Cloud-init bi-weekly status
Posted on Mon 19 March 2018 in status-meeting-minutes • 11 min read
Meeting information
- #cloud-init: Cloud-init bi-weekly status meeting, 19 Mar at 16:02 — 17:09 UTC
- Full logs at [[http://ubottu.com/meetingology/logs/cloud-init/2018/cloud-init.2018-03-19-16.02.log.html]]
Meeting summary
Recent Changes
The discussion about "Recent Changes" started at 16:05.
In-progress Development
The discussion about "In-progress Development" started at 16:12.
- LINK: https://code.launchpad.net/~lp-markusschade/cloud-init/+git/cloud-init
- LINK: https://code.launchpad.net/~lp-markusschade/cloud-init/+git/cloud-init/+merge/338439
- LINK: https://code.launchpad.net/~raharper/cloud-init/+git/cloud-init/+merge/341662
Office hours (next ~30 mins)
The discussion about "Office hours (next ~30 mins)" started at 16:43.
- LINK: https://trello.com/c/vD1em9WP/698-jenkins-job-to-run-tox-tip-pylint-weekly-pin-all-lint-versions-otherwise
- LINK: https://cloud-init.github.io
Vote results
Done items
- (none)
People present (lines said)
- blackboxsw (85)
- smoser (57)
- stanguturi (16)
- ubot5` (8)
- ajorg (5)
- rharper (5)
- dpb1 (4)
- meetingology (3)
Full Log
16:02 <blackboxsw>
#startmeeting Cloud-init bi-weekly status meeting
16:02 <meetingology>
Meeting started Mon Mar 19 16:02:30 2018 UTC. The chair is blackboxsw. Information about MeetBot at http://wiki.ubuntu.com/meetingology.
16:02 <meetingology>
16:02 <meetingology>
Available commands: action commands idea info link nick
16:03 <blackboxsw>
ok, let's kick off this cloud-init bi-weekly meeting. welcome all!
16:04 <dpb1>
o/
16:04 <blackboxsw>
it's been a busy couple weeks for a few of us w/ planning meetings and vacation, but let's see what progress we've made on cloud-init.
16:05 <blackboxsw>
#topic Recent Changes
16:05 <dpb1>
smoser vacation specifically
16:05 <blackboxsw>
hehe. Generally we're tracking high-points of what lands in our trello board
16:05 <blackboxsw>
#link http://trello.com/b/hFtWKUn3/daily-cloud-init-curtin
16:06 <blackboxsw>
but from changelogs folks have made progress on azure, vmware and FreeBSD deployment targets
16:06 <blackboxsw>
- netplan: render bridge port-priority values (LP: #1735821)
16:06 <blackboxsw>
- net: recognize iscsi root cases without ip= on kernel command line.
16:06 <blackboxsw>
(LP: #1752391)
16:06 <blackboxsw>
- util: Fix subp regression. Allow specifying subp command as a string.
16:06 <blackboxsw>
(LP: #1755965)
16:06 <blackboxsw>
- This commit fixes get_hostname on the AzureDataSource.
16:06 <blackboxsw>
[Douglas Jordan] (LP: #1754495)
16:06 <ubot5
>` Launchpad bug 1735821 in nplan (Ubuntu Artful) "netplan needs bridge port-priority support" [Medium,Fix committed] https://launchpad.net/bugs/1735821
16:06 <blackboxsw>
- shellify: raise TypeError on bad input.
16:06 <blackboxsw>
- FreeBSD: Set hostname to FQDN. [Dominic Schlegel] (LP: #1753499)
16:06 <blackboxsw>
- Make salt minion module work on FreeBSD.
16:06 <blackboxsw>
[Dominic Schlegel] (LP: #1721503)
16:06 <ubot5
>` Launchpad bug 1752391 in cloud-init "cloud-init does not recognize initramfs provided network config in all cases" [Medium,Fix committed] https://launchpad.net/bugs/1752391
16:06 <blackboxsw>
- set_hostname: When present in metadata, set it before network bringup.
