=== paride1 is now known as paride [11:16] join #cloudinit [11:16] sorry [13:14] this just popped up on showhn: https://news.ycombinator.com/item?id=23310063 [13:34] hi folks, is there a way to ask cloud-init what datasource was used? I see that with`cloud-init status --long` there is a "details" section with this information, but is that the right way to do it and is that in a format that can be depended on ? [14:06] ijohnson: For programmatic use, I wouldn't depend on it. It reads from /run/cloud-init/result.json, so that would be a better place to look. [14:06] Odd_Bloke: so is /run/cloud-init/result.json a stable format that I can rely on for just reading the datasource ? [14:08] ijohnson: Yep, there's a v1 key which we'll keep stable. [14:08] ijohnson: I'd be interested to know a little more about your use case, too, for future reference. [14:11] is this known? [14:11] http://paste.ubuntu.com/p/mDk8TNvgSc/ [14:12] launched a fresh azure instance of 18.04. the clock jumps during boot. [14:12] 2020-05-17 17:02:58,229 - netlink.py[DEBUG]: Wait for media disconnect and reconnect to happen [14:12] 2020-05-26 13:52:56,326 - netlink.py[DEBUG]: netlink socket ready for read [14:13] Wow. [14:13] that seems a little "broken" to me. Why wouldn't you set your vm bios to the correct time to start? [14:14] but ... ok, ntp does the rigth thing and fixes it. [14:14] but what seems *more* broken is that uptime seems unaware of the jump [14:14] uptime [14:14] 14:13:00 up 8 days, 21:10, 3 users, load average: 0.00, 0.11, 0.15 [14:14] i guess its possible that this was a "pre-provisioning" [14:15] but i would not have expected that... I suspect on the order of 5 ubuntu vms have been launched on thsi account in a year [14:17] smoser: `journalctl -o short-monotonic` might give you a separate indication of how long the machine has actually been up? [14:17] yeah. just ran that [14:17] it shows the hump [14:17] pasting [14:17] http://paste.ubuntu.com/p/66SRvpJY3k/ [14:18] so the jump occurs right around the dcp. so i'd certainly think its ntp related. [14:23] smoser: Does `journalctl -u chrony.service` shed any light? [14:24] -- Logs begin at Sun 2020-05-17 17:02:46 UTC, end at Tue 2020-05-26 14:23:17 UTC. -- [14:24] -- No entries -- [14:24] so.. no [14:30] smoser: Oh, I was looking at GCE where I think they requested chrony, so perhaps: `journalctl -u systemd-timesyncd.service`? [14:32] / [14:32] https://paste.ubuntu.com/p/fdMpW8BXNy/ [14:33] am i just wrong? is uptime supposed to jump like that? [14:34] I'm not really sure TBH. [14:35] The fact the monotonic timestamp also jumps has me wondering if it's actually just the NTP clock setting that's going on. [14:39] https://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/1880704 [14:39] Ubuntu bug 1880704 in cloud-init (Ubuntu) "jump in kernel clock during boot on azure" [Undecided,New] [14:54] AnhVoMSFT: Would you be able to take a look at ^ and let us know what you think, please? [14:58] smoser: some time dameons do an initial time warp if the delta is above some threshhold [14:58] not sure what timesyncd does [14:58] is the final time after the sync the real time that makes sense? [14:59] cpaelzer: right. that didn't surprise me. [14:59] but i dont think the kernel clock is supposed to jump like that [14:59] is'nt that what "monotonic" means ? [15:00] * smoser realizes monotonic might *only* mean not-ever-decreasing [15:01] well, per metacpan "A clock source that only increments and never jumps" [15:01] which is what I thought too [15:01] and since a perl documentation site agrees with me, i'm clearly right. or should i just be embarrased. [15:03] exactly - monotonic is split in two way [15:03] never decrease [15:05] Hi All , we are trying to use cloud-init to initialize our images , but mostly our images have