| *** mjturek has quit IRC | 00:07 | |
| prometheanfire | ianw: I can reboot after growroot is run just fine (it seems) | 00:11 |
|---|---|---|
| ianw | prometheanfire: ok ... was that a problem? | 00:12 |
| prometheanfire | I was led to believe it was | 00:13 |
| clarkb | prometheanfire: no... | 00:13 |
| clarkb | the images are booting fine I siad | 00:13 |
| clarkb | the problem is if you then try to do anything useful in the image that requires more than a little bit of disk | 00:13 |
| prometheanfire | ah, so if I dd to a file? | 00:14 |
| prometheanfire | then reboot | 00:14 |
| clarkb | rebooting is orthogonal | 00:14 |
| prometheanfire | sure, just a useful test | 00:14 |
| clarkb | I was just warning you that you might end up with a non grow rooted root | 00:14 |
| clarkb | booting will work fine | 00:14 |
| prometheanfire | hmm, growroot seemed to work | 00:15 |
| clarkb | huh gentoo growroot is less discerning I guess? | 00:16 |
| prometheanfire | it's the same growroot | 00:17 |
| ianw | what does fdisk /dev/vda1 show? | 00:18 |
| prometheanfire | Device Boot Start End Sectors Size Id Type | 00:19 |
| prometheanfire | /dev/vda1 * 2048 41943006 41940959 20G 53 OnTrack DM6 Aux3 | 00:19 |
| prometheanfire | ya, that's not quite right | 00:19 |
| clarkb | fix is on its way, prereq for fix is finally in the gate | 00:20 |
| prometheanfire | yarp | 00:20 |
| ianw | that's the problem ... possibly gentoo's sfdisk patches out the check that considers "OnTrack DM6 Aux3" a "disk manager" and dies | 00:21 |
| prometheanfire | anyway, systemd is having issues with the console | 00:21 |
| prometheanfire | maybe | 00:21 |
| prometheanfire | [[0;1;31mFAILED[0m] Failed to start Glean for interface ens3. | 00:25 |
| prometheanfire | See 'systemctl status glean@ens3.service' for details. | 00:25 |
| prometheanfire | booo | 00:25 |
| openstackgerrit | Merged openstack/diskimage-builder master: Remove tripleo jobs https://review.openstack.org/555037 | 00:31 |
| clarkb | ianw ^ I guess approve fix now? | 00:33 |
| ianw | clarkb: ++ | 00:35 |
| ianw | i'm just working on a test for it in the nodepool boot tests | 00:35 |
| clarkb | do you want that first? | 00:36 |
| ianw | clarkb: no, just for ongoing sanity | 00:41 |
| clarkb | ok I've approved the fix | 00:45 |
| ianw | hmm, it's harder to come up with a good test for this than i first thought | 00:59 |
| clarkb | ianw: maybe just reexecute growroot and check the return code? | 01:00 |
| ianw | we have two sizes of disk, 512 & 1024 which are used fairly randomly. i *think* just checking the root is exapnded to 512mib covers it | 01:00 |
| ianw | yeah, i'd sort of like to verify it *actually* worked and the disk is bigger, if we can | 01:01 |
| ianw | see what happens with https://review.openstack.org/555103 | 01:02 |
| prometheanfire | ianw: you think devuser should install sudo? | 01:29 |
| openstackgerrit | Merged openstack/diskimage-builder master: Fix default partition type https://review.openstack.org/554771 | 01:30 |
| prometheanfire | clarkb: with that merge are we good? | 01:33 |
| clarkb | prometheanfire: need a release too | 01:33 |
| prometheanfire | says you :P | 01:34 |
| ianw | prometheanfire: there's a DIB_DEV_USER_PWDLESS_SUDO ... i've been stuck before not providing that and having a devuser with only key login. | 01:40 |
| ianw | it probably should bring in sudo ... i guess on every other platform that's already part of base? | 01:40 |
| ianw | prior coment is to say, i've never been 100% on the semantics of devuser/sudo etc | 01:41 |
| prometheanfire | it was part of it by default on gentoo, but a user made the point that a minimal image shouldn't include it | 01:43 |
| prometheanfire | already tagged? | 01:48 |
| prometheanfire | got a new task to package 2.12.1 | 01:48 |
