. | 08:19 |
| takadayuiko | soren: If you do baremetal provisioning using Nova(for example, "nova boot" command), Ironic access to Neutron as you say, but this blog doesn't use Nova. That's the reason. Wait for a minute... | 08:22 |
| takadayuiko | soren: check the IRC log :D http://eavesdrop.openstack.org/irclogs/%23openstack-ironic/%23openstack-ironic.2014-06-24.log | 08:24 |
| takadayuiko | soren: devananda teached me the system, it's easy to understand. | 08:25 |
| *** lucasagomes has joined #openstack-ironic | 08:37 | |
| lifeless | soren: which docs ? | 08:37 |
| lifeless | soren: that blog post is flat out wrong | 08:37 |
| lifeless | soren: it *might* have been right in january | 08:38 |
| lifeless | soren: but I thought the neutron integration was done quite a bit earlier. | 08:38 |
| Nisha | hi lifeless : | 08:39 |
| *** bvivek has quit IRC | 08:39 | |
| Nisha | lifeless: Did you had a chance to review https://review.openstack.org/100951 | 08:39 |
| lifeless | Nisha: nope | 08:39 |
| Nisha | lifeless: Would you be able to look at? | 08:40 |
| *** coolsvap|afk is now known as coolsvap | 08:41 | |
| lifeless | Nisha: I will try to get to it, probably early next week | 08:41 |
| *** coolsvap is now known as coolsvap|afk | 08:41 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 08:46 |
| *** martyntaylor has joined #openstack-ironic | 08:47 | |
| Nisha | lifeless: Thanks. The review is for Discover node properties at node-create/node-update (generic one) | 08:48 |
| takadayuiko | Many people seem like not being able to run Ironic with "nova boot"... http://www.gossamer-threads.com/lists/openstack/dev/39381 | 08:51 |
| *** martyntaylor has quit IRC | 08:52 | |
| *** martyntaylor has joined #openstack-ironic | 08:53 | |
| agordeev | good morning Ironic :) | 09:03 |
| *** foexle has joined #openstack-ironic | 09:06 | |
| *** bvivek has joined #openstack-ironic | 09:15 | |
| *** bvivek2 has joined #openstack-ironic | 09:20 | |
| *** bvivek has quit IRC | 09:21 | |
| *** pelix has joined #openstack-ironic | 09:23 | |
| *** kpavel has joined #openstack-ironic | 09:24 | |
| kpavel | When i'm trying to use ironic (in mocked devstack environment) it successfully boots. But i see that when i restart it it continues to use kernel/ramdisk (not deploy_kernel) from pxe. Changing boot order to boor from hard drive didn't help. Machine hangs on "Booting from Hard Disk". Is it working as designed? I mean, does machine booted once from network has to boot always from network? Thanks | 09:28 |
| *** coolsvap|afk is now known as coolsvap | 09:28 | |
| rameshg87 | kpavel: ironic boots the machine from network always | 09:29 |
| rameshg87 | kpavel: it is expected. | 09:29 |
| rameshg87 | kpavel: the kernel/ramdisk booted from the network is the kernel/ramdisk of the final image. it mounts the root partition on the disk and continues booting | 09:30 |
| kpavel | rameshg87: i see. so in case ironic node goes down the baremetal machines can't be rebooted? | 09:30 |
| rameshg87 | kpavel: the machine will boot from network again after you have rebooted. it will boot over network until you 'tear down' the node from ironic | 09:31 |
| rameshg87 | kpavel: there is no boot loader installed on disk, so it can't boot from disk | 09:32 |
| kpavel | rameshg87: oh...how do i "tear down" the node from ironic? is there a place/wiki i can read about those things? | 09:32 |
| rameshg87 | kpavel: 'tear down' is the ironic term for destroying an instance. you can use corresponding nova command to delete the install..i guess nova-delete | 09:33 |
| rameshg87 | kpavel: type, you can use corresponding nova command to delete the node i meant :-) | 09:33 |
| kpavel | rameshg87: thanks! is it documented anywhere so i could read about it? | 09:35 |
| rameshg87 | kpavel: i don't see tear-down mentioned anywhere :-( | 09:40 |
| rameshg87 | kpavel: i hope you already have this http://docs.openstack.org/developer/ironic/ | 09:40 |
| rameshg87 | kpavel: and all documentation is available here as well, https://github.com/openstack/ironic/tree/master/doc/source | 09:40 |
| kpavel | rameshg87: yes. maybe i missed it but didn't see anywhere that baremetal nodes don't have own bootloader. | 09:40 |
| rameshg87 | kpavel: okay | 09:41 |
| takadayuiko | rameshg87, kpavel: I'm sorry to interrupt you, your conversation is very useful. I didn't know that loader is not installed, but I thought that MBR was not updated... | 09:41 |
| kpavel | rameshg87: do you know if it was same way before ironic in nova baremetal? | 09:41 |
| soren | lifeless: I'm trying to piece information together from a variety of sources. I suppose some of them are ancient and/or for nova-baremetal. | 09:42 |
| rameshg87 | takadayuiko: currently ironic doesn't install bootloader on the local disk of the baremetal node. | 09:45 |
| rameshg87 | kpavel: i haven't work on nova baremetal before, so i don't know :-) | 09:46 |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic-specs: iLO Virtual Media Deploy Driver https://review.openstack.org/97744 | 09:46 |
| takadayuiko | rameshg87, I understood, thank you :) | 09:48 |
| *** max_lobur has quit IRC | 09:48 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 09:49 |
| *** lazy_prince has quit IRC | 09:50 | |
| kpavel | rameshg87: Thank you! | 09:50 |
| *** loki184 has joined #openstack-ironic | 09:50 | |
| takadayuiko | I'm trying to use Ironic with "nova boot" command, but failed with "501 Not Implemented" error in neutronclient. mmm... | 09:56 |
| takadayuiko | perhaps my neutron is old | 09:57 |
| *** lucasagomes has quit IRC | 10:01 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/python-ironicclient: Add bash completion support for ironic cli https://review.openstack.org/101765 | 10:05 |
| *** Nisha has quit IRC | 10:06 | |
| *** jcoufal has quit IRC | 10:26 | |
| *** jcoufal has joined #openstack-ironic | 10:26 | |
| *** lazy_prince has joined #openstack-ironic | 10:30 | |
| soren | I end up with a KeyError: 'baremetal:deploy_kernel_id' | 10:37 |
| soren | Is there something obvious I should be looking at? | 10:37 |
| soren | The node's driver_info has pxe_image_source, pxe_deploy_kernel, pxe_deploy_ramdisk set. | 10:39 |
| soren | Oh, is this a problem with the image I'm trying to boot there rather than the deploy kernel, perhaps? | 10:39 |
| soren | Are there any special requirements for the images being booted? | 10:40 |
| *** Alexei_987 has joined #openstack-ironic | 10:48 | |
| *** killer_prince has joined #openstack-ironic | 10:49 | |
| *** openstack has joined #openstack-ironic | 10:51 | |
| *** lazy_prince has quit IRC | 10:55 | |
| *** killer_prince is now known as lazy_prince | 10:55 | |
| *** agordeev has quit IRC | 10:58 | |
| *** ramineni has quit IRC | 10:58 | |
| *** overlayer has joined #openstack-ironic | 11:00 | |
| soren | Hm... Ok, I've created a new flavor. I set these extra_specs: {"cpu_arch": "x86_64", "baremetal:deploy_kernel_id": "3732b36d-22be-4083-9178-d09b3749b4b5", "baremetal:deploy_ramdisk_id": "f5473f1e-ec5d-4687-8f65-19eccaabb212"} | 11:01 |
| soren | disk: 30 ephemeral: 200 ram: 32000 vcpus: 10 | 11:02 |
| *** agordeev has joined #openstack-ironic | 11:02 | |
| soren | nova hypervisor-show <uuid> says: free_disk_gb: 900 free_ram_mb: 63879 local_gb: 900 local_gb_used: 0 memory_mb: 64391 running_vms: 0 vcpus: 40 (and a few other data points that I doubt are relevant). | 11:03 |
| soren | Yet I get NoValidHost. | 11:04 |
| soren | Oh. I got this in my debug log: "extra_spec requirement 'x86_64' does not match 'NotFound' _satisfies_extra_specs " | 11:06 |
| soren | Is this something I've failed to configure for the node in Ironic? | 11:06 |
| *** overlayer has quit IRC | 11:09 | |
| *** sabah has quit IRC | 11:13 | |
| *** overlayer has joined #openstack-ironic | 11:41 | |
| *** overlayer has quit IRC | 11:52 | |
| *** athomas has quit IRC | 11:52 | |
| *** overlayer has joined #openstack-ironic | 11:54 | |
| *** amitpp has quit IRC | 11:55 | |
| *** overlayer has quit IRC | 12:01 | |
| *** ajc_ has quit IRC | 12:17 | |
| *** jcoufal has quit IRC | 12:19 | |
| *** dhellmann is now known as dhellmann_ | 12:19 | |
| *** mitz_ has quit IRC | 12:25 | |
| *** lynxman has quit IRC | 12:26 | |
| soren | Ok, I got further. I see some glance stuff in the ironic logs, but nothing furhter seems to happen. | 12:27 |
| *** lynxman has joined #openstack-ironic | 12:28 | |
| *** overlayer has joined #openstack-ironic | 12:29 | |
| *** linggao has joined #openstack-ironic | 12:29 | |
| soren | ...and the node never gets turned on.. Any ideas? | 12:32 |
| *** overlayer has quit IRC | 12:35 | |
| *** jdob has joined #openstack-ironic | 12:37 | |
| romcheg | dtantsur: Morning! | 12:38 |
| romcheg | dtantsur: could you please remind me the link to our mid-cycle meetup? | 12:38 |
| *** jcoufal has joined #openstack-ironic | 12:39 | |
| *** martyntaylor1 has joined #openstack-ironic | 12:48 | |
| *** martyntaylor has quit IRC | 12:48 | |
| *** coolsvap is now known as coolsvap|afk | 12:50 | |