multiple disks (due to xenserver server limitation on max single disk size) and we need to have lvm that span multiple disks to show as 1 disk for end users, I see that could-init plan t support mdadm raid in the future , so my question do you know if [15:05] there is any ETA for that ? and if yes do think the user case of having raid1 just to have all disks shown as 1 will be supported ? [15:11] cpaelzer: did you miss "never jumps" above? [15:12] Hani: hey, advanced storage config is definitely on the roadmap, but it's not coming soon; in your use-case, are you configuring the root disk or some other disk for end users? [15:14] rharper most users prefer to have all the storage directly in root , so it should appear as 1 logical volume for the client [15:15] smoser: no I didn't miss it, it just never goes down [15:15] and strictly speaking there also are "strictly monotonic" clocks [15:15] the max limit for vdi in xenserver is 2TB and we have some vms with 20TB storage , which will appear as 10x 2TB disks attached to the vm [15:15] smoser: which also guarantee to never report the same value twice [15:16] which can be hard if you need to hold that guarantee across all CPUs [15:16] not sure if Linux provides a strictly monotonic clock to userspace [15:16] some HW does provide one to the kernel [15:17] cpaelzer: i'm still confused. [15:18] google definitely says that a monotonic clock should not jump. === blackboxsw changed the topic of #cloud-init to: pull-requests https://git.io/JeVed | Meeting minutes: https://goo.gl/mrHdaj | Next status meeting June 02 16:15 UTC | 20.1 (Feb 18) | 20.2 (Apr 28) | https://bugs.launchpad.net/cloud-init/+filebug [15:25] there is clearly confusion on this. beyond just my head. [15:30] rharper: https://github.com/canonical/cloud-init/pull/391/ [15:30] meena: ^ is the proper location for that PR I pinged you with y'day. [15:34] Hani: assembling a root volume during boot isn't quite the storage plan for cloud-init; cloud-init does not run in the initramfs where one might be able to adjust a rootfs. One can construct a two-stage approach here; using cloud-init and curtin; (https://curtin.readthedocs.io/en/latest/); in curtin our testing harness does exactly this (boot an ephemeral environment with a configuration to assemble a rootfs from various disks); [15:36] Hani: cloud-init's initial storage scope is targetting non-root volumes; [15:42] rharper that what i was looking for , will check curtin Thanks [15:51] Hani: https://curtin.readthedocs.io/en/latest/topics/integration-testing.html and https://git.launchpad.net/curtin/tree/tests/vmtests #curtin if you want to discuss further [16:03] hey. i'm looking at http://paste.ubuntu.com/p/SY9n2SC6cW/ (66-azure-ephemeral.rules) [16:03] where does ATTRS{device_id} come from ? [16:04] smoser: I believe blkid/udev export those values [16:04] sda on that system gets named 'fabric_root'. [16:04] https://paste.ubuntu.com/p/4QRHZTzCdt/ [16:04] that is udevadm info --quer=all and its not there. [16:05] nor is it in /run/udev/data/b8:0 [16:07] I don't have an azure instance up just yet, but their scsi device exports those (or is supposed to export those values) [16:07] rharper: smoser@51.143.89.81 if you want to look [16:07] k [16:10] there it is in find /sys | grep f8b3781 [16:10] what rule file ? [16:11] ah, yeah, VMBUS [16:12] there are actually 2 rules files. one cloud-init, one walinux-agent [16:12] lib/udev/rules.d/66-azure-ephemeral.rules /lib/udev/rules.d/66-azure-storage.rules [16:12] $attr{file}, %s{file} [16:12] The value of a sysfs attribute found at the device, where all keys of [16:12] that's what I was looking for [16:12] one can directly read sysfs attrs [16:12] in udev rules [16:13] smoser: ok, I'm off, thanks [16:13] annoying that those sysfs attrs aren't part of udev export