Perl 6 - the future is here, just unevenly distributed

IRC log for #puppet-openstack, 2014-08-22

| Channels | #puppet-openstack index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:30 rmoe joined #puppet-openstack
00:33 rmoe joined #puppet-openstack
01:01 topshare joined #puppet-openstack
01:07 topshare_ joined #puppet-openstack
01:28 topshare joined #puppet-openstack
01:33 xarses joined #puppet-openstack
01:35 aimon joined #puppet-openstack
02:10 xingchao joined #puppet-openstack
02:18 rmoe joined #puppet-openstack
02:22 britthouser joined #puppet-openstack
02:37 topshare joined #puppet-openstack
03:17 rmoe joined #puppet-openstack
03:39 nibalizer i think this might fix the problem with the gate
03:39 nibalizer idunno
03:39 nibalizer https://review.openstack.org/116169
03:41 crinkle hogepodge: ^
04:16 rmoe joined #puppet-openstack
04:29 tfz joined #puppet-openstack
04:31 sputnik13 joined #puppet-openstack
04:46 britthouser joined #puppet-openstack
05:43 sanjayu joined #puppet-openstack
06:29 shakamunyi joined #puppet-openstack
06:38 tfz joined #puppet-openstack
06:49 sputnik13 joined #puppet-openstack
07:06 sputnik13 joined #puppet-openstack
07:08 sputnik13 joined #puppet-openstack
07:11 sputnik13 joined #puppet-openstack
07:14 saju_m joined #puppet-openstack
07:16 sputnik13 joined #puppet-openstack
07:20 giorgiodinapoli joined #puppet-openstack
07:30 shakamunyi joined #puppet-openstack
07:31 EmilienM good morning folks!
07:31 EmilienM hogepodge: pong
07:39 sbadia morning!
07:41 xarses morning / good night =)
08:14 derekh joined #puppet-openstack
08:15 michchap_ morning :)
08:16 michchap_ fyi there is a thread on the openstack-dev ML about 'Integrated Release' where the puppet modules are mentioned. It's long but if you're bored it is worth reading if you have time.
08:20 EmilienM michchap_: it's a very long thread...
08:20 EmilienM where do you see puppet mentionned?
08:23 michchap_ EmilienM: starts with Jay Pipes' third response.
08:25 mmagr joined #puppet-openstack
08:26 EmilienM michchap_: nice catch
08:26 EmilienM michchap_: if I understand, we could try to move our modules to github/openstack
08:27 michchap_ EmilienM: that's the suggestion. I don't think it's appropriate while our master branch is not really tested much vs the upstream master branches.
08:28 jgrassler Good morning.
08:28 jgrassler michchap_: I.e. everybody tests the modules against Icehouse currently?
08:29 michchap_ jgrassler: that's my assumption yes. The majority of module consumers and a large chunk of developers are testing master modules vs stable openstack, and that needs to change to master modules vs master openstack for both parties to seriously consider making the puppet modules an integrated release.
08:30 shakamunyi joined #puppet-openstack
08:31 michchap_ jgrassler: otherwise I just don't think the quality is there. I also question if there's actually any benefit over what we do now.
08:31 michchap_ jgrassler: EmilienM our ad-hoc governance has worked pretty well for the last 2 years, I don't see any particular need to start having PTL elections and the other stuff that the official projects have.
08:32 jgrassler michchap_: well, things can change. It's what I've seen from my logging patches at any rate.
08:33 jgrassler michchap_: I have a nagging suspicion that more than just the five services I mentioned on the list will be using openstack.common.logging in Juno.
08:33 jgrassler michchap_: And yes, kudos on the ad-hoc governance.
08:33 jgrassler michchap_: more process would have been off-putting in my case.
08:36 jgrassler michchap_: being able to say 'hey, I've got some code you might be interested in' and being welcomed with "just put it out there, we'll review it" is a lot better than "well...first you'll have to write a blueprint..."
08:37 jgrassler And I kind of suspect _that_ would change with tighter integration.
08:38 michchap_ jgrassler: I'm glad the process hasn't been too painful. And as far as getting to master puppet vs master openstack I *am* working on a generic package build tool that can replace arbitrary parts of the package dependency chain with references to git repos, and will output replacement packages in deb or rpm using fpm. It is probably at least 6 months away from being more than a prototype though
08:38 michchap_ jgrassler: that would make running a version of the code that isn't from a downstream a lot easier, with the goal of doing CD
08:39 jgrassler michchap_: now that tool sounds real nice
08:39 michchap_ jgrassler: absolutely. I like how casual we can be.