| *** loki184 has quit IRC | 12:50 | |
| *** martyntaylor has joined #openstack-ironic | 12:50 | |
| *** martyntaylor1 has quit IRC | 12:53 | |
| *** jbjohnso__ has joined #openstack-ironic | 12:59 | |
| *** jbjohnso_ has joined #openstack-ironic | 12:59 | |
| *** overlayer has joined #openstack-ironic | 13:02 | |
| *** jbjohnso__ has quit IRC | 13:03 | |
| *** takadayuiko has quit IRC | 13:06 | |
| jroll | devananda: I think it's safe to assume we'll be writing operational docs, yes | 13:07 |
| *** blamar has quit IRC | 13:09 | |
| jroll | soren: I'd suggest posting logs :) | 13:09 |
| soren | jroll: That's the problem, I guess. There's nothing to post, since nothing happens. | 13:10 |
| soren | After the glanceclient bit, there's only the usual periodic_tasks stuff. | 13:10 |
| jroll | well, there's *some* logs | 13:11 |
| *** blamar has joined #openstack-ironic | 13:11 | |
| jroll | you might also look at the console of the node being provisioned | 13:11 |
| soren | There's nothing. It's turned off. | 13:11 |
| jroll | ok | 13:11 |
| jroll | are you running at DEBUG log level? | 13:11 |
| soren | I believe so. | 13:12 |
| soren | I can totally post the logs, no problem. | 13:12 |
| soren | Gimme a sec. | 13:12 |
| jroll | sure | 13:12 |
| jroll | I can't promise I can help, I don't work with the standard driver much. but I'll try :) | 13:12 |
| *** lazy_prince is now known as killer_prince | 13:13 | |
| soren | http://paste.ubuntu.com/7731353/ | 13:16 |
| *** sseago has quit IRC | 13:19 | |
| *** sseago has joined #openstack-ironic | 13:19 | |
| jroll | soren: anything interesting in the nova logs? | 13:20 |
| jroll | I don't see where nova actually tells it to provision | 13:20 |
| jroll | (nova-compute, to be specific) | 13:21 |
| jroll | or ironic-api logs, even | 13:21 |
| soren | I set debug=True in the global part of ironic.conf. | 13:22 |
| soren | And specified a log_file. | 13:22 |
| jroll | oh, I do see it | 13:22 |
| jroll | ok yeah | 13:25 |
| jroll | this worries me: 2014-07-01 17:03:36.344 58570 DEBUG requests.packages.urllib3.connectionpool [-] "GET /v2.0/tokens/revoked HTTP/1.1" 200 698 _make_request /usr/local/lib/python2.7/dist-packages/requests/packages/urllib3/connectionpool.py:350 | 13:25 |
| jroll | not sure if that's normal | 13:25 |
| soren | I'm pretty sure it is. | 13:28 |
| soren | It's meant to check if the token I'm presenting has been revoked. | 13:28 |
| soren | I'm grabbing the nova log file right now. Gimme a minute. | 13:28 |
| *** matty_dubs|gone is now known as matty_dubs | 13:28 | |
| jroll | mmm | 13:28 |
| jroll | soren: I'm not sure nova logs are relevant :| | 13:29 |
| jroll | downloading the images and powering up the machine happen in the same method | 13:30 |
| soren | Well, note that the glanceclient calls don't GET. | 13:30 |
| soren | They HEAD. | 13:30 |
| soren | So they're probably just validating the request. | 13:30 |
| soren | ...and never get around to actually attempting to fulfill it. But I also have no indication that the validation failed. | 13:31 |
| jroll | ohhh. interesting. | 13:31 |
| jroll | ... why would they do that. grr. | 13:32 |
| soren | Oh wait! | 13:33 |
| jroll | it is validating things | 13:33 |
| * jroll waits | 13:33 | |
| soren | 2014-07-01 17:00:02.067 58273 TRACE nova.compute.manager [instance: 4930ddd3-a2fb-444a-916e-761bb752e7b3] InstanceDeployFailure: Failed to provision instance 4930ddd3-a2fb-444a-916e-761bb752e7b3: Failed to deploy. Error: 'kernel_id' | 13:33 |
| jroll | ah ha | 13:33 |
| soren | Not the most useful of errors, but it's something to go by at least. | 13:33 |
| jroll | devs: pls to log all the exceptions in the conductor | 13:33 |
| jroll | ok, so | 13:34 |
| jroll | the image needs that in its properties | 13:34 |
| jroll | not in extra_specs | 13:34 |
| jroll | afaict | 13:34 |
| soren | Ironic can't do whole-disk images? | 13:35 |
| soren | It needs kernel/ramdisk/filesystem? | 13:35 |
| jroll | yeah | 13:35 |
| jroll | because of the whole 'pxe boot the instance image' thing | 13:36 |
| jroll | nothing boots from disk | 13:36 |
| soren | I thought that was just for the "deploy kernel". | 13:36 |
| jroll | that said, there's this awesome new agent-based deploy model :) | 13:36 |
| soren | Maybe I've misunderstood. | 13:36 |
| jroll | it's just not well-documented... at all | 13:36 |
| soren | Oh, that's a new thing? | 13:36 |
| jroll | yeah, this struck me as weird, too | 13:36 |
| soren | I thought that was just how things worked. | 13:36 |
| *** jgrimm has joined #openstack-ironic | 13:37 | |
| *** overlayer has quit IRC | 13:37 | |
| jroll | actually, pquerna made some great diagrams just yesterday | 13:37 |
| jroll | sec | 13:37 |
| soren | So what is the deploy_kernel and deploy_ramdisks for? | 13:37 |
| jroll | here's how ironic works today, with the pxe driver: https://www.evernote.com/shard/s184/sh/041ab958-2ce7-477b-97e5-f504e51c0362/27b115a020543253a631ed3b95881956 | 13:38 |
| jroll | here's how it works with the agent driver (which hasn't landed yet): https://www.evernote.com/shard/s184/sh/0bedd6a4-02a2-4199-8c64-33983d06aae5/f42ebabc08537c2e55901578cabe147d | 13:38 |
| soren | Hm. My nodes never get turned on. | 13:38 |
| *** overlayer has joined #openstack-ironic | 13:38 | |
| soren | So I don't even get to the initial IPMI Power on. | 13:38 |
| jroll | they both use ramdisks, but one exposes iscsi mounts, the other exposes a rest api | 13:38 |
| jroll | right, because it failed validating the image | 13:38 |
| *** pcrews has joined #openstack-ironic | 13:39 | |
| soren | Well... Right, yeah. | 13:39 |
| soren | I just.. | 13:39 |
| jroll | you got to here, which runs just before deploy(): https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe.py#L487 | 13:39 |
| soren | What purpose do the deploy_kernel+ramdisk serve? | 13:40 |
| soren | If you PXE boot forever, I understand the use for separate kernel/ramdisk, but if you intend to write something bootable to... | 13:41 |
| jroll | so the deploy ramdisk is a small image that boots and exposes /dev/sda as an iscsi mount, which ironic-conductor dd's the instance image to, and then uses instance kernel/ramdisk to pxe boot that image | 13:41 |
| jroll | so it pxe boots forever, yes | 13:41 |
| soren | Oh. You do expect to PXE boot forever, right? | 13:41 |
| soren | Yeah, I just realised that. | 13:41 |
| soren | Erk, so we're back to not being able to do kernel upgrades using the normal mechanisms? | 13:41 |
| jroll | took me months before I noticed that | 13:41 |
| jroll | ooo | 13:42 |
| jroll | I didn't think of that | 13:42 |
| jroll | that's... fun | 13:42 |
| soren | This should be possible. You can instruct pxelinux to just boot from local disk. | 13:42 |
| jroll | if you write a boot partition :) | 13:42 |
| soren | Right, exactly. | 13:43 |
| jroll | there's specs up for some of this | 13:43 |
| jroll | but I think they're mostly going to rely on the new agent stuff | 13:43 |
| soren | ...which I believe is common nowadays. | 13:43 |
| soren | Anyway, for now I'll just do kernel/ramdisk/filesystem. | 13:43 |
| jroll | right | 13:43 |
| soren | I'm sure diskimage-builder can help me with that somehow. | 13:43 |
| jroll | yep, that's the idea | 13:43 |
| soren | Do you happen to have a command line that will build exactly that? | 13:47 |
| soren | I seem to keep finding stale docs, so I'd rather skip that part this time :) | 13:47 |
| jroll | I'm sorry, I don't. we're using... full disk images and booting from disk | 13:47 |
| soren | Lucky :) | 13:48 |
| jroll | I would encourage you to use IPA and that driver, but there's not too many docs on it | 13:48 |
| soren | IPA? | 13:48 |
| jroll | the best docs we have right now are this blog post: http://developer.rackspace.com/blog/how-we-run-ironic-and-you-can-too.html | 13:48 |
| jroll | yeah, ironic-python-agent | 13:48 |
| soren | Ah, that. | 13:48 |
| jroll | https://github.com/openstack/ironic-python-agent | 13:48 |
| soren | I'm too close to getting this working to really entertain the idea of tossing it before I see it work at least once :) | 13:49 |
| jroll | heh | 13:49 |
| jroll | I don't blame you | 13:49 |
| soren | Do you work on OnMetal? | 13:49 |
| jroll | yes I do | 13:49 |
| jroll | if you stick around for a bit, I'm sure someone can help you out with building the images etc | 13:50 |
| jroll | I'm also tempted to file a bug "takes days to figure out how to deploy ironic" | 13:50 |
| *** overlayer has quit IRC | 13:54 | |
| *** overlayer has joined #openstack-ironic | 13:58 | |
| *** pradipta_away is now known as pradipta | 13:59 | |
| *** martyntaylor has quit IRC | 14:04 | |
| *** overlayer has quit IRC | 14:08 | |
| NobodyCam | good morning Ironic | 14:09 |
| jroll | mornin NobodyCam :) | 14:09 |
| NobodyCam | morning jroll | 14:09 |
| matty_dubs | Morning NobodyCam, jroll | 14:10 |
| jroll | when you're up for it, soren here needs some help getting/building kernel/ramdisk for an instance image | 14:10 |
| jroll | if you know how to do so | 14:10 |
