=== shardy is now known as shardy_lunch | ||
uppock | Does cloud init support setting up ipv6 via dhcp6. Im having problems getting the correct network config in a ubuntu intance on openstack | 13:19 |
---|---|---|
=== shardy_lunch is now known as shardy | ||
jgrimm | thanks rbasak | 14:11 |
smoser | uppock, if you use config drive on openstack and it provides network_data.json then it should work. | 14:20 |
smoser | however, if you just want to do "dhcp v6 on the first nic", then there is not currently a way to do that without modifying an image. | 14:21 |
smoser | i recently just did this in bug | 14:21 |
smoser | 14:21 | |
smoser | https://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/1657940/comments/4 | 14:21 |
smoser | see the 'sudo tee' line there. thats how you can configure the system to do it for a spicific NIC by mac. i am pretty sure if you know the name the nic will get you can drop the mac | 14:22 |
=== rangerpbzzzz is now known as rangerpb | ||
uppock | Oh | 14:49 |
uppock | I do have network_data.json via the metadata server, but that does not help? | 14:51 |
smoser | uppock, currently no it does not. | 14:58 |
smoser | that is a change that can be done, but its not working now. | 14:58 |
uppock | smoser: ok | 15:02 |
uppock | thanks for the help | 15:16 |
smoser | rharper, https://code.launchpad.net/~wesley-wiedenmeier/cloud-init/+git/cloud-init/+merge/317589 | 16:40 |
smoser | so... | 16:40 |
smoser | NETWORK_CONFIG_V2 is a bad name i think | 16:40 |
smoser | NETPLAN_PASSTHROUGH ? | 16:40 |
smoser | no. its not passthrough. | 16:40 |
smoser | what shoudl we call that. | 16:40 |
=== logan_ is now known as Guest29589 | ||
smoser | rharper, also i had https://code.launchpad.net/~smoser/cloud-init/+git/cloud-init/+merge/319506 for you | 16:57 |
smoser | which is terribly rendered there because i rebased to master | 16:58 |
rharper | smoser: I don't have a better suggestion; I had talked with magicalChicken about having maybe a list of network renderers; as well as paththrough features | 16:58 |
rharper | smoser: oh, the ntp test | 17:00 |
smoser | hm.. i guess the name isnt terrible in itsel.f | 17:01 |
smoser | as i think cloud-init is basically stating that it can render | 17:01 |
smoser | http://paste.ubuntu.com/24152513/ | 17:01 |
smoser | which is "network config v2" | 17:01 |
smoser | an interesting thing though is that we cannot render the sysconfig for that yet, right? | 17:02 |
rharper | not yet | 17:02 |
smoser | we have a generic v2-to-v1 ? | 17:02 |
rharper | yes | 17:02 |
smoser | so we could v2-to-v1-to-sysconfig | 17:02 |
smoser | outside of features not supported. | 17:03 |
rharper | yes, v2 to network-state, and network-state to *whatever* | 17:03 |
smoser | right | 17:03 |
rharper | this branch adds a v2 parser -> netstate; and then a state -> netplan | 17:03 |
=== Guest29589 is now known as logan- | ||
rharper | we have a v2 -> output skipping state | 17:04 |
rharper | aka, 'passthrough' | 17:04 |
powersj | smoser: test regression wrt chpasswd last night LP: #1671883 | 17:10 |
powersj | https://bugs.launchpad.net/cloud-init/+bug/1671883 | 17:10 |
smoser | powersj, should be fixed in trunk | 17:11 |
powersj | smoser: oh look at that :) | 17:12 |
powersj | I'll launch the tests again then. thanks! | 17:12 |
smoser | yeah, horay for that integration test. | 17:13 |
smoser | rharper, https://code.launchpad.net/~smoser/cloud-init/+git/cloud-init/+merge/319605 | 17:13 |
smoser | magicalChicken, ^ | 17:14 |
smoser | that is simplified and more-inert version of https://code.launchpad.net/~wesley-wiedenmeier/cloud-init/+git/cloud-init/+merge/317589 | 17:14 |
magicalChicken | smoser: calling it NETWORK_CONFIG_V1 might be confusing though, as there are versions of cloud-init with network config support but without the feature flags | 17:31 |
powersj | smoser: https://paste.ubuntu.com/24152754/ | 17:31 |
smoser | "it" ? | 17:31 |
rharper | smoser: would we fill that out with other input formats we support? I sorta liked the idea of specifying the list of inputs and outputs supported | 17:31 |