| prometheanfire | can confirm good on that last patch | 02:48 |
| prometheanfire | /dev/vda1 * 2048 41943006 41940959 20G 83 Linux | 02:48 |
| *** andreas_s has joined #openstack-dib | 02:48 | |
| *** andreas_s has quit IRC | 02:53 | |
| greghaynes | prometheanfire: +1 on devuser should install sudo | 03:16 |
| prometheanfire | greghaynes: you happen to know simple-init? | 03:19 |
| greghaynes | uhhh, not sure im willing to admit to it | 03:19 |
| prometheanfire | it's a simple question :P | 03:20 |
| prometheanfire | what do we expect `ifup INTERFACE` to do in https://github.com/prometheanfire/glean/blob/master/glean/init/glean%40.service#L13 run dhcp? | 03:20 |
| greghaynes | prometheanfire: I think the idea is for each interface we first call glean to potentially write out a networking config for that interface, then ifup to trigger the actual bring up the interface, run whatever we configured (dhcp client, etc) | 03:23 |
| greghaynes | so its just the bit to poke distro network bringup for the interface glean just wrote config for | 03:24 |
| prometheanfire | greghaynes: I'd like to convert that to networkd | 03:24 |
| greghaynes | in general or add support for networkd? | 03:25 |
| greghaynes | the issue is networkd doesnt work cleanly atm on many distros | 03:25 |
| prometheanfire | if gentoo is using systemd for it's init instead of openrc we are using networkd (or the user selects something else like networkmanager) | 03:27 |
| greghaynes | yea I think itd be great to support networkd where its standardish | 03:27 |
| greghaynes | its just on fedora and deb theres competing systems which are the default | 03:28 |
| greghaynes | so for supporting it - does networkd not have a bring interface X up? | 03:29 |
| prometheanfire | ya, have to research | 03:29 |
| prometheanfire | will ask cloudnull :P | 03:29 |
| openstackgerrit | Merged openstack/diskimage-builder master: remove portage git directory https://review.openstack.org/553914 | 03:32 |
| cloudnull | prometheanfire: in osa we're actively making networkd the option we use in test https://review.openstack.org/#/c/552721 | 03:42 |
| prometheanfire | cloudnull: ya, need to figure out how do work between systemd and glean/simple-init | 03:45 |
| prometheanfire | cloudnull: what I need to do is talk to whoever is the lead for glean... | 03:50 |
| cloudnull | i dont know anything about glean. but i can help with the networkd bits if you need | 03:50 |
| * prometheanfire doesn't think anyone is leading glean | 03:52 | |
| ianw | it's the type of thing that just works ... until it doesn't | 04:13 |
| *** chhagarw has joined #openstack-dib | 04:13 | |
| clarkb | what I would do is have a unit that reads the config drive and writes out all the interface configs and set that up before networkd | 04:15 |
| clarkb | Then in the simple init element have it select for that over the distro specific network config (this part is hand wavy but in thoery networkd would work on all the systemd distros) | 04:16 |
| prometheanfire | clarkb: for networkd it can be one file iirc | 04:17 |
| prometheanfire | clarkb: I'll talk to cloudnull tomorow about it (probably), afk for now... | 04:18 |
| clarkb | prometheanfire: the docs I'm reading seem to suggest a .network file per interface but that may just be for human readability | 04:18 |
| clarkb | but ya single glean unit. write all the .network file(s) content set that before networkd done | 04:19 |
| cloudnull | clarkb: prometheanfire: we have an ansible role which will configure the interfaces https://github.com/openstack/ansible-role-systemd_networkd | 04:33 |
| cloudnull | idk if you can do ansible on localhost? however if you can that might make things easy enough | 04:33 |