| jroll | NobodyCam: ^ | 14:10 |
| jroll | heya matty_dubs | 14:11 |
| soren | I think I *just* worked it out. | 14:11 |
| jroll | welp | 14:11 |
| *** rakesh_hs has quit IRC | 14:13 | |
| *** nosnos has quit IRC | 14:13 | |
| *** nosnos has joined #openstack-ironic | 14:14 | |
| *** nosnos has quit IRC | 14:18 | |
| *** martyntaylor has joined #openstack-ironic | 14:19 | |
| *** max_lobur has joined #openstack-ironic | 14:20 | |
| *** martyntaylor has left #openstack-ironic | 14:20 | |
| NobodyCam | moring matty_dubs | 14:21 |
| NobodyCam | soren: your using diskimage builder? | 14:21 |
| *** max_lobur has quit IRC | 14:24 | |
| *** dwalleck has joined #openstack-ironic | 14:27 | |
| NobodyCam | rameshg87: you happen to be around? | 14:45 |
| NobodyCam | wanyen: you here? | 14:49 |
| *** chuckC has joined #openstack-ironic | 14:50 | |
| *** Mikhail_D_ltp has quit IRC | 14:56 | |
| *** mitz_ has joined #openstack-ironic | 14:57 | |
| *** mitz_ has quit IRC | 14:57 | |
| *** romcheg has quit IRC | 14:59 | |
| *** romcheg has joined #openstack-ironic | 15:00 | |
| *** max_lobur has joined #openstack-ironic | 15:06 | |
| *** mitz has quit IRC | 15:08 | |
| *** mitz has joined #openstack-ironic | 15:12 | |
| *** coolsvap|afk is now known as coolsvap | 15:18 | |
| matty_dubs | Has anyone run into issues the past few days with devstack failing with Ironic, with an error "worlddump.py: error: argument -d/--dir: expected one argument" ? | 15:27 |
| matty_dubs | I don't know devstack well enough to know what's going wrong | 15:29 |
| Shrews_ | matty_dubs: hrm, that seems familiar. i had issues with it the other day after doing a 'git pull' on devstack that caused all sorts of havoc. i ended up just removing all of /usr/local/lib/python2.7 and starting fresh | 15:29 |
| Shrews_ | So, with regards to mitigating the effects of NodeLocked errors that we sometimes see (and since the async api is far off)... Why don't we just build in a retry mechanism around the dbapi.reserve_node() call in TaskManager.__init__()? | 15:33 |
| Shrews_ | Wouldn't completely eliminate the problem, but should reduce how often clients see it | 15:36 |
| *** Shrews_ is now known as Shrews | 15:36 | |
| *** nikunj2512 has quit IRC | 15:37 | |
| *** foexle has quit IRC | 15:38 | |
| *** mdorman has joined #openstack-ironic | 15:59 | |
| *** linggao has quit IRC | 16:00 | |
| *** Ng has quit IRC | 16:02 | |
| *** Ng has joined #openstack-ironic | 16:03 | |
| *** Ng has quit IRC | 16:04 | |
| *** Ng has joined #openstack-ironic | 16:06 | |
| *** matty_dubs is now known as matty_dubs|lunch | 16:06 | |
| *** nikunj2512 has joined #openstack-ironic | 16:07 | |
| *** coolsvap is now known as coolsvap|afk | 16:09 | |
| devananda | morning, all | 16:10 |
| romcheg | Goog morning devananda! | 16:10 |
| NobodyCam | good morning devananda | 16:10 |
| NobodyCam | morning romcheg | 16:10 |
| romcheg | …and NobodyCam and everyone else :) | 16:11 |
| *** Nisha has joined #openstack-ironic | 16:15 | |
| Nisha | NobodyCam: there? | 16:15 |
| NobodyCam | maybe | 16:16 |
| Nisha | NobodyCam: Hi | 16:16 |
| NobodyCam | thou I have a call in about 15 | 16:16 |
| NobodyCam | Hi | 16:16 |
| Nisha | NobodyCam: saw your comments on https://review.openstack.org/#/c/100951 | 16:17 |
| Nisha | NobodyCam: just a clarification on one comment | 16:17 |
| Nisha | NobodyCam: it should not take much time | 16:17 |
| Nisha | NobodyCam: "many servers may have unconnected NICs that are not used" | 16:17 |
| Nisha | NobodyCam: Could you clarify this more | 16:17 |
| NobodyCam | sure | 16:18 |
| Nisha | NobodyCam: How do you find if NIC is connected? by IP address? | 16:18 |
| NobodyCam | the option --create-ports will add all nics found even if they dont have a cable pluged in | 16:18 |
| Nisha | Yes, whatever ILO/driver returns | 16:19 |
| *** dwalleck has quit IRC | 16:19 | |
| NobodyCam | Nisha: is there a link status field when you check for mac addresses | 16:19 |
| Nisha | i will paste an output from ilo here | 16:19 |
| Nisha | <NIC> <NETWORK_PORT VALUE = "Port 1"/> <PORT_DESCRIPTION VALUE = "N/A"/> <LOCATION VALUE = "Embedded"/> <MAC_ADDRESS VALUE = "9c:b6:54:bb:01:90"/> <IP_ADDRESS VALUE = "N/A"/> <STATUS VALUE = "Unknown"/> </NIC> | 16:20 |
| NobodyCam | please use paste.openstack.org | 16:20 |
| Nisha | NobodyCam: a small one | 16:20 |
| NobodyCam | :) | 16:20 |
| Nisha | NobodyCam: How do you determine from this? | 16:20 |
| devananda | Nisha: "how do yhou find if NIC is connected?" -- LLDP | 16:20 |
| NobodyCam | Nisha: is this sample from a connected or unconnected nic? | 16:21 |
| NobodyCam | :) | 16:22 |
| Nisha | NobodyCam: This is the sample from ilo where server is not deployed | 16:22 |
| Nisha | NobodyCam: devananda frankly speaking i dont know how to determine this info if NIC is connected or not :) | 16:22 |
| NobodyCam | lldp = Link Layer Discovery Protocol | 16:23 |
| *** bvivek2 has quit IRC | 16:23 | |
| *** hemna has joined #openstack-ironic | 16:23 | |
| devananda | Nisha: is the host running an operating system, or is this via OOB management (iLO) ? | 16:23 |
| Nisha | This is from OOB(iLO) | 16:24 |
| Nisha | devananda: the host is not running any OS | 16:24 |
| Nisha | devananda: its not deployed | 16:24 |
| devananda | Nisha: I see. Seems very odd to me that iLO would not expose that, but you probably know that better than I | 16:25 |
| NobodyCam | Nisha: what is status_value field for? | 16:26 |
| *** stendulker has joined #openstack-ironic | 16:26 | |
| Nisha | devananda: NobodyCam may be i can try from an ilo where OS is deployed and see if i get more info | 16:27 |
| NobodyCam | Nisha: I know with an os you can get that info | 16:27 |
| NobodyCam | I would need to look at ilo more closly | 16:27 |
| NobodyCam | jumping on call...brb | 16:27 |
| Nisha | NobodyCam: i think it means if the NIC configuration is OK or not. for ilo dedicated port the status value is "OK" | 16:28 |
| Nisha | NobodyCam: ok. Thanks | 16:28 |
| *** ndipanov has quit IRC | 16:31 | |
| *** amitpp has joined #openstack-ironic | 16:32 | |
| devananda | Nisha: from within Linux, you should be able to inspect /sys/class/net/*/operstate and other files in that directory to determine connection state | 16:33 |
| Nisha | devananda: Thanks. I will see this one. Let me see if iLO will be able to figure it out. | 16:34 |
| *** romcheg has quit IRC | 16:34 | |
| Nisha | devananda: request you also to look at the spec and provide your comments. | 16:35 |
| Nisha | devananda: https://review.openstack.org/#/c/100951 | 16:35 |
| stendulker | Hello jroll | 16:36 |
| *** Alexei_987 has quit IRC | 16:37 | |
| Nisha | devananda: NobodyCam the node-update can be run after the BM is deployed? | 16:37 |
| Nisha | devananda: NobodyCam i asked this because the intention of port creation in the spec is before OS provisioning | 16:37 |
| *** geekyogi has quit IRC | 16:38 | |
| *** dwalleck has joined #openstack-ironic | 16:39 | |
| *** loki184 has joined #openstack-ironic | 16:40 | |
| NobodyCam | Nisha: some quick checking with google look like there is a link status field in the web ui: Link State—The current link speed of the selected iLO network interface. The default value is | 16:41 |
| devananda | Nisha: yes, node-update can be run any time | 16:41 |
| NobodyCam | Auto-Negotiate. | 16:41 |
| NobodyCam | thou I am not sure how you would access that info | 16:41 |
| devananda | Nisha: what is the impact of running node-update --discover while an OS is already on the node? | 16:41 |
| *** geekyogi has joined #openstack-ironic | 16:41 | |
| devananda | i need to step away for half an hour... bbs | 16:42 |
| NobodyCam | :) | 16:42 |
| Nisha | devananda: as of nwo the proposal is to rediscover whenever node-update --discover is issued | 16:42 |
| Nisha | devananda: NobodyCam what shall be the expected behavior? it shud update or not? | 16:42 |
| NobodyCam | Nisha: what would it update.. things like ram size, disk, cpus? | 16:45 |
| NobodyCam | I'm a little un clear of the question | 16:45 |
| NobodyCam | brb | 16:46 |
| Nisha | NobodyCam: yes all the node properties . | 16:46 |
| Nisha | NobodyCam: deva asked this just few min back "what is the impact of running node-update --discover while an OS is already on the node?" | 16:46 |
| Nisha | NobodyCam: so the proposal is to discover node properties with this option (as of now all the properties whatever a driver will implement) | 16:48 |
| Nisha | NobodyCam: generally there may be two use cases for node-update here: 1. user may not have discovered during node-create as now even node-create will have option --discover. 2. some h/w changes have happened and user wants to discover them again using node-update | 16:49 |
| *** pradipta is now known as pradipta_away | 16:50 | |
| NobodyCam | so ya I don't see why it wouldn't update | 16:50 |
| Nisha | NobodyCam: sorry i didnt get... | 16:50 |
| NobodyCam | Nisha: yes the node should update | 16:51 |
| NobodyCam | brb | 16:51 |
| Nisha | NobodyCam: ok Thanks. | 16:51 |