magicalChicken | smoser: sorry, the feature for network support | 17:31 |
smoser | magicalChicken, there, i just took your suggestion, but currently if i merged this into trunk, we do not support NETWORK_CONFIG_V2 | 17:32 |
smoser | but we *do* support NETWORK_CONFIG_V1 | 17:32 |
smoser | so i listed that support as a feature | 17:32 |
magicalChicken | smoser: oh, right that makes sense | 17:32 |
magicalChicken | yeah, it may be good t fill out that list | 17:32 |
magicalChicken | the mp i had filed had been into ~raharper/cloud-init:netconfig-v2-passthrough, once that branch lands in trunk it will support v2 though | 17:33 |
smoser | magicalChicken, right.l but i was just grabbing the "add feature list" | 17:34 |
magicalChicken | yeah, that makes sense | 17:34 |
smoser | rharper, i dont know what to do there.. | 17:34 |
smoser | wrt listing "sysconfig" or "ifupdown" | 17:35 |
rharper | right; those might be per distro | 17:35 |
smoser | i think if we follow curtin's lead, we end up with: | 17:35 |
magicalChicken | i like the features output command | 17:35 |
smoser | NETWORK_CONFIG_V1_IFUPDOWN | 17:35 |
rharper | it could be a KEY into a list of distros that support those renderes | 17:35 |
smoser | and | 17:35 |
smoser | NETWORK_CONFIG_V1_SYSCONFIG | 17:36 |
smoser | unfortunately, that then possibly leads to | 17:36 |
smoser | NETWORK_CONFIG_V1_SYSCONFIG_WITH_BONDS | 17:36 |
smoser | or something like that | 17:36 |
magicalChicken | rharper: i think we had discussed this before, the per-distro features vs global features | 17:36 |
rharper | RENDERERS = 'eni, sysconfig, netplan, netcfg'; distro_renderers = { 'Ubuntu': ['eni', 'netplan'], 'Fedora': ['sysconfig'] } | 17:36 |
rharper | yeah | 17:36 |
smoser | powersj, [embarrased] how do i run integration test for specific test ? | 17:42 |
smoser | rharper, so i dont want to over complicate things. | 17:42 |
powersj | As an example: python3 -m tests.cloud_tests run -v -n xenial -t tests/cloud_tests/configs/modules/set_password.yaml --deb cloud-init_daily_all.deb | 17:43 |
smoser | i dont think t here is any reason to query cloud-init and say "can you render Fedora/sysconfig" | 17:43 |
rharper | smoser: agreed; certainly for now, a flag or two makes sense | 17:43 |
smoser | i can see a reason to say "if i gave you v2, can you render that for *THIS SYSTEM*" | 17:43 |
rharper | we may want something more complex later | 17:43 |
rharper | smoser: from curtin perspective, it will know what it has (v1 or v2) and it will want to know if cloud-init can render it , and it *will* know what the target os is | 17:44 |
rharper | so, I do think it's cloud-init, can you render v2 -> fedora/sysconfig | 17:44 |
rharper | because, they may want to see if we can v2 -> fedora/networkd | 17:45 |
rharper | I do think it's os + networkcfg format | 17:45 |
smoser | powersj, what am i doing wrong | 17:45 |
smoser | http://paste.ubuntu.com/24152788/ | 17:45 |
powersj | smoser: what version of pylxd do you have? If I recall 2.1.3 is still the one that works best until we get magicalChicken's branch added | 17:47 |
smoser | 2.2.3-0ubuntu1 | 17:47 |
magicalChicken | smoser: you need to run from the current branch then | 17:48 |
magicalChicken | smoser: the old cloud_tests can't handle 2.2 | 17:48 |
magicalChicken | wait, if 2.2.3 is out that means that the fix for the leaked file handle bug is released, so I can go ahead and update the tox env in the new version | 17:49 |
magicalChicken | that means we should be able to run 2.2.3 on jenkins | 17:49 |
powersj | yep looks like that just got released! | 17:49 |
magicalChicken | nice | 17:49 |
powersj | getting that + the tox commands would make smoser's life easier too | 17:51 |
smoser | yeah, do you have a example tox env ? | 17:51 |
magicalChicken | smoser: 1 sec | 17:52 |
magicalChicken | http://paste.ubuntu.com/24152810/ | 17:52 |
magicalChicken | smoser: ^ | 17:52 |
magicalChicken | powersj: i am going to be able to work almost full time next week, so i may go ahead and add support for distro and platform feature flags and write some of the kvm platform | 17:53 |
