/srv/irclogs.ubuntu.com/2020/09/22/#cloud-init.txt

ebal[m]Hello07:52
* ebal[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/JnMMsgLMbxHfZyrsRfQMorxp/message.txt >07:55
meenaebal[m]: what's the issue on kernel > 5.7?08:52
meenai'm trying to remember why we made those exceptions for fallocate on XFS rather than just always using dd to begin with08:55
* ebal[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/ShEKrtiBPTAtwGBBWmfiRklG/message.txt >09:10
meenawhy not truncate, i wonder…09:14
meenaebal[m]: that's two minutes?!09:15
ebal[m]test on wsl  ;)  (work laptop)09:18
Odd_Blokeotubo: Ack, thanks!13:06
Odd_Bloke[42]: Are you using netplan to configure your Debian system, or you were just referring to the configuration format?13:09
Odd_BlokeI would expect cloud-init to accept any valid netplan configuration; if you aren't seeing that then a bug would be much appreciated!13:10
Odd_Blokeebal[m]: It looks to me from the link to linux-fsdevel on that Bugzilla ticket that you're hitting a kernel bug, not a cloud-init bug; have you raised the issue with your distro's kernel maintainers?13:16
rharperebal[m]: meena: re: fallocate on ext4 on 5.8 kernels, https://bugs.launchpad.net/ubuntu/+source/linux/+bug/189491014:23
ubot5Ubuntu bug 1894910 in linux (Ubuntu) "fallocate swapfile has holes on 5.8 ext4, causes: swapon failed: Invalid argument" [Undecided,Confirmed]14:23
ebal[m]There is already a couple (or a few) bug reports for this issue (linux/fallocate) and for archlinux also.14:26
ebal[m]Nevertheless, I wanted to openly discuss it here and exchange ideas or even a workaround.14:26
ebal[m]personally I am blocked - but it is not important14:27
rharperdd is the workaround; since this is a regression in the 5.8 kernel, I wouldn't expect any changes to cloud-init here14:27
ebal[m]it is important to see if there is of any value to suggest that cloud-init can switch to dd as a primary choice instead of fallocate or even provide a flag in cloud-init configuration file so you can choose how the swap partition will be created14:29
ebal[m]cause in the end, it is a choice to use fallocate against dd ( I guess ).14:29
rharperI don't see cloud-init switching;  it's a regression in the filesystem in the kernel;    as a workaround for right now, one could use boothook to run a command to move fallocate out of $PATH  (mv /usr/bin/fallocate /root);  when cc_mount runs create_swap with fallocate, this will fail, and cloud-init falls back on dd;14:34
rharperebal[m]: I suppose a PR to allow swap: config to take a command would give the user some control over what to use;14:35
* ebal[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/UpoXgazQtXpvigmKMIJDmRWS/message.txt >14:39
* ebal[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/VRvIjGKmKAISzTkTHdxRcyJj/message.txt >14:40
[42]Odd_Bloke: yes also using netplan to configure it14:43
Odd_Bloke[42]: And does the configuration work if you manually apply it?14:51
[42]yes14:53
[42]this is an example for a machine without public ipv4 https://paste.debian.net/plainh/a1eb40cd14:55
Odd_Bloke[42]: Are you able to pastebin cloud-init.log from an affected instance?15:02
[42]i can launch another run in a bit, will need to make sure to set a password to be able to login via console :)15:03
Odd_Bloke:)15:03
Odd_Blokerobjo: Are the Python versions that you care about for SUSE still 2.7/3.4 for 12.4 and 3.6 for 15?  (Is cloud-init still(?) being backported to 12.4?)15:08
rharper[42]: thanks for posting, I don't believe your target os supports netplan, I can reproduce the error with your config when rendering from netplan input to eni;15:08
rharperhttps://paste.ubuntu.com/p/wZYpgq498f/15:08
rharper[42]: do you have what you expect /etc/network/interfaces should look like?   or did you expect netplan to work in which case, netplan detection on debian may need fixing15:09
[42]i did expect netplan to work15:09
[42]i'm already using it with netplan manually on some machines15:10
[42]as in non-cloud-init setups15:10
robjoOdd_Bloke: yes, I am in the process of switching the build for SLES 12 to Python 3.415:11
[42]the error in your paste looks like it's exactly what i saw in the console when i tried15:11
[42]netplan rendering from the config above results in https://paste.debian.net/plainh/ed1efb5215:14
rharper[42]: ok, the boot logs will be helpful:  it sounds like you have both ifupdown and netplan installed, cloud-inits default policy for net renderers will try eni first, if found it'll render that;15:14
[42]that could be the case, i'm using a modified debian cloud-init image which i added netplan.io to15:15
rharperif you;'re creating your own image, then you can modify /etc/cloud/cloud.cfg and set network: renderers: ['netplan']15:16
rharperunder system_info:15:17
[42]ok, i'll have a look at that after work, thanks15:17
meenarobjo, Odd_Bloke: does cloud-init even work with 3.4??15:51
powersjduring last years cloud-init summit, as we deprecated python 2 support we agreed to start python 3 support with 3.415:54
robjoI have a few more package dependencies to resolve but do not expect any issues on 3.415:55
Odd_Blokemeena: Yep, 3.4 is our lowest supported version.16:19
Odd_Bloke(Else we'd have much better type annotations. :p)16:19
meenai was wondering…17:19
meenayah, in libioc we went with 3.6 as the lowest bar17:19
meenai think… mostly… cuz that's what the lowest bar in FreeBSD was17:20
=== ijohnson is now known as ijohnson|lunch
=== ijohnson|lunch is now known as ijohnson
meenasomebody at redhat created a job for me: https://global-redhat.icims.com/jobs/80986/senior-deployment-engineer---openstack/job19:21

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!