| *** pelix has quit IRC | 16:52 | |
| *** nikunj2512 has quit IRC | 16:54 | |
| *** jcoufal has quit IRC | 16:54 | |
| *** Ng has quit IRC | 16:55 | |
| *** Ng has joined #openstack-ironic | 16:57 | |
| stendulker | Hello NobodyCam, I have update design spec for firmware settings. Have addressed the review comments. Please have a look https://review.openstack.org/#/c/101122/ | 16:57 |
| *** rakesh_hs has joined #openstack-ironic | 17:00 | |
| *** derekh_ has quit IRC | 17:01 | |
| *** Poornima has joined #openstack-ironic | 17:01 | |
| *** matty_dubs|lunch is now known as matty_dubs | 17:02 | |
| *** jdob has quit IRC | 17:02 | |
| *** jdob has joined #openstack-ironic | 17:02 | |
| *** harlowja_away is now known as harlowja | 17:06 | |
| *** romcheg has joined #openstack-ironic | 17:06 | |
| *** Penick has joined #openstack-ironic | 17:09 | |
| NobodyCam | stendulker: will do in a meeting atm | 17:12 |
| *** Penick has quit IRC | 17:15 | |
| jroll | stendulker: hi | 17:16 |
| *** jdob has quit IRC | 17:17 | |
| *** jdob has joined #openstack-ironic | 17:17 | |
| *** linggao_ has joined #openstack-ironic | 17:25 | |
| *** linggao_ has quit IRC | 17:26 | |
| *** athomas has joined #openstack-ironic | 17:33 | |
| stendulker | jroll: I have update design spec for firmware settings. Have addressed the review comments. Please have a look https://review.openstack.org/#/c/101122/ | 17:41 |
| stendulker | NobodyCam: Ok. | 17:41 |
| stendulker | jroll: Wanted to understand the security concerns raised by you. | 17:42 |
| *** dwalleck_ has joined #openstack-ironic | 17:45 | |
| *** dwalleck has quit IRC | 17:49 | |
| *** dwalleck_ has quit IRC | 17:52 | |
| *** dhellmann_ is now known as dhellmann | 17:53 | |
| *** dwalleck has joined #openstack-ironic | 17:53 | |
| *** stendulker has quit IRC | 17:55 | |
| *** Penick has joined #openstack-ironic | 18:04 | |
| *** harlowja is now known as harlowja_away | 18:10 | |
| *** dwalleck has quit IRC | 18:12 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 18:16 |
| *** dwalleck has joined #openstack-ironic | 18:18 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 18:18 |
| Nisha | NobodyCam: just reposted the spec after the addressing the comments, please have a look... | 18:19 |
| NobodyCam | Nisha: will do just started to look at stendulker's | 18:20 |
| NobodyCam | then I'll loop back to that one | 18:20 |
| Nisha | NobodyCam: Ok. :) | 18:21 |
| soren | NobodyCam: Yes, using diskimage-builder. | 18:25 |
| NobodyCam | hey soren did you get the image built? | 18:26 |
| *** bvivek has joined #openstack-ironic | 18:27 | |
| soren | NobodyCam: Yeah, I don't think images are my problem anymore. | 18:27 |
| NobodyCam | :) | 18:27 |
| *** harlowja_away is now known as harlowja | 18:33 | |
| jroll | stendulker left me :( | 18:34 |
| sirushti | Hi all, could I get some reviews on https://review.openstack.org/#/c/97150/, it's the whole disk image spec. It feels like it's ready after addressing all your comments :-) | 18:36 |
| Shrews | NobodyCam: devananda: We need to try to push https://review.openstack.org/103670 through today to unbreak us. I just did a recheck on it. | 18:42 |
| devananda | Shrews: ack | 18:42 |
| NobodyCam | Shrews: will take look now | 18:43 |
| NobodyCam | Shrews: lol py26 failed on your recheck | 18:44 |
| Shrews | gah | 18:45 |
| Shrews | ERROR: py26: InterpreterNotFound: python2.6 | 18:45 |
| Shrews | nice | 18:46 |
| NobodyCam | lol | 18:46 |
| JayF | Man, that's some broken code | 18:46 |
| *** dwalleck has quit IRC | 18:46 | |
| JayF | Managed to uninstall python, that's impressive | 18:46 |
| matty_dubs | LOL | 18:46 |
| Shrews | this wouldn't happen if we weren't using python... just saying | 18:46 |
| NobodyCam | Shrews: you could rewrite all of openstack in, oh say, bash :-p | 18:47 |
| Shrews | https://bugs.launchpad.net/openstack-ci/+bug/1282795 | 18:47 |
| NobodyCam | importance: Undecided → Low | 18:48 |
| devananda | Shrews: +2'd, waiting on Jenkins anyway | 18:48 |
| NobodyCam | humm youd think some one was trying to get rid of 26 | 18:48 |
| NobodyCam | devananda: ya. LGTM too, I'll land whance mr J is happy | 18:49 |
| devananda | NobodyCam: ack, ty | 18:50 |
| NobodyCam | s/whance/once/ | 18:50 |
| NobodyCam | -p | 18:50 |
| * Shrews likes "whance" | 18:50 | |
| * NobodyCam 's coffee pot broke the other day :( | 18:50 | |
| JoshNang | NobodyCam: gotta have fail over coffee pots. it's critical infrastructure | 18:51 |
| Shrews | NobodyCam: i issued another recheck hoping it would short-circuit the other. that's not a thing, apparently | 18:51 |
| NobodyCam | Shrews: ya, I think you have to wait, or at least you use to ... | 18:55 |
| NobodyCam | JoshNang: ++++++++++++++++^2 | 18:56 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 18:57 |
| Shrews | NobodyCam: JoshNang: http://www.toadworld.com/cfs-file.ashx/__key/communityserver-blogs-components-weblogfiles/00-00-00-00-29/For-Blog.png | 18:57 |
| NobodyCam | :) | 18:57 |
| JoshNang | heh | 18:57 |
| *** loki184 has quit IRC | 18:57 | |
| devananda | Shrews: uploading a new change will interrupt it, IIRC | 18:59 |
| *** derekh_ has joined #openstack-ironic | 18:59 | |
| NobodyCam | Shrews: secondary = > http://www.starbucks.com/store-locator/search/geolocate | 18:59 |
| NobodyCam | devananda: yep that works | 19:00 |
| Shrews | So, no thoughts on my suggestion earlier to wrap dbapi.reserve_node() in a retry to help mitigate the NodeLocked problems? | 19:01 |
| devananda | adam_g: our pep8 cleanup patches only had 1 line of conflict | 19:03 |
| devananda | adam_g: mind if I fixup and rebase yours? | 19:03 |
| *** linggao has joined #openstack-ironic | 19:04 | |
| adam_g | devananda, sure, go for it | 19:04 |
| * Shrews just files a spec | 19:05 | |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: pep8 cleanup of Nova code https://review.openstack.org/103001 | 19:05 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: PEP fixes for the Nova driver https://review.openstack.org/103162 | 19:05 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 19:06 |
| NobodyCam | Nisha: ahh you are able to detect if a cable is connected :) | 19:07 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 19:08 |
| linggao | Hi devananda, could you please help me checking why my patch failed Jenkins again? | 19:10 |
| *** geekyogi has quit IRC | 19:10 | |
| linggao | https://review.openstack.org/#/c/102318/ | 19:10 |
| Shrews | linggao: that's due to a bug that https://review.openstack.org/103670 is fixing | 19:11 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 19:12 |
| *** rakesh_hs has quit IRC | 19:12 | |
| linggao | Shrews, thanks. should I do "recheck bug 103670" ? | 19:12 |
| linggao | Or I should wait. | 19:13 |
| Shrews | linggao: no. 103670 is not a bug number, it's a review. you must wait for that review to merge, then "recheck no bug" | 19:13 |
| linggao | ? | 19:13 |
| Shrews | linggao: we are trying to push it through now | 19:14 |
| *** dwalleck has joined #openstack-ironic | 19:14 | |
| linggao | ok. Shrews thanks. I'll wait. | 19:14 |
| NobodyCam | Nisha: reviewed | 19:16 |
| soren | Are there any guidelines for Neutron configuration to get it to work with Ironic? An example config would be wonderful. | 19:16 |
| soren | I *think* my problems are in that domain now. | 19:17 |
| NobodyCam | soren: like the neutron.conf or other configs? | 19:18 |
| soren | Oh. | 19:18 |
| * soren stumbles on tripleo-incubator/doc/source/neutron_notes.rst | 19:18 | |
| NobodyCam | :) | 19:18 |
| soren | NobodyCam: I honestly don't know. Neutron's configuration directory makes very little sense to me. | 19:19 |
| soren | That file says to run dnsmasq manually, though. Is that still accurate? | 19:20 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 19:22 |
| *** geekyogi has joined #openstack-ironic | 19:22 | |
| NobodyCam | soren: also see http://docs.openstack.org/developer/tripleo-incubator/deploying.html | 19:23 |
| Nisha | NobodyCam: Thanks for reviewing the spec. Please could you have a look at https://review.openstack.org/#/c/103007/ also. It is to implement the generic one using iLO driver | 19:25 |
| Nisha | NobodyCam: "ahh you are able to detect if a cable is connected" regarding this comment | 19:26 |
| Shrews | NobodyCam: 103670 requeue'd *crosses fingers* | 19:26 |
| Nisha | NobodyCam: i couldnt find it, but i think there shall be a way. I will get in touch with iLO team to find it out. But i have mentioned it in generic spec as others can find it :) | 19:27 |
| NobodyCam | :) | 19:28 |
| Nisha | NobodyCam: Thats why i have mentioned defaul action will be creating for all the discovered MACs | 19:28 |
| Nisha | %s/defaul/default | 19:29 |
| NobodyCam | yep :) | 19:29 |
| NobodyCam | Shrews: watching zuul | 19:29 |
| NobodyCam | py27 pass / pep8 pass | 19:35 |
| NobodyCam | lol like watching paint dry | 19:36 |
| NobodyCam | woo hoo py26 pass | 19:36 |
| NobodyCam | not just doc's and tempest to go | 19:37 |