magicalChicken | powersj: i have been talking to jgrimm about getting centos tests running smoothly so we can start integration testing there too | 17:54 |
powersj | magicalChicken: no beach vacation for spring break? ;) | 17:54 |
magicalChicken | powersj: haha no, probably not :) | 17:54 |
powersj | magicalChicken: sounds great -- let me know what you start on. I can jump between bugs and the kvm stuff | 17:55 |
magicalChicken | powersj: sure, sounds good | 17:55 |
magicalChicken | i'll get all thise done in a new branch so that mp doesn't get any bigger | 17:55 |
jgrimm | thanks magicalChicken | 17:55 |
magicalChicken | jgrimm: no problem :) | 17:56 |
smoser | powersj, was that the test that failed ? tests/cloud_tests/configs/modules/set_password.yaml | 17:56 |
powersj | smoser: yes | 17:56 |
smoser | ok. | 17:56 |
smoser | magicalChicken, can you just put a MP for the citest toxenv right now ? | 17:58 |
smoser | that is very useful on its own. | 17:58 |
magicalChicken | smoser: sure i'll pull that out | 17:59 |
magicalChicken | smoser: if it would make it easier to review the main one, i can split it up into a few smaller ones. they'll all still be needed to get everything working but it may make the diff easier to read | 18:00 |
powersj | magicalChicken: if there are some things that we could divide out it probably would be good to do anyway | 18:02 |
magicalChicken | powersj: part of the problem is that if i divide things out, the test suite as a whole won't work until almost all of it is merged in | 18:03 |
smoser | magicalChicken, patch series are easier, yeah. | 18:03 |
smoser | you can just work it as a patch series... | 18:04 |
smoser | in a single mp, and rebase ... commit --fixup.... like that. | 18:04 |
magicalChicken | smoser: the commit history is already pretty heavily modified, there aren't really any junk commits left | 18:04 |
magicalChicken | i can do it a bit further though | 18:05 |
smoser | ok. | 18:05 |
smoser | if you want, i can just do the tox.ini change for trunk right now. | 18:05 |
smoser | powersj, that wont cause harm to antying in c-i right ? | 18:06 |
smoser | if i aadded 'citest' tox | 18:06 |
magicalChicken | smoser: that should be fine | 18:06 |
powersj | smoser: correct that is fine | 18:06 |
* powersj greatly wants that change too | 18:06 | |
magicalChicken | smoser: it would have to just be the pylxd 2.1.3 env for now though ,the 2.2.3 env would break | 18:06 |
smoser | ok | 18:07 |
smoser | magicalChicken, powersj | 18:11 |
smoser | http://paste.ubuntu.com/24152915/ | 18:11 |
smoser | that look sane ? i'll push that now. | 18:12 |
smoser | w | 18:12 |
magicalChicken | smoser: yeah, that looks good, thanks | 18:12 |
powersj | +1 | 18:12 |
smoser | pushed. | 18:12 |
magicalChicken | at some point we should probably update the docs again, but there's enough stuff changing for now that it can probably wait | 18:13 |
smoser | and powersj trunk should hopefully pass again | 18:15 |
powersj | smoser: thank you | 18:16 |
smoser | rharper, you want to ack https://code.launchpad.net/~smoser/cloud-init/+git/cloud-init/+merge/319605 ? and i'll pull it | 18:16 |
jgrimm | magicalChicken, fwiw i generally like to see doc changes accompany said code/test changes. | 18:19 |
jgrimm | but please open a bug as reminder if known gaps | 18:19 |
magicalChicken | jgrimm: sure, i'll open a bug to track integration-testing doc updates | 18:20 |
jgrimm | thanks | 18:20 |
magicalChicken | i'll see if i can get to some of those changes next week | 18:20 |
jgrimm | +1 thanks | 18:20 |
rharper | smoser: done | 18:39 |
smoser | powersj, i kicked https://jenkins.ubuntu.com/server/job/cloud-init-integration-z/ and it ran successful | 18:48 |
powersj | smoser: good, I had gone into torkoal after I saw the failure and blew away the offending lxc image | 18:48 |
powersj | that means tot is fixed as well :) \o/ | 18:49 |
smoser | ah. i didnt realize there had been lxc issues | 18:49 |
powersj | "Failed to create ZFS filesystem: cannot mount '/var/lib/lxd/images/(uuid of image)" | 18:49 |