16:06 <blackboxsw>
(LP: #1746455) VMWare
16:06 <ubot5
>` Launchpad bug 1755965 in cloud-init "util.subp regression: no longer accept commands as string" [High,Fix committed] https://launchpad.net/bugs/1755965
16:06 <blackboxsw>
- cc_snap: Add new module to install and configure snapd and snap
16:06 <blackboxsw>
packages.
16:06 <blackboxsw>
- doc: fix all warnings issued by 'tox -e doc'
16:06 <blackboxsw>
- tests: Make pylint happy and fix python2.6 uses of assertRaisesRegex.
16:06 <ubot5
>` Launchpad bug 1755965 in cloud-init "duplicate for #1754495 util.subp regression: no longer accept commands as string" [High,Fix committed] https://launchpad.net/bugs/1755965
16:06 <blackboxsw>
- tests: fix run_tree and bddeb
16:06 <blackboxsw>
- tests: Fix some warnings in tests that popped up with newer python.
16:06 <ubot5
>` Launchpad bug 1753499 in cloud-init "hostname in FreeBSD should prefere FQDN" [Undecided,Fix committed] https://launchpad.net/bugs/1753499
16:06 <blackboxsw>
- tests: fix flakes warning for unused variable
16:06 <blackboxsw>
- tests: patch leaked stderr messages from snap unit tests
16:06 <ubot5
>` Launchpad bug 1721503 in cloud-init "salt module not able to be used on FreeBSD" [Medium,Fix committed] https://launchpad.net/bugs/1721503
16:06 <blackboxsw>
- tests: Centralize and re-use skipTest based on json schema presense.
16:06 <ubot5
>` Launchpad bug 1746455 in cloud-init "cloud-init vSphere cloud provider DHCP unique hostname issue" [High,Fix committed] https://launchpad.net/bugs/1746455
16:08 <blackboxsw>
a big thanks to dojordan (Azure) and Dominic Schlegel (FreeBSD) for patching some gaps in support as cloud-init master progresses
16:10 <blackboxsw>
On the ubuntu side of the house we got tip of tree published into Bionic thusday & friday, we are awaiting cloud-image builds which look like they are stale at 03-15-2018. once those builds are published, all clouds should be getting latest cloud-init on Bionic
16:11 <blackboxsw>
I think that's probably it for 'done' work. We have a few things in flight at the moment
16:12 <blackboxsw>
#topic In-progress Development
16:13 <blackboxsw>
Ubuntu is getting a number of new cloud-config modules:
16:13 <blackboxsw>
- new cc_snap module (deprecated cc_snappy and cc_snap_config modules) the ability to install and manage snap package
16:14 <blackboxsw>
- new cc_ubuntu_drivers: support to install 3rd party drivers on install
16:15 <blackboxsw>
- new cc_ubuntu_advantage: manage Ubuntu Advantage subscriptions for services such as Extended Security Mainenance (trusty), canonical livepatch and FIPS PPAs
16:15 <blackboxsw>
these should be landing in the week(s) to come
16:15 <blackboxsw>
and cc_snap landed already
16:17 <blackboxsw>
also there are a couple of branches that we are trying to wrap up for first class chrony support (per rharper, inspired by robjo's work)
16:17 <rharper>
blackboxsw: smoser: on the lander emails, the subject could include the git hash (or branch name); it's currently only in the body;
16:17 <smoser>
rharper: i had suggested to blackboxsw that it should acutally change to not send a subject. so it threads in your email reader with the other MP mails.
16:18 <rharper>
heh
16:18 <blackboxsw>
maybe we can toggle between the two modulus 2 :)
16:18 <rharper>
sorry, didn't meant to disturb the flow
16:19 <rharper>
continue
16:19 <blackboxsw>
yeah, we've also touched a little bit of our code landing automation this last week. powersj also is working on a git lander plugin that we might be able to use to automate landing of approved branches w/ tox test runs
16:19 <blackboxsw>
anything to free up developer time will give us more time for reviews/code
16:19 <ajorg>
should vendor-specific modules be shipped in a separate package?