| NobodyCam | s/not/now/ | 19:37 |
| *** Poornima has quit IRC | 19:39 | |
| devananda | soren: with ironic, if you are using neutron, you do not need to run a separate dnsmasq process | 19:41 |
| *** amitpp has quit IRC | 19:41 | |
| soren | *if* I'm using Neutron? | 19:41 |
| devananda | right | 19:41 |
| soren | I have a choice? | 19:41 |
| devananda | soren: the OnMetal team isn't using neutron. they've blogged a bit about that, i think | 19:41 |
| devananda | and I believe there are patches up to make that pluggable | 19:42 |
| soren | If I'm using Ironic from the Icehouse release.. Are there any docs to tell me how to not have to deal with the madnass that is Neutron? | 19:43 |
| soren | If no, any pointers? | 19:44 |
| soren | Or must I cry myself to sleep again tonight? | 19:44 |
| devananda | soren: no docs on that. as i said, the work to decouple is in flight now | 19:44 |
| soren | devananda: So Neutron isn't optional right now? | 19:45 |
| JayF | devananda: we are using neutron, with a custom plugin though which means our stuff may be less useful to others right now http://developer.rackspace.com/blog/how-we-run-ironic-and-you-can-too.html | 19:45 |
| JayF | jroll or morgabra would be able to speak more in detail about it, I only understand from a high level | 19:46 |
| jroll | hi | 19:46 |
| jroll | we're using neutron for network isolation | 19:46 |
| jroll | but not for dhcp | 19:46 |
| JayF | gah, I missed some of the context | 19:47 |
| * soren sobs | 19:47 | |
| devananda | JayF: ah, sorry - thought you guys were not using neutron | 19:48 |
| devananda | soren: i stand corrected. sorry :( | 19:48 |
| jroll | well | 19:48 |
| JayF | We're using neutron, just not for dhcp/dnsmasq. Which was what the conversation was about. So I'm wrong (too?) | 19:48 |
| devananda | soren: also this is what I was thinking of -- https://review.openstack.org/#/c/102296/ -- it's a spec for external DHCP | 19:49 |
| soren | If I can't get rid of Neutron altogether, I think I'd better stay on the beaten path.. (?) | 19:49 |
| jroll | was *just* going to link that | 19:49 |
| jroll | soren: our stuff would make it totally optional | 19:49 |
| jroll | soren: you just won't be able to isolate provisioning network from customer network | 19:49 |
| jroll | soren: which might not be an issue for you; idk | 19:49 |
| soren | jroll: And I look forward to that :) Right now, I'm trying to get IceHouse deployed. Just to not have a moving target to deal with. | 19:49 |
| jroll | fwiw, ironic has been fairly stable day to day... just need to pay attention a bit | 19:50 |
| soren | jroll: No, that's not a concern at all right now. This is a single tenant ironic cloud. | 19:50 |
| jroll | and yes, icehouse is a no-go for avoiding neutron, without changing code | 19:50 |
| soren | jroll: Thanks for the clarification. | 19:51 |
| jroll | np | 19:51 |
| soren | Ok, so with that in mind, the final line of tripleo-incubator/doc/source/neutron_notes.rst is all lies? | 19:51 |
| soren | I don't actually have to run dnsmasq on my own? | 19:51 |
| soren | Ever? At all? | 19:51 |
| soren | (Assuming the Ice House release) | 19:52 |
| jroll | sounds like it, from what devananda said | 19:53 |
| * devananda reads that doc | 19:54 | |
| devananda | wait - why does a tripleo-incubator doc file say "after starting devstack"?? | 19:54 |
| jroll | are you saying devstack isn't meant for production | 19:55 |
| jroll | uh oh | 19:55 |
| devananda | soren: that file is referring to several things specific to nova-baremetal | 19:55 |
| devananda | soren: in other words, there's no guarantee it has anything to do with ironic | 19:55 |
| *** dwalleck has quit IRC | 19:56 | |
| devananda | soren: for instance, "bm-helper" is dead. long live bm-helper. | 19:56 |
| soren | .. | 19:56 |
| soren | Ok. | 19:56 |
| lifeless | soren: neutron_notes.rst is waaaay old. I suspect we should just delete it | 19:58 |
| soren | devananda: Ok, so I'll just forget I ever saw that doc. | 19:58 |
| soren | Any hints at all for how I can set up Neutron to make Ironic (and me!) happy? | 19:58 |
| lifeless | soren: so the automation in setup-neutron for either devtest_seed.sh or devtest_undercloud.sh should do what you want | 19:59 |
| NobodyCam | LOL who is: "Welcome, new contributor!" listed on https://review.openstack.org/#/c/100951 | 19:59 |
| * soren stares at setup-neutron for a while | 19:59 | |
| jroll | NobodyCam: isn't that awesome? :) | 20:00 |
| jroll | it posts for new contributors now | 20:00 |
| *** max_lobur has quit IRC | 20:00 | |
| Nisha | NobodyCam: Its my first contribution to openstack :) | 20:00 |
| *** krtaylor has quit IRC | 20:02 | |
| soren | lifeless: What's the difference between seed and undercloud? | 20:02 |
| *** bvivek has quit IRC | 20:03 | |
| NobodyCam | Nisha: congratz!!!! and welcome to openstack community | 20:05 |
| Nisha | NobodyCam: :) | 20:06 |
| NobodyCam | Nisha: looking at 103007.. what info is in chche status? | 20:07 |
| Nisha | NobodyCam: chche? | 20:07 |
| NobodyCam | Cache Information :-p | 20:08 |
| *** krtaylor has joined #openstack-ironic | 20:08 | |
| NobodyCam | sorry fingers faster then brain | 20:08 |
| Nisha | NobodyCam: Cache Information will hold L1, L2, L3 cache sizes | 20:11 |
| Nisha | :) | 20:11 |
| soren | lifeless: Sorry, I'm having trouble following the code.. setup-neutron creates all the required resources... But where does Neutron get its configuration? | 20:11 |
| NobodyCam | ahh so Cpu Cache info | 20:12 |
| Nisha | Yes | 20:12 |
| NobodyCam | or is that also for disk? | 20:12 |
| Nisha | No only for processors | 20:13 |
| lifeless | soren: what configuration? | 20:17 |
| *** geekyogi has quit IRC | 20:19 | |
| lifeless | Nisha: do you expect people to schedule on that? | 20:20 |
| lifeless | soren: seed and undercloud are the same from a neutron perspective | 20:20 |
| soren | lifeless: You know.. The thing that configuration files specify. | 20:21 |
| soren | lifeless: Like.. Use these classes for this kind of nonsense, these other classes for other kinds of nonsense, and here are my environmental details that everything must submit to. | 20:22 |
| lifeless | oh, defaults | 20:22 |
| lifeless | the only thing you need config file wise for neutron is provider network mappings | 20:22 |
| lifeless | bridge_mappings = ctlplane:br-ctlplane | 20:22 |
| lifeless | network_vlan_ranges = ctlplane | 20:22 |
| lifeless | in your ovs config file | 20:23 |
| soren | Ok, so openvswitch is given? | 20:23 |
| lifeless | there's no hard dependency | 20:23 |
| lifeless | should work with any plugin you want | 20:23 |
| lifeless | we've just never used downlevel ones, or proprietary ones :) | 20:23 |
| soren | "downlevel"? | 20:24 |
| lifeless | less capable | 20:24 |
| lifeless | eg linuxbridge | 20:24 |
| NobodyCam | Nisha: for HP ProLiant servers via iLO using iLO client python library? should that be ... using Ironic python library? | 20:24 |
| devananda | NobodyCam: no - there is an hp proliant python lib on github | 20:25 |
| NobodyCam | devananda: ever where else in the spec it reffered to as Depends on proliantutils library. | 20:25 |
| devananda | NobodyCam: ah. that wording may be confusing then | 20:26 |
| NobodyCam | s/proliantutils library.// | 20:26 |
| NobodyCam | :) | 20:26 |
| NobodyCam | will leave comment on spec | 20:26 |
| Nisha | NobodyCam: Ok :) | 20:26 |
| lifeless | soren: you're testing out ironic ? | 20:26 |
| soren | lifeless: I'm trying to. Desperately. | 20:27 |
| *** mdorman has quit IRC | 20:27 | |
| Nisha | lifeless: I didnt get your question? is it for cache info | 20:27 |
| Nisha | ? | 20:27 |
| *** dwalleck has joined #openstack-ironic | 20:27 | |
| lifeless | yes | 20:27 |
| lifeless | Ironic isn't a general purpose machine database | 20:28 |
| lifeless | I'm wondering why you want to collect so much information | 20:28 |
| Nisha | lifeless: if thats the case i can remove that info i.e. cache info , rest is already included in the generic spec. | 20:29 |
| *** Penick has quit IRC | 20:30 | |
| NobodyCam | Shrews: 103670 +a'd | 20:30 |
| NobodyCam | devananda: ^^^^ | 20:30 |
| jroll | \o/ | 20:31 |
| Shrews | woohoo | 20:31 |
| NobodyCam | still has to get through the gate .. lol | 20:31 |
| *** dwalleck has quit IRC | 20:32 | |
| NobodyCam | Nisha: reviewed | 20:33 |
| Nisha | NobodyCam: Thanks . Just for one comment. I am removing cache info as per lifeless suggestion as of now. If needed that can be added later on. | 20:34 |
| NobodyCam | sure | 20:34 |
| lifeless | Nisha: that wasn't my suggestion. I was asking *why* | 20:34 |
| lifeless | Nisha: I'll be asking that on the generic one too | 20:35 |
| Nisha | lifeless: generic one doesnt mention that attribute | 20:35 |
| Nisha | only ilo specific i added it as ilo can discover that information. I am not sure how that property can be used by people later on , may not apply for scheduling | 20:36 |
| Nisha | for scheduling only the required ones get updated | 20:36 |
| lifeless | Nisha: what review # is the generic one? | 20:37 |