| ianw | cloudnull: the point of glean is to be a minimal thing that brings up the network, so ansible kind of breaks that model :) | 04:35 |
| cloudnull | fair enough. | 04:36 |
| cloudnull | generating the netdev and network files for networkd is fairly small should be simple enough to template in just about anything | 04:39 |
| prometheanfire | cloudnull: ya, I just didn't want to be the person to jump on this grenade :P | 05:00 |
| prometheanfire | cloudnull: I'll get osa on gentoo, one day... | 05:00 |
| openstackgerrit | Matthew Thode proposed openstack/diskimage-builder master: install sudo in the devuser element https://review.openstack.org/555149 | 05:18 |
| prometheanfire | greghaynes: ^ | 05:18 |
| prometheanfire | greghaynes: feel free to comment or add other distro package mappings for sudo | 05:18 |
| *** andreas_s has joined #openstack-dib | 07:21 | |
| *** andreas_s has quit IRC | 07:25 | |
| *** rajinir has quit IRC | 07:25 | |
| *** rajinir_ has joined #openstack-dib | 07:26 | |
| *** rajinir_ has quit IRC | 07:26 | |
| *** rajinir_ has joined #openstack-dib | 07:26 | |
| *** rajinir_ is now known as rajinir | 07:26 | |
| *** andreas_s has joined #openstack-dib | 07:27 | |
| *** yolanda_ is now known as yolanda | 08:35 | |
| *** jesusaur has quit IRC | 09:20 | |
| *** jesusaur has joined #openstack-dib | 09:29 | |
| *** hashar has joined #openstack-dib | 09:58 | |
| *** jesusaur has quit IRC | 11:02 | |
| *** jesusaur has joined #openstack-dib | 11:12 | |
| *** dhill_ has quit IRC | 13:35 | |
| *** pickle has joined #openstack-dib | 13:37 | |
| *** mjturek has joined #openstack-dib | 14:00 | |
| *** hashar is now known as hasharAway | 15:13 | |
| *** vmlinuz has joined #openstack-dib | 15:27 | |
| *** andreas_s has quit IRC | 16:19 | |
| *** andreas_s has joined #openstack-dib | 16:33 | |
| prometheanfire | ianw: happen to know why umount of the build_dir is not working (it seems to still have things like /tmp/dib_build.kG6NA6mS/mnt/proc mounted) | 16:46 |
| *** andreas_s has quit IRC | 16:47 | |
| *** mjturek has quit IRC | 16:58 | |
| *** mjturek has joined #openstack-dib | 16:58 | |
| *** mjturek has quit IRC | 17:01 | |
| *** mjturek has joined #openstack-dib | 17:01 | |
| openstackgerrit | Doug Hellmann proposed openstack/diskimage-builder master: add lower-constraints job https://review.openstack.org/555412 | 18:27 |
| prometheanfire | ianw: was finally able to reproduce the [[ foo ]] && bar failure | 18:37 |
| prometheanfire | not sure why it's happening still, but meh | 18:37 |
| prometheanfire | I'll change that pr to a full if | 18:37 |
| * prometheanfire wonders if package-isntalls can handle installopts | 18:42 | |
| openstackgerrit | Matthew Thode proposed openstack/diskimage-builder master: enable systemd profile for Gentoo https://review.openstack.org/552952 | 18:58 |
| ianw | prometheanfire: no ... i'm interested in any patches for unmount issues that don't involve just making it lazy ... it's been a persistent problem but hard to track down | 20:12 |
| openstackgerrit | Matthew Thode proposed openstack/diskimage-builder master: enable systemd profile for Gentoo https://review.openstack.org/552952 | 20:13 |
| prometheanfire | ianw: ya, I saw the review wanting to add -l | 20:13 |
| prometheanfire | and I aggree that is the wrong solution | 20:13 |
| ianw | prometheanfire: what did you do differently to trigger it locally? i'm interested if it's something we can dib-lint or bashate | 20:13 |
| prometheanfire | ianw: can't dib just build is a process namespace, that way you can kill everything in that process namespace, so nothing should be held open | 20:14 |
| prometheanfire | I didn't do anything | 20:14 |
| prometheanfire | it's confusing, I had people at work run the code happily too | 20:14 |