16:20 <smoser>
vendor specific modules ?
16:20 <ajorg>
ubuntu_advantage
16:22 <blackboxsw>
good question/point. I hand't thought about that separation as a lot of the modules cloud-init delivers support a subset of distros
16:23 <blackboxsw>
each module has a distro attribute defined as to whether or not it will even run
16:23 <blackboxsw>
so we have spacewalk, zypper_repos etc
16:24 <smoser>
and cloud.cfg is rendered based on knowledge of the distro
16:24 <smoser>
so ubuntu_advantage wont even be in the list of config modules
16:24 <smoser>
having a static config module list is WIN in this case (but pain elsewhere)
16:25 <smoser>
at some point whe may have a more dynamic config module list.
16:25 <smoser>
but anyway... at the moment the only cost to non-ubuntu of that module being shipped is bytes on disk.
16:26 <blackboxsw>
if/when we do define that more dynamic config module list, I'd like us also to look at having configurable/separate plugin directories defined for folks providing vendor-specific content.
16:26 <ajorg>
agree that having a more dynamic config module list is prerequisite to being able to parcel out modules to other packages
16:26 <blackboxsw>
so that we don't expect folks to add plugins directly into /usr/lib/python3/dist-packages/cloudinit/config/ for instance
16:26 <smoser>
yeah. at the point when it is dynamic, the module would still declare its support for a list of distros and would be filtered out.
16:29 * ajorg is satisfied
16:30 <blackboxsw>
:). the only other thing I can think of in progress two more datasources softlayer cloud support by smoser and hetzner cloud
16:30 <blackboxsw>
so cloud-init is getting it's grubby hands into a couple of more clouds shortly.
16:31 <blackboxsw>
s/it's/its/
16:31 <blackboxsw>
it's nice to see the adoption continue to grow
16:31 <smoser>
looks like someone followed up on hetzner
16:31 <smoser>
so that hopefully is ready to land
16:31 <blackboxsw>
#link https://code.launchpad.net/~lp-markusschade/cloud-init/+git/cloud-init
16:32 <blackboxsw>
rharper: also has a couple of branches to allow cloud-init to work a bit better when rendering netplan configuration
16:33 <blackboxsw>
I think that's all for in-progress development at the moment.
16:33 <smoser>
man we need to fix that pylint thing.
16:33 <smoser>
did you mention ?
16:33 <rharper>
blackboxsw: yeah, I just pushed a fix for v1 global dns entries to get rendered under interfaces without any dns configuration
16:33 <blackboxsw>
Anything else that should be noted by anyone?
16:33 <blackboxsw>
#link https://code.launchpad.net/~lp-markusschade/cloud-init/+git/cloud-init/+merge/338439
16:33 <blackboxsw>
oops
16:34 <smoser>
cloudinit/config/cc_puppet.py:143: [W1505(deprecated-method), handle] Using deprecated method readfp()
16:34 <blackboxsw>
#link https://code.launchpad.net/~raharper/cloud-init/+git/cloud-init/+merge/341662
16:34 <smoser>
that needs fixing. it has come to us due to a new version of some of our tox environemnts. we do not fully pin the versions , only the top level packages. Ie, pylint's dependencies changed, but we only pin pylint version.
16:35 <blackboxsw>
yeah how much should we freeze our deps?
16:35 <blackboxsw>
it's kindof annoying to have your branch locally pass ci, and a fresh build of CI deps fail when you try to land
16:36 <blackboxsw>
but I don't really know whether it's worth us 'pinning' everything
16:36 <smoser>
i think pinning everything is generally the best practice for this sort of thing now.
16:37 <smoser>
oh my.
16:37 <blackboxsw>
so if we pin the world, should we also just make it a habit to occasionally tox -e tip-pylint?
16:37 <smoser>
blackboxsw: did you know you accidently fixed that in trunk ?