| NobodyCam | brb quick walkies | 20:37 |
| Nisha | lifeless: A driver may chose to discover more attributes than listed in generic one , correct? | 20:37 |
| NobodyCam | lifeless: 100951 | 20:37 |
| Nisha | lifeless: https://review.openstack.org/#/c/100951 | 20:37 |
| Nisha | NobodyCam: Thanks :) | 20:38 |
| soren | lifeless: Do I have to create br-ctlplane? Or will something do that for me? | 20:40 |
| *** Penick has joined #openstack-ironic | 20:43 | |
| NobodyCam | jroll: happen to be around? | 20:44 |
| jroll | sup NobodyCam | 20:44 |
| NobodyCam | was looking at your comment on 103136 | 20:44 |
| jroll | yeah | 20:44 |
| jroll | I forgot to loop back and check that out | 20:45 |
| NobodyCam | have you looked at https://github.com/openstack/ironic/blob/master/ironic/nova/scheduler/filters/exact_ram_filter.py | 20:45 |
| NobodyCam | line 29 | 20:45 |
| jroll | yes | 20:45 |
| NobodyCam | that was my concern | 20:45 |
| jroll | so that looks at host_state.free_ram_mb | 20:45 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic-specs: Add support retry on NodeLocked exceptions. https://review.openstack.org/103996 | 20:45 |
| jroll | and the dict we return in get_info() doesn't have that key | 20:46 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 20:46 |
| jroll | so I think it's unrelated | 20:46 |
| jroll | but /me looks at code | 20:46 |
| jroll | seems like it's related to get_available_resource, not get_info | 20:47 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic-specs: Add support for retry on NodeLocked exceptions https://review.openstack.org/103996 | 20:47 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties for iLO driver https://review.openstack.org/103007 | 20:48 |
| jroll | NobodyCam: here's what the filter looks at https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L190 | 20:48 |
| Nisha | NobodyCam: Have just updated the spec for iLO driver and posted back :) | 20:49 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission interfaces https://review.openstack.org/102953 | 20:50 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add Node decommission in conductor https://review.openstack.org/102984 | 20:50 |
| soren | Am I supposed to create br-ctlplane myself? Or is OVS supposed to be told about it from neutron and then it'll create it? | 20:50 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic-specs: Add support for retry on NodeLocked exceptions https://review.openstack.org/103996 | 20:51 |
| *** jbjohnso_ has quit IRC | 20:52 | |
| NobodyCam | jroll: ahh | 20:53 |
| jroll | :) | 20:54 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic-specs: Add support for retry on NodeLocked exceptions https://review.openstack.org/103996 | 20:55 |
| NobodyCam | jroll: changed my vote | 20:55 |
| jroll | \o/ | 20:56 |
| *** jdob has quit IRC | 20:56 | |
| Shrews | heading out for the day. good evening @all | 20:56 |
| jroll | night Shrews | 20:57 |
| Nisha | lifeless: NobodyCam Leaving for tonight. Its already early morning here. (3 am) :) | 20:57 |
| Nisha | lifeless: will see your comments in the spec and if needed will discuss tomorrow. | 20:58 |
| NobodyCam | have a good night Nisha | 20:58 |
| NobodyCam | anyone happen to know the bug number for recheck on the glance endpoint issue | 20:59 |
| jroll | did we make one? | 21:00 |
| *** ccrouch1 has quit IRC | 21:00 | |
| *** ccrouch has joined #openstack-ironic | 21:01 | |
| matty_dubs | Is that the "AttributeError: 'Client' object has no attribute 'endpoint'" one? I was _just_ noticing that it keeps coming up | 21:01 |
| jroll | yes | 21:01 |
| matty_dubs | (I don't have the number, though.) | 21:01 |
| jroll | NobodyCam: there wasn't a bug on that patch, so I would assume there's no bug for this issue | 21:01 |
| NobodyCam | :-p | 21:01 |
| * NobodyCam is trying to not use no bug | 21:02 | |
| *** Nisha has quit IRC | 21:02 | |
| NobodyCam | humm use no bug or add a bug (after the fix is already approved) to track the rechecks? | 21:03 |
| * NobodyCam notes to dbl check for bug # on patches that are fixing bugs | 21:04 | |
| NobodyCam | brb | 21:04 |
| jroll | +1 | 21:04 |
| *** ccrouch has quit IRC | 21:05 | |
| openstackgerrit | A change was merged to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 21:05 |
| matty_dubs | ^ rock | 21:06 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 21:07 |
| matty_dubs | Oops, I'm an idiot | 21:08 |
| matty_dubs | I put 103670 as the bug # | 21:08 |
| matty_dubs | That's a bug for a SIGSEGV in rhythmbox from 2007 | 21:08 |
| *** max_lobur has joined #openstack-ironic | 21:10 | |
| NobodyCam | lol someone is going to get a strange email :-p | 21:10 |
| * matty_dubs succeeds in making the world a more interesting place ;) | 21:11 | |
| *** dwalleck has joined #openstack-ironic | 21:12 | |
| matty_dubs | NobodyCam: Obviously, "Fix the section name in CONTRIBUTING.rst" cannot be merged until rhythmbox is fixed | 21:13 |
| *** ccrouch has joined #openstack-ironic | 21:14 | |
| matty_dubs | Having successfully sown confusion, I've now got to head out for the day. Have a great afternoon/evening, all. | 21:16 |
| *** matty_dubs is now known as matty_dubs|gone | 21:16 | |
| NobodyCam | lol have a good day matty_dubs|gone | 21:17 |
| NobodyCam | s/day/night/ | 21:17 |
| *** dwalleck has quit IRC | 21:20 | |
| *** dwalleck has joined #openstack-ironic | 21:21 | |
| lifeless | hey so | 21:22 |
| lifeless | this removal of ramdisk selection from flavors | 21:22 |
| lifeless | where is the spec for it? I"d like to argue very much against it | 21:22 |
| lifeless | or at least make sure I really understand why - it seems entirely unnecessary today | 21:23 |
| lifeless | and we're going to make activating a new deploy ramdisk a very slow process (min one call per node) | 21:23 |
| *** dwalleck has quit IRC | 21:26 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add a decommission command for the agent https://review.openstack.org/103694 | 21:26 |
| jroll | lifeless: I think that's not happening until L or something... giving it deprecation time in K | 21:27 |
| jroll | lifeless: that review is -2'd, no? | 21:27 |
| *** dwalleck has joined #openstack-ironic | 21:29 | |
| *** jrist has quit IRC | 21:30 | |
| lifeless | jroll: devananda filed a bug about it in tripleo, saying that it was deprecated, not was-to-be deprecated | 21:30 |
| lifeless | jroll: I'm arguing we don't need to deprecate it at all, and shouldn't. | 21:31 |
| devananda | lifeless: ah. perhaps my wording was unclear. we're supporting both right now | 21:31 |
| devananda | lifeless: the spec was already discussed in meetings and approved, a while ago | 21:31 |
| lifeless | so I want to have a discussion about why we feel we need to deprecate it | 21:31 |
| devananda | lifeless: because flavor != driver | 21:31 |
| devananda | and different drivers require different ramdisks (IPA and iLO and PXE deploy disk images are not compatible with each other) | 21:32 |
| lifeless | devananda: that seems entirely orthogonal | 21:32 |
| devananda | not at all | 21:32 |
| devananda | well | 21:32 |
| devananda | they are | 21:32 |
| lifeless | devananda: I still don't see the problem. Perhaps in spec form. 'Because X, Y'. | 21:32 |
| devananda | flavor and driver are orthogonal | 21:32 |
| devananda | but putting the deploy ramdisk on the flavor means it is tied to a driver | 21:32 |
| lifeless | why does it mean that? | 21:32 |
| devananda | which the nova scheduler is unaware of | 21:32 |
| lifeless | I don't see how it means that. | 21:32 |
| devananda | 21:32:12 < devananda> and different drivers require different ramdisks (IPA and iLO and PXE deploy disk images are not compatible with each other) | 21:32 |
| lifeless | So? Still not seeing the connection. I must be really dumb or something. | 21:33 |
| devananda | i'll walk through it | 21:33 |
| devananda | user requests flavor X, which specifies a deploy ramdisk taht is compatible with IPA | 21:33 |
| lifeless | wait a sec | 21:34 |
| lifeless | it specifically says | 21:34 |
| lifeless | baremetal:deploy_kernel_id and baremetal:deploy_ramdisk_id, right ? | 21:34 |
| lifeless | those were the nova-bm key's, which specify nova-baremetal deploy ramdisks. | 21:34 |
| devananda | which are compatible with the PXE driver | 21:35 |
| lifeless | We have a bug right now - that this conversation has made me realise - that Ironic is incompatible with nova-BM | 21:35 |
| devananda | oh? | 21:35 |
| lifeless | because the ironic deploy ramdisk requires the token accessed over tftp | 21:35 |
| devananda | oh?ahhh right | 21:35 |
| devananda | :( | 21:35 |
| devananda | so | 21:35 |
| devananda | that's also covered in my spec proposal to nova | 21:35 |
| lifeless | its why we have a baremetal-deploy and ironic-deploy element in dib | 21:35 |
| devananda | for the upgrade path | 21:35 |
| lifeless | one fix for that would be to say ironic:pxe-deploy-kernel-id and ironic:pxe-deploy-ramdisk-id | 21:36 |