| ianw | some sort of containierisation is probably a good way forward | 20:15 |
| clarkb | process wouldn't necesasrily die right away in that case would it? | 20:16 |
| prometheanfire | kill -9 ppid, then recheck for processes in namespace | 20:16 |
| prometheanfire | do that a couple times for good measure | 20:17 |
| *** vmlinuz has quit IRC | 21:08 | |
| *** chhagarw has quit IRC | 21:12 | |
| openstackgerrit | greghaynes proposed openstack/diskimage-builder master: [WIP] new diskimage-build command with config https://review.openstack.org/554002 | 21:26 |
| openstackgerrit | greghaynes proposed openstack/diskimage-builder master: Move fail early checks in to python https://review.openstack.org/555462 | 21:26 |
| openstackgerrit | greghaynes proposed openstack/diskimage-builder master: [WIP] Move environment create/cleanup to python https://review.openstack.org/555463 | 21:26 |
| openstackgerrit | greghaynes proposed openstack/diskimage-builder master: [WIP] Move environment create/cleanup to python https://review.openstack.org/555463 | 21:28 |
| greghaynes | ^ might be a nice way to get to cleaner process management / using namespaces | 21:29 |
| openstackgerrit | greghaynes proposed openstack/diskimage-builder master: Move fail early checks in to python https://review.openstack.org/555462 | 21:45 |
| ianw | greghaynes: one thing that always worries me a little about turning python into a shell runner is you basically lose tracing, which is so important for figuring things out | 22:17 |
| ianw | when what you're doing comes down to running a series of shell commands in a certain order, unsurprisingly shell does pretty good at that :) | 22:17 |
| ianw | what it doesn't do well at is anything involving data structures | 22:18 |
| greghaynes | ianw: you should have a look at the bits im removing - they include things like writing python blocks to be executed via pyton -c ') | 22:19 |
| ianw | yeah, oh i agree on moving forwards with something else | 22:19 |
| greghaynes | the other thing I really hope is this makes logging a lot nicer | 22:20 |
| ianw | just by the time you add in tracing, logging, error handling, and some configuration ... you're starting to look a lot like ansible :) | 22:20 |
| greghaynes | I think you dont really want 'tracing' here, you just want better logging (which probably includes logs of every time you shell out) | 22:21 |
| greghaynes | but I dunno, I kind of started it as an experiment to look at 'hey is this significantly easier to write / read than the corresponding bash' | 22:21 |
| ianw | right, that's fundamentally what i mean by tracing | 22:22 |
| greghaynes | oh thats super easy though - just make a wrapper for os.popen or whatnot | 22:22 |
| ianw | even as we've shown, reliably getting output from the console can be non-trivial | 22:22 |
| ianw | yeah, we already have one layer of it in dib-block-device | 22:23 |
| greghaynes | I really want to see where it breaks down in to being more of a pain than bash, but so far ISTM its pretty far on the side of this is way easier to follow | 22:24 |
| ianw | cool, not trying to be overly negative :) | 22:24 |
| greghaynes | yea no worries - I totally get the fear | 22:24 |
| ianw | in dib's defence, i guess that when it started, i doubt anyone was really thinking about containers | 22:25 |
| ianw | well, chroot was the container at the time | 22:25 |
| greghaynes | yea, still is kinda, it just also has gotten a lot more complex at which point having nicer interfaces between bits of code becomes more valuable | 22:26 |
| *** andreas_s has joined #openstack-dib | 22:47 | |
| *** andreas_s has quit IRC | 22:52 | |
| *** calebb has quit IRC | 23:50 | |
| *** calebb has joined #openstack-dib | 23:52 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!