16:38 <blackboxsw>
smoser: I know I intentionally added a pylint ignore on that to come back and address it today.
16:38 <ajorg>
i tend to believe it's better to stay current and take your punches a few at a time so you don't have a major upset when you have to upgrade.
16:38 <smoser>
oh ko. i see.
16:39 <smoser>
ajorg: well, sor tof. if you have c-i that you want to be green, and consider it bad when it is not, then you dont want dude-on-the-internet to break you
16:40 <smoser>
there is the "good" break, where new upload to pypi identifies some lingering bug
16:40 <blackboxsw>
+1 ajorg, but I'm good (on avoiding a avalanche) if we agree to run tip-pylint target fairy regularly to avoid the landslide
16:40 <smoser>
but also the "bad" break where some upload breaks your c-i for invalid reason.
16:42 <smoser>
one huge advantage to pinning is the ability to re-create things.
16:42 <blackboxsw>
it definitely felt like last week was a lot of c-i breaks for changes unrelated to the code up for review
16:42 <smoser>
ie, if you were looking to it bisect something...
16:42 <smoser>
git bisect
16:43 <blackboxsw>
I should transition to the office hours topic so we can continue discussion
16:43 <smoser>
you can't really do that if trunk from a point in the past does not pass C-I because an external dependency changed.
16:43 <smoser>
sure we can transition to office hours.
16:43 <blackboxsw>
#topic Office hours (next ~30 mins)
16:43 <smoser>
but yeah... i want c-i on tip to not just start failing.
16:44 <stanguturi>
@blackboxsw, I have couple of requests. First, https://bugs.launchpad.net/ubuntu/+source/open-vm-tools/+bug/1724128 discussed this in last meeting as well. Any help is greatly appreciated
16:44 <ubot5
>` Ubuntu bug 1724128 in open-vm-tools (Ubuntu) "Need a Success / Failure notification mechanism when cloud-init finishes." [Undecided,New]
16:44 <blackboxsw>
that's fair. smoser can we maybe add a jenkins job to run tip-pylint then weekly. So, we don't have a huge backlog of lint failures against tip?
16:45 <smoser>
blackboxsw: i'mi fine with that... thats why we added the tip-* targets. so it was easy enough to keep current.
16:45 <blackboxsw>
+1 smoser I'll put a card up for that
16:46 <smoser>
stanguturi: your suggestion there is not a bad idea at all.
16:46 <blackboxsw>
#link https://trello.com/c/vD1em9WP/698-jenkins-job-to-run-tox-tip-pylint-weekly-pin-all-lint-versions-otherwise
16:46 <smoser>
the desire to have cloud-init tell the platform/datasource that it failed or succeeded is valid.
16:47 <smoser>
with MAAS, that his done via reporting
16:47 <blackboxsw>
stanguturi: hiya. I think we talked after that meeting about trying to allow the datasource to subscribe to a callback when cloud-init exists
16:47 <blackboxsw>
stanguturi: hiya. I think we talked after that meeting about trying to allow the datasource to subscribe to a callback when cloud-init exits
16:47 <smoser>
cloud-init reports status and results via a Reporter.
16:47 <smoser>
dojordan (i think) had also put up a request for a reproter module on azure.
16:47 <smoser>
so we do kind of have the function you're after in place.
16:48 <stanguturi>
smoser: ok. Any inputs / examples of using it will be really great.
16:50 <smoser>
stanguturi: well, the reporter interface is pretty simple. you can cloudinit/reporting/handlers.py
16:51 <smoser>
blackboxsw: http://paste.ubuntu.com/p/6KjDX8WHQH/
16:51 <smoser>
did you intend boht of those changes ?
16:51 <stanguturi>
smoser: ok. Then do I need to write a new handler for our DataSource?
16:52 <smoser>
stanguturi: well you write a reporting Handler, ankd then either system confi or optionally datasource config would turn that reporter on.