| devananda | https://review.openstack.org/#/c/95025/ | 21:36 |
| lifeless | if we do that it should be clear how IPA can coexist | 21:36 |
| devananda | lifeless: so this has me thinking, perhaps another way would be to specify the list of all possible k&r pairs for any driver in use on the flavor | 21:36 |
| devananda | except again this is really exposing an abstraction layer violation in Ironic's API | 21:37 |
| devananda | by requiring nova (a client in this case) to pass in driver-specific info | 21:37 |
| devananda | the architectural reason here is: these valuse are as internal and operator-centric as the IPMI credentials | 21:37 |
| devananda | and should not be changed on a node between one deployment and the next | 21:37 |
| lifeless | I don't agree with your last statement | 21:38 |
| lifeless | the rest I agree with | 21:38 |
| devananda | they can be changed when teh operator is upgrading their deploy environment | 21:38 |
| devananda | but not by a client | 21:38 |
| *** dwalleck has quit IRC | 21:38 | |
| lifeless | sure | 21:38 |
| lifeless | Anyhow, my main beef is | 21:38 |
| devananda | if at some point we want t oseparate a "management" and a "deploy" interface -- from a policy/provilege perspective -- I think we need to do this | 21:39 |
| lifeless | I don't want to have to run a script over 1500 nodes when I build a new deploy ramdisk | 21:39 |
| lifeless | the current 'fix' is a big step backwards in efficiency | 21:39 |
| jroll | ... why not? that's not that large of a load | 21:39 |
| jroll | as you say, it's scriptable | 21:39 |
| jroll | pushing a new button is not a big step backwards | 21:39 |
| lifeless | jroll: aesthetics? room for failure? more moving parts? | 21:40 |
| devananda | lifeless: if you change it on the flavor, nova will just be doing that for you | 21:40 |
| devananda | it's no less efficient in terms of API calls to ironic | 21:40 |
| devananda | it merely shifts the point-in-time to be more clearly visible to the operator regarding which node is being updated | 21:41 |
| devananda | making it, IMO, easier to manage, not harder | 21:41 |
| lifeless | devananda: its a) local (so way less latency*count hit), amortised over time (only hits when you do a deploy) and wrapped in retry etc logic so invisible to me | 21:41 |
| *** linggao has quit IRC | 21:41 | |
| lifeless | devananda: I've shifted from arguing *for* flavor to arguing about the specific thing that concerns me. I recognise that user-definable flavors make this a security issue. | 21:42 |
| devananda | ack | 21:42 |
| devananda | lifeless: so efficiency is "the thing that concerns you" ? | 21:42 |
| lifeless | the shift from define-once to define-N times; the denormalisation. | 21:43 |
| lifeless | efficiency is part of it | 21:43 |
| lifeless | the ability to have races where one node doesn't get updated because it was added by process A while process B was updating the ramdisk references is another | 21:43 |
| lifeless | which means from an ops perspective I'll have to write a tool to assert 'machines with shape X get ramdisk Y' and do that in a big loop checking for exceptions and fixing them. | 21:44 |
| lifeless | ^ that worries me | 21:44 |
| lifeless | Oh, and to permit testing new ramdisks I'll need to have exceptions in that tool to carve out test machines temporarily | 21:45 |
| lifeless | and use scheduler hints to do the testing | 21:45 |
| lifeless | that also worries me | 21:45 |
| lifeless | with flavor, I create a new flavor referencing the new ramdisk, do a deploy to it, verify it works, then delete the flavor and update the public flavor, done. | 21:46 |
| lifeless | ^ I want that easy a story. | 21:46 |
| devananda | lifeless: if you can create new flavors, then you probably also have access to the force-host extension | 21:51 |
| lifeless | devananda: which is way less easy. I have to query ironic for free nodes, and then race against the scheduler. | 21:52 |
| lifeless | devananda: I kindof feel like you're arguing purity and I'm arguing user experience. | 21:52 |
| devananda | lifeless: i think that's a fair statement | 21:52 |
| devananda | lifeless: which is an interesting turn of roles for us :) | 21:52 |
| NobodyCam | :O | 21:53 |
| lifeless | possibly; I do try to bear both things in mind always | 21:53 |
| devananda | lifeless: i'm also arguing in favor of increased transparency and privilege separation. | 21:53 |
| lifeless | sometimes its harder than others. | 21:53 |
| * BadCub watches intently | 21:53 | |
| lifeless | devananda: I don't see that it increases transparency at all. with flavor (as one way to get UX) its entirely transparent. | 21:54 |
| devananda | carrying ironic-driver-specific information in an external system (nova's flavors) is illogical | 21:54 |
| * NobodyCam now has feel he's being watched | 21:54 | |
| NobodyCam | s/feel/feeling/ | 21:54 |
| lifeless | devananda: I agree. I'm not arguing for flavor. I'm arguing that we don't make the UX worse when addressing the concerns flavor has for us. | 21:54 |
| devananda | lifeless: the UX you seem to want is one that uses groups to associate metadata with nodes | 21:55 |
| devananda | lifeless: such that you can atomically change driver-specific metadata for all nodes in the group | 21:55 |
| devananda | lifeless: i would support that | 21:55 |
| devananda | if it were implemented within ironic | 21:55 |
| lifeless | devananda: thats one facet of it yes. Another facet is letting me run with different driver specific metadata *on a nova boot request*, to let me test the impact of changes without making global config changes. | 21:55 |
| devananda | lifeless: but that's a layer violation | 21:56 |
| lifeless | devananda: not necessarily. | 21:56 |
| devananda | lifeless: rephrase, as I parsed taht too quickly | 21:56 |
| devananda | lifeless: you *can* do taht today | 21:56 |
| lifeless | devananda: I was very careful to phrase it in a way that doesn't presume one implementation. | 21:56 |
| devananda | lifeless: by changing the driver specific metadata on a specific node, then requesting nova to boot *that node* | 21:56 |
| lifeless | devananda: requesting nova to boot *that node* is a layer violation in it's own right. And its racy with the scheduler. | 21:57 |
| devananda | indeed, but that's another topic :) | 21:57 |
| lifeless | devananda: unless you put the node into maintenance first - we're stacking headache on headache. | 21:57 |
| devananda | lifeless: so, placing ironic driver metadata anywhere but inside ironic is counter-intuitive | 21:57 |
| devananda | lifeless: or you set the node properties such that it only matches your "i'm an admin, here's my test flavor" flavor | 21:57 |
| NobodyCam | devananda: not to distract. I'm looking at https://review.openstack.org/#/c/92416 (Add the remaining unittests to the ClientWrapper class) in the nova section of our tree.. so checking with you if that's one to land. | 21:57 |
| devananda | which would be easy to do if there were an extra_specs check in the scheduler | 21:58 |
| lifeless | mmm, I think thats too broad a statement. Consider port details that might live in neutrons topology db. But thats a distraction :) | 21:58 |
| devananda | find any free node, ironic node-update add extra/admin_only=True | 21:58 |
| devananda | or what ever | 21:58 |
| *** jgrimm has quit IRC | 21:58 | |
| lifeless | devananda: compare that to the UX we have today. Ignore the illogical. | 21:58 |
| devananda | lifeless: i can't ignore that | 21:58 |
| lifeless | devananda: I'm asking you to hold the UX to *that standard*, in assessing replacements. | 21:58 |
| devananda | lifeless: the UX changes, but it's not *worse* | 21:59 |
| lifeless | devananda: yes you can, you just don't want to. | 21:59 |
| lifeless | :) | 21:59 |
| lifeless | devananda: my perception of it is that its worse. | 21:59 |
| lifeless | devananda: I wish you'd been active in the tripleo region administration over the last 9 months, I think we wouldn't be having this discussion like this. | 21:59 |
| devananda | lifeless: I wish more tripleo region admins had been active in Ironic over the last 9 months -- their input is invaluable to me | 22:00 |
| lifeless | devananda: been doing my best... | 22:00 |
| devananda | lifeless: but tripleo isn't the only user (or use case) for ironic | 22:01 |
| devananda | lifeless: and i've been tracking as much of each user and use-case as I can ... | 22:01 |
| lifeless | devananda: I know that; the key word there wasn't tripleo, it was region admin | 22:01 |
| *** mrda-away is now known as mrda | 22:02 | |
| mrda | Morning all | 22:02 |
| NobodyCam | morning mrda | 22:03 |
| *** romcheg has left #openstack-ironic | 22:03 | |
| jroll | heya mrda | 22:03 |
| mrda | \o | 22:04 |
| lifeless | devananda: so where do we go from here? I feel like I've failed at helping you understand the operational impact | 22:04 |