powersj | this is the issue cpaelzer and I keep hitting on the test systems | 18:50 |
powersj | I am trying to reproduce locally, but waiting for new daily images as I think that is when it happens | 18:51 |
cpaelzer | ? | 19:33 |
cpaelzer | powersj: so we hit it again? | 19:33 |
powersj | cpaelzer: on torkoal yeah :( | 19:33 |
cpaelzer | without touching LXD versions | 19:33 |
powersj | correct | 19:33 |
cpaelzer | just as I promised | 19:33 |
cpaelzer | are you giving stgraber access again? | 19:34 |
cpaelzer | probably the best debug chance | 19:34 |
powersj | well, I already blew the failed image away | 19:34 |
cpaelzer | did that make it working? | 19:34 |
powersj | yes | 19:34 |
cpaelzer | e.g. after sync of the same image it is ok | 19:34 |
cpaelzer | hmm | 19:34 |
cpaelzer | interesting | 19:34 |
cpaelzer | but that is a great indicator where stgraber can check next time it happens | 19:35 |
powersj | yeah | 19:35 |
cpaelzer | powersj: did yu update the github issue I filed` | 19:35 |
cpaelzer | ? | 19:35 |
powersj | no, but I can | 19:35 |
cpaelzer | thanks | 19:35 |
cpaelzer | uh, I just got suggested that I#M in EOD (EOW) for a while ... | 19:35 |
* cpaelzer hushing back to family | 19:36 | |
powersj | :D | 19:36 |
powersj | magicalChicken: for integration tests, the only system requirements right now are python3, lxd, and pylxd==2.1.3? | 19:40 |
magicalChicken | powersj: as far as i know yeah | 19:41 |
powersj | Updating the docs with how to run via tox + putting small section on dependencies | 19:42 |
magicalChicken | powersj: that's all that is set up in the tox env and that works | 19:42 |
magicalChicken | powersj: oh, nice | 19:42 |
powersj | magicalChicken: https://paste.ubuntu.com/24153367/ | 19:42 |
magicalChicken | i'll pull those changes in with the pylxd 2.2 stuff and update them for that as well | 19:42 |
magicalChicken | powersj: yeah, that looks good | 19:43 |
powersj | ok | 19:43 |
powersj | thx! | 19:43 |
magicalChicken | powersj: it may be worth changing the full path to the test config to just moduules/user_groups | 19:43 |
powersj | I don't need the "tests/cloud_tests/configs" part? | 19:44 |
magicalChicken | it doesn't really make a difference, but i find it easier to read that way | 19:44 |
magicalChicken | powersj: no, the argument normalizer will handle that for you | 19:44 |
powersj | O.o didn't know this | 19:44 |
magicalChicken | i built in relative path -> absolute path stuff pretty much everywhere because i'm lazy typing out commands :) | 19:45 |
powersj | let me go try this then | 19:45 |
smoser | rharper, http://paste.ubuntu.com/24153396/ | 19:48 |
rharper | ok | 19:50 |
rharper | smoser: re the virtual stuff; that's just super annoying | 19:50 |
powersj | magicalChicken: very nice | 19:51 |
rharper | I don't think rename in lxc is an issue, no? it's created with the name it needs | 19:51 |
rharper | smoser: so not sure why that matters | 19:51 |
magicalChicken | powersj: :) the code that handles args in the new version is kinda cool | 19:53 |
smoser | well, definitely one way to interact with lxc is to request it to name your network devices. | 19:53 |
smoser | but another is to tell cloud-init via network config. | 19:54 |
rharper | smoser: I updated my comment in the bug; we need to somehow detect duplicate macs and sort out the *real* interface that should be renamed | 19:55 |
smoser | yeah. | 19:56 |
rharper | smoser: that's for the review, replying to it is somewhat awkward | 19:56 |
smoser | agree | 19:56 |
smoser | why did i do it like that.. | 19:57 |
smoser | because it wasnt submitted for merge | 19:57 |
smoser | right ? | 19:57 |
smoser | https://code.launchpad.net/%7Ecurtin-dev/curtin/trunk/+activereviews | 19:57 |
smoser | bah. well it woudlt be there | 19:58 |
smoser | for some reason i thought you hadn't clicked submit | 19:58 |
smoser | i can replay my comments onto the mp | 19:58 |
smoser | https://code.launchpad.net/~raharper/cloud-init/+git/cloud-init/+merge/319259 | 19:58 |
smoser | rharper, updated. https://code.launchpad.net/~raharper/cloud-init/+git/cloud-init/+merge/319259 | 20:07 |
rharper | smoser: I sent a mail reply | 20:14 |