16:52 <blackboxsw>
wow smoser, no
16:52 <blackboxsw>
wow, ok, I'll put up a branch to fix that
16:52 <stanguturi>
smoser: ok. Will work on that.
16:52 <smoser>
ok.
16:53 <stanguturi>
I have another quick request about https://code.launchpad.net/~sankaraditya/cloud-init/+git/cloud-init/+ref/set_hwclock_module
16:53 <smoser>
stanguturi: do you find this is actually needed ?
16:53 <smoser>
to my knowledge the only time anyone would ever set their hardware clock to something other than UTC would be dual booting with windows.
16:54 <smoser>
which i can't seem to believe is all that a common situation in VMs
16:54 <stanguturi>
smoser: Yeah. We need this for 'VMware guest customization workflow'.
16:54 <dpb1>
smoser: oh man, I hope not
16:54 <stanguturi>
smoser: if you think, this is not worth for the base cloud-init modules, I can modify to do this change in our datasource specific modules.
16:54 <smoser>
stanguturi: does it actually solve a current problem for you ?
16:55 <stanguturi>
smoser: For 'VMware managed VMs', customers can specify in the specification file if they want UTC or localtime for the hardware clock.
16:55 <smoser>
or one that originally came in from a decade ago
16:55 <stanguturi>
smoser: Our existing customization (non cloud-init) engine does it. If we want to move to cloud-init, we want to port all our changes from our engine to our datasource in cloud-init.
16:55 <smoser>
hm... so my sugestion is really to stop allowing that :)
16:56 <stanguturi>
smoser: Oh ok. Can you please add a comment to that merge request just for the record.
16:56 <smoser>
i very well could be wrong
16:56 <smoser>
but the only time that i ever had to deal with this was when dual booting
16:56 <smoser>
with windows specifically
16:57 <stanguturi>
smoser: ok.
16:57 <smoser>
am i wrong there ?
16:57 <smoser>
i really could be.
16:58 <stanguturi>
smoser: I can discuss this within our team. But to be on par with our existing engine, want to port the changes.
16:58 <smoser>
and even if you get it wrong, generally speaking you havhe some sort of ntpdate or ntp that will fix your system clock anyway.
16:58 <smoser>
stanguturi: yeah. i understand that.
16:58 <stanguturi>
I have another request. For Ubuntu 18.04, we are planning to set 'disable_vmware_customization' flag to False by default in /etc/cloud/cloud.cfg file.
16:59 <stanguturi>
Want to know your opinion, shall we set it in cloud-init installation phase or request Ubuntu maintainers to set it in 18.04
17:00 <stanguturi>
smoser: And when is the cloud-init 18.2 scheduled for release? 3/22?
17:01 <blackboxsw>
probably a good time for us to bring that up
17:01 <smoser>
yeah. whoops.
17:01 <smoser>
:)
17:01 <smoser>
18.2 is scheduled for 3/22 (thursday)
17:02 <blackboxsw>
We cloud-init 18.2 have it scheduled for an arbitrary 3/22 date, we'd like to slip that out to next week Tuesday 3/27
17:02 <smoser>
but amoung our internall team we decided to push that to 3/27
17:02 <stanguturi>
smoser: ok. Thanks for the update.
17:02 <blackboxsw>
there a a few in flight branches, azure, softlayer etc that we'd like to get in and get tested before 18.2
17:02 <smoser>
we will send an email today or tomrrow witih "pending release" like subject like we've done before.
17:03 <blackboxsw>
dpb1: others any objections to cutting the 18.2 release on Tuesday 3/27?
17:03 <dpb1>
none
17:04 * blackboxsw adds the upcoming date to the topic
17:05 <blackboxsw>
... ok, folks interested in discussing today?
17:08 <blackboxsw>
#link https://cloud-init.github.io
17:09 <blackboxsw>
The above link will have our captured logs for this meeting.
17:09 <blackboxsw>
thanks again for tuning in
17:09 <blackboxsw>
#endmeeting
Generated by MeetBot 0.1.5 (http://wiki.ubuntu.com/meetingology)