though. [16:14] I guess it's somewhat duplicate if you know the device path, which I think is knowable [16:20] /sys/bus/scsi/devices and /sys/class/block/sda -> ../../devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A03:00/device:07/VMBUS:01/00000000-0000-8899-0000-000000000000/host0/target0:0:0/0:0:0:0/block/sda [17:55] yeah... obvious. [17:58] mhm, yeah. very. [18:09] Odd_Bloke: rharper: uss-tableflip functional change to add a new fix-daily-recipe script https://github.com/canonical/uss-tableflip/pull/50 [18:09] I want to add docs around this release process. I'm adding notes there on testing: [18:12] blackboxsw: cool [18:33] rharper: push testing instructions [18:33] pushed rather [18:34] I can fix daily focal builds with this and then we can iterate on the README docs (or I can add it to this PR) [18:34] I had wanted to get the 'fix' in to ensure daily build recipes are fixed for focal before a new-upstream-snapshot today [18:34] * blackboxsw adds doc changes now. as it shouldn't be too much [18:37] yeah [18:37] * rharper will test [18:40] blackboxsw: https://paste.ubuntu.com/p/fnV3Jv9rTr/ [18:41] rharper: sorry that should have been fix-daily-branch -s origin/ubuntu/focal -d ubuntu/daily/focal [18:41] missing origin [18:42] ah, ok [18:42] trying again [18:42] but we should sort that and provide an upstream [18:42] +1 [18:42] that worked [18:43] continueing with instructions [18:43] remote param could probably be used for both source_branch and the daily_branch params if provided [18:43] ok can tweak it to add the $remote if not already the prefix of source_branch or daily_branch [18:44] https://paste.ubuntu.com/p/MpZnDnBHJD/ [18:44] does that look right? (it looks right to me) [18:44] schweet! [18:44] yeah [18:44] that should fix daily focal at the moment. [18:44] I'm scrubbing docs [18:44] \o/ [18:45] and Odd_Bloke is +1 on us fixing focal "manually" and we'll build/revise documents/tools ? [18:49] I think Odd_Bloke wanted us to review docs first and fix script later. they are both small so I'll push the doc changes to this branch so we can review them. [18:49] then we can land them as one [18:49] almost done w/ docs. probably 10 mins [18:53] rharper: docs pushed [18:54] rharper: one thing we could to is make sure scripts/cherry-pick emits the message (now run `fix-daily-branch -s ubuntu/ -d ubuntu/daily/`) [18:54] so that the scripts document next step (just like new-upstream-snapshot does) [18:54] that sounds nice [19:37] blackboxsw: Reviewed that PR; I'm +1 on using it locally in it's current state to get focal fixed. [19:58] thanks Odd_Bloke ! [21:25] Odd_Bloke: thanks, late , but I've pushed the changes to require and positional parameters [21:25] good thoughts [21:30] blackboxsw: Nice, thanks! [21:46] Thanks Odd_Bloke got it. [22:46] rharper: I resolved Odd_Bloke's review comments on https://github.com/canonical/uss-tableflip/pull/50 if you give it a final thumbs up I can fix ubuntu/focal daily builds [22:47] blackboxsw: yeah [22:57] thanks rharper reverted that line in the docs [22:59] ok [23:00] blackboxsw: where does gitcmd come from ? I don't see a source wrapper or anything [23:00] just curious [23:00] nm [23:00] right in my face [23:00] blackboxsw: ok, +1 [23:02] heh rharper right gitcmd was your suggestion for me on a previous PR, to catch failures :) [23:02] I figured I'd apply it here too [23:02] and thx [23:02] * blackboxsw fixes ubuntu/daily/focal now [23:04] ok daily focal recipe trying to build now. https://code.launchpad.net/~cloud-init-dev/+archive/ubuntu/daily/+recipebuild/2574992 [23:13] d'oh forgot to sync github -> LP [23:19] ok daily build recipe working on focal now [23:27] and daily build recipe for groovy was broken, it was building packages for focal ~20.04 instead of groovy 20.10 so there is a conflict on upload.