| jroll | lifeless: I'm still not sure I see this as a problem, in terms of operational impact. you need to update the kernel/ramdisk id in ironic rather than nova (this can even be in mysql if you don't want to put load on the api). testing in production is harder (that's why test/staging environments exist). anything I'm missing? | 22:06 |
| jroll | lifeless: in our environment, we need to reboot unprovisioned nodes when we update the ramdisk, and it hasn't caused us any pain | 22:07 |
| lifeless | jroll: 'testing in production is harder' <- | 22:07 |
| lifeless | jroll: many smaller environments won't have test/staging baremetal labs | 22:08 |
| jroll | that's what staging environments are for | 22:08 |
| lifeless | jroll: many smaller environments won't have test/staging baremetal labs | 22:08 |
| lifeless | jroll: and won't be willing to have them | 22:08 |
| lifeless | since the cost / utilisation is too high/too low | 22:08 |
| jroll | ... really? we're supporting users insane enough to use production to test code? | 22:09 |
| jroll | if you're at the point of automating bare metal provisioning | 22:09 |
| jroll | you have, say, $20k to spend on a small staging environment | 22:09 |
| JayF | Eh, I don't think that's an unreasonable as you make it seem jroll | 22:09 |
| jroll | let me put it this way | 22:10 |
| JayF | I think we should have ironic be easy enough to use that someone with <100 physical servers could use it to manage boxes | 22:10 |
| jroll | if you're willing to go through the pain of testing new code in production | 22:10 |
| jroll | you're probably willing to go through the pain of using scheduler hints to test | 22:10 |
| jroll | I would hope | 22:10 |
| JayF | However in context of this discussion, I do agree that specifying deploy ramdisk seems like a node specific thing, and specifically well targeted for driver_info | 22:10 |
| NobodyCam | hummm #notice should show up on meeting summary | 22:11 |
| lifeless | jroll: welcome to users. | 22:12 |
| lifeless | jroll: who said anything about testing new code in production. | 22:12 |
| lifeless | jroll: there's this thing called change control | 22:12 |
| JayF | deploy ramdisk and kernel is code, at least to us | 22:12 |
| lifeless | sure | 22:12 |
| lifeless | but consider a user | 22:12 |
| lifeless | upstream release a new version | 22:13 |
| JayF | I am a user fwiw as well ;) | 22:13 |
| lifeless | a user that isn't a dev | 22:13 |
| lifeless | they get a new ramdisk + kernel from their vendor | 22:13 |
| lifeless | RH, HP, RAX, whoever. | 22:13 |
| lifeless | anyhow, I get whinged at by product folk for having 2 extra nodes for an HA management plane. | 22:14 |
| lifeless | an extra full set of infrastructure for a staging environment in customer premises... I would be laughed out of the country | 22:15 |
| JayF | /o\ | 22:15 |
| lifeless | I mean, there are some customers where that totally makes sense. But there are many where it doesn't. | 22:16 |
| JayF | We have an entire cab of each hardware profile in our staging environment currently :) | 22:16 |
| lifeless | where the rate of change would lead to such infrastructure being 99% idle, and it being a high fraction of the total hardware pool | 22:16 |
| JayF | and I know that's somewhat a luxury, because I've worked ops for a small enough place that 'staging' was a couple of servers running vms, and was code-only | 22:16 |
| lifeless | yup | 22:17 |
| lifeless | ok, time to go focus on vlan support in tripleo | 22:17 |
| lifeless | au revoir | 22:17 |
| JayF | However, I don't think that that using hinted schedulers, or adding information to a node (like node/beta=true to get it to ID to a specific flavor) in ironic is too onerous | 22:18 |
| JayF | I was actually very surprised to see that driver-specific things leaked back into nova flavors | 22:18 |
| devananda | JayF: that was a vestige of nova-baremetal | 22:24 |
| devananda | JayF: when it was all in nova anyway | 22:24 |
| JayF | that makes sense | 22:25 |
| JayF | You have to remember I live in a post-ironic world ;) | 22:26 |
| JayF | er, I mean, post-bm? i.e. whatever the right term would be for ironic has always existed :) | 22:26 |
| *** chuckC has quit IRC | 22:26 | |
| NobodyCam | ˁ˚ᴥ˚ˀ <- does that look like a Polar Bear | 22:27 |
| jroll | lol, looks like a black lab from here | 22:28 |
| jroll | or a black bear | 22:28 |
| NobodyCam | lol | 22:28 |
| JayF | it looks like a guy with really small ears | 22:28 |
| JayF | and no mouth | 22:28 |
| lifeless | I'm with JayF on that | 22:28 |
| NobodyCam | from: http://1lineart.kulaone.com | 22:28 |
| jroll | oh I love this site | 22:28 |
| lifeless | JayF: sure, I'm not arguing for flavor, its just a reference point to say 'that was easy, can I please still have easy' | 22:29 |
| JayF | lifeless: something we meant to mention in the blog post but didn't, is that OnMetal is 'triple o' (in the philisophical sense), becuase we have our entire Ironic control plane running on our internal installation of Nova | 22:29 |
| JayF | the only physical machines that are running for our product are the machines that are the product | 22:29 |
| lifeless | JayF: nice | 22:29 |
| *** radsy has joined #openstack-ironic | 22:30 | |
| * NobodyCam clicks back to reviews | 22:30 | |
| lifeless | JayF: do you run a dedicated cell or region for OnMetal ? | 22:30 |
| jroll | NobodyCam: those are useful for reviews you know :) | 22:30 |
| JayF | lifeless: we run it inside it's own cell | 22:30 |
| NobodyCam | lol | 22:30 |
| jroll | NobodyCam: (>ლ) | 22:30 |
| jroll | also this is very relevant for some reviews: ( •_•)O*¯`·.¸.·´¯`°Q(•_• ) | 22:31 |
| NobodyCam | lol | 22:31 |
| NobodyCam | ︻╦╤─ \(^-^)/ <- gun to head | 22:32 |
| jroll | yeah, I've been iterating on that one outside of channel | 22:33 |
| jroll | I like this better: ༼ つ ◕_◕ ༽つ︻デ┳═ー (╯︵╰,) | 22:33 |
| devananda | $W(*TU($% | 22:39 |
| * devananda drops his laptop | 22:39 | |
| NobodyCam | LOL | 22:40 |
| NobodyCam | brb | 22:40 |
| *** derekh_ has quit IRC | 22:41 | |
| jroll | hahaha | 22:43 |
| BadCub | Ouch | 22:49 |
| *** ccrouch has quit IRC | 22:51 | |
| NobodyCam | adam_g: happen to be around ish? | 22:54 |
| *** athomas has quit IRC | 22:58 | |
| *** chuckC has joined #openstack-ironic | 23:07 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Fix email in setup.cfg https://review.openstack.org/104015 | 23:10 |
| * NobodyCam just gave his first -2 to a patch | 23:10 | |
| jroll | \o/ | 23:11 |
| NobodyCam | lol makes me sad | 23:11 |
| NobodyCam | :( | 23:11 |
| jroll | aww | 23:11 |
| jroll | you can +2 that one I just uploaded :P | 23:11 |
| jroll | to even things out | 23:11 |
| NobodyCam | lol hahhaha | 23:11 |
| mrda | jroll: subtle | 23:11 |
| jroll | tis easy :) | 23:11 |
| jroll | if someone -1's that patch... they have issues | 23:12 |
| NobodyCam | lol ... jroll is this land my patch because I know where you live... | 23:12 |
| jroll | lol | 23:13 |
| jroll | no, really | 23:13 |
| jroll | please +2 | 23:13 |
| jroll | https://review.openstack.org/104015 | 23:13 |
| jroll | I fixed the email in setup.cfg | 23:13 |
| NobodyCam | jroll: I could -1 it | 23:15 |
| NobodyCam | https://github.com/openstack/nova/blob/master/setup.cfg#L7 | 23:15 |
| jroll | ... | 23:15 |
| jroll | file a bug :P | 23:15 |
| * jroll fixes | 23:15 | |
| NobodyCam | lol | 23:15 |
| * NobodyCam is reviewed out at this point | 23:16 | |
| NobodyCam | lol | 23:16 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Fix author/email in setup.cfg https://review.openstack.org/104015 | 23:16 |
| jroll | one more? | 23:16 |
| jroll | please? | 23:16 |
| NobodyCam | heheheh | 23:16 |
| jroll | (really just want to test an infra job) | 23:16 |
| NobodyCam | sure... | 23:16 |
| jroll | JoshNang: please review again ^ | 23:17 |
| NobodyCam | devananda: when you have a free minute: https://review.openstack.org/#/c/103137 ready to land.. But is in the nova section of our repo.. | 23:17 |
| JoshNang | jroll: +2 | 23:17 |
| jroll | danke | 23:18 |
| NobodyCam | +a'd | 23:19 |
| NobodyCam | :) | 23:19 |
| NobodyCam | it was an easy one | 23:19 |
| jroll | thanks :) | 23:19 |
| * jroll stares at zuul | 23:19 | |
| NobodyCam | lol | 23:19 |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Fix author/email in setup.cfg https://review.openstack.org/104015 | 23:20 |
| kylestev | jroll: ^ \o/ | 23:22 |
| jroll | whee | 23:22 |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/102126 | 23:32 |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic-python-agent: Updated from global requirements https://review.openstack.org/88722 | 23:32 |
| devananda | Shrews: replied on https://review.openstack.org/#/c/103162/ | 23:40 |
| NobodyCam | in an effort to do some house cleaning: Anyone with a free minute and an intrest in helping clear the Old Review of the Day (ORD or ORotD) please have a look over: https://review.openstack.org/#/c/97731 | 23:53 |
| * NobodyCam notes he just made up ORotD it is not offical in any way | 23:54 | |
| jroll | should have just run with it | 23:54 |
| jroll | :P | 23:54 |
| NobodyCam | lol | 23:54 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