dtp | hello. cloud-init-0.7.8-5.fc25 is failing to write net config for me on Fedora 25. error is [ValueError: Unknown subnet type 'loopback' found for interface 'lo'] at net/sysconfig.py:238 | 20:22 |
harlowja | larsks have u seen anything like that ( dtp was poking me - also works at godaddy, but not sure if u've already fixed that, lol ) | 20:23 |
larsks | harlowja: I haven't seen that previously, but 'loopback' is clearly not one of the supported types in NetInterface.iface_types... | 20:24 |
harlowja | ya, lol | 20:24 |
larsks | ...but why are they passing in configuration for lo? | 20:24 |
larsks | That sounds goofy. | 20:24 |
harlowja | agreed | 20:25 |
dtp | my content/0000 looks like ubuntu w/ an [iface lo inet loopback] line | 20:25 |
dtp | http://imgur.com/a/qEElh | 20:26 |
rharper | isn't the content/XXXX path where a complete eni is being injected rather than using network_data.json ? | 20:26 |
dtp | added network_data.json image to that link, too | 20:27 |
dtp | i don't see a loopback in network_data.json; so must be coming from content/0000? | 20:28 |
rharper | I need to review the code path, but I thought if we got a content eni file, we'd just pass that through; but maybe we're attempting to inject the eni into network_State and then re-render something else (like sysconfig) | 20:28 |
rharper | which sounds like what's happening | 20:29 |
* harlowja didn't think we did that (but maybe we do, ha) | 20:29 | |
rharper | well | 20:29 |
dtp | added most of the stack trace to that link | 20:29 |
rharper | but if openstack sends a network_data.json, I'd think we prefer that to injesting eni into network_state | 20:29 |
larsks | rharper: and yet, the network_data.json doesn't container any reference to lo... | 20:30 |
rharper | it shouldn't | 20:30 |
larsks | well, of course not. | 20:30 |
larsks | So the exception dtp is seeing suggests that we *don't* prefer network_data.json. | 20:31 |
rharper | yeah | 20:31 |
rharper | I wonder why they send both | 20:31 |
larsks | rharper: huh, looking at the code, it seems as if network_json should take precedence: https://github.com/cloud-init/cloud-init/blob/master/cloudinit/sources/DataSourceConfigDrive.py#L142 | 20:36 |
larsks | dtp: do you see any of those message in your logs (on lines 145 or 150)? | 20:36 |
dtp | not that i recall larsks, but i will dbl check now | 20:37 |
dtp | not that i can see | 20:38 |
larsks | Hmm. I need to run off, but if you haven't already, open a bug so that we don't forget (including details of cloudinit version, distro and version, cloud provider, etc) | 20:39 |
dtp | ok | 20:39 |
dtp | https://bugs.launchpad.net/cloud-init/+bug/1671927 | 20:51 |
rharper | larsks: not quite; the code I see checks if the we have a self._network_config; if it's None, then we check if we have a network_json and convert; so the eni file mapped in the metadata content will get used first | 20:54 |
=== dtp is now known as dtp-afk | ||
larsks | rharper: ah, good catch. | 20:55 |
rharper | larsks: but I think we're meant to | 20:56 |
rharper | given the network_eni | 20:56 |
rharper | in distro/rhel.py we do this; so I suspect we're getting the eni passed in here; entries = net_util.translate_network(settings) | 20:59 |
rharper | something funny is going on w.r.t the type of network config that was found and passed in; more investigation needed; | 21:17 |
rharper | dtp-afk: if you can, attaching the configdrive to the bug would be super helpful | 21:18 |
smoser | rharper, sorry for the noise above with the patch comments... can you re-play on the MP ? | 21:20 |
rharper | yes | 21:20 |
smoser | i'ms orry i just didnt' think you'd proposed for merge | 21:20 |
smoser | :-( | 21:20 |
rharper | no worries | 21:20 |
* rharper forklifts into web form | 21:21 | |
* rharper mumbles something about gerrit | 21:21 | |
=== dtp-afk is now known as dtp | ||
dtp | rharper you want the directory listing or ? | 22:37 |
=== rangerpb is now known as rangerpbzzzz | ||
dtp | uploaded 0000 and meta_data.json | 22:52 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!