08:39 jgrassler michchap_: it's something we've been drooling over increasingly over the past couple of weeks ("if only we could use stuff from master _here_" has been uttered a lot)
08:39 michchap_ jgrassler: https://github.com/jordansissel/fpm
08:40 michchap_ jgrassler: if you just want the openstack repos it's pretty easy - fpm can just make a package from setup.py, then you duplicate the rpm/deb dependencies from the downstream you're replacing
08:40 jgrassler michchap_: so that works already?
08:41 michchap_ jgrassler: yes, but no init scripts.
08:41 michchap_ jgrassler: I use fpm today to build haproxy, ruby and kibana, which I couldn't find appropriate rpms for.
08:41 EmilienM michchap_: I tend to agree with you
08:41 jgrassler michchap_: still close enough, thanks :-)
08:42 michchap_ jgrassler: here's the tool (by the same guy) for init scripts: https://github.com/jordansissel/pleaserun
08:42 michchap_ jgrassler: so you glue them together and can build a cross platform generic build system.
08:42 michchap_ EmilienM: you build your own packages right?
08:42 jgrassler michchap_: nice - I think that's just what we need
08:43 EmilienM michchap_: no
08:43 EmilienM michchap_: for RH, we use OSP5, and for debian we build upstream packages for Debian community and use them.
08:43 michchap_ jgrassler: there's another guy on here, beddari, who is also interested in this space. He advocates the omnibus approach which is a little different.
08:43 jgrassler michchap_: omnibus?
08:44 michchap_ EmilienM: what about for prod customers who need security patches straight away?
08:44 EmilienM michchap_: for debian, we fix it right away. For RH it's another story for now. We depend of RHEL & OSP5
08:44 michchap_ jgrassler: you install a minimum OS, make a single package that contains everything you want and install it under /opt
08:45 michchap_ jgrassler: it's very similar to containerisation of everything - there's a common train of thought.
08:47 michchap_ jgrassler: the general gist is that package dependencies are a pain in the ass, so let's just use packages to distribute code in a checksummed and secure manner, and offload the dependency management to the build system.
08:47 jgrassler michchap_: Ah, I've heard of this concept before. I just didn't know that name.
08:47 mmagr joined #puppet-openstack
08:48 jgrassler michchap_: I have to say I like the non-omnibus approach better.
08:48 michchap_ EmilienM: Is that where you want to be? Are you also interested in CD using packages? Maybe I'll send a mail to your colleagues about working on something together?
08:48 michchap_ jgrassler: I tend to agree. I feel it's throwing too much away. I don't like to confuse my operators.
08:48 jgrassler michchap_: for if lots of folks use the omnibus approach it's an incentive for developers to care less and less about proper release engineering.
08:49 jgrassler michchap_: i.e. packages will get rarer and rarer.
08:50 michchap_ jgrassler: That's a good point. Something I've noticed is that more and more orgs need to run a particular piece of software from source, that is not their app but merely a dependency. A good example is when Centos 5 was getting long in the tooth I had people asking for newer versions of Nginx. Having an easy way to just slip in a newer package from a source build is better than having stuff like rpmforge, which is essentially ra
08:50 michchap_ ndom versions of things.
08:51 michchap_ jgrassler: yes. It's difficult trying to convince some openstack developers to care about installs that are not from source, for example
08:53 beddari ... reading up, interesting topic yes :)
08:54 michchap_ beddari: I am going to invent a yaml format for fpm ingestion - I don't like the cookbook style ruby thing.
08:55 beddari michchap_: I figured you didn't like something yes
08:55 beddari michchap_: in my experience that will be hard, at best, because of the flexibiliy you need across platforms
08:55 michchap_ beddari: also I wrote a puppet face that pulls packages and repos out of the catalog for image building, and also for making partial local repos.
08:56 michchap_ beddari: yeah...atm I'm just using bash while I see what things I can abstract and what things need to be 'eval style'
08:56 beddari michchap_: yeah I know others that do similar ...
08:57 beddari michchap_: there's a small community around fpm-cookery at the moment, I'd love to combine efforts
08:57 beddari michchap_: but I do think what we are looking at are closer to a Makefile than a yaml definition
08:58 beddari michchap_: thus you could say I'm a proponent of continuing it beeing cookbook-style, for my use cases
08:59 michchap_ beddari: interesting. I need to mess around more to understand the problem space better. Make some mistakes and learn :) Is there a mailer for fpm-cookery or irc channel?
08:59 shakamunyi joined #puppet-openstack
09:00 beddari michchap_: mostly twitter sadly ... we're trying to do something to meet up at devopsdays Belgium
09:00 michchap_ beddari: what's your handle?
09:00 beddari same, beddari
09:01 mmaglana_ joined #puppet-openstack
09:01 michchap_ beddari: I will keep an eye out.
09:03 michchap_ beddari: I have been making slow progress since my job seems to mostly involve rewriting bad modules from the forge atm :(
09:03 beddari michchap_: this wishlist is what we got trying to move the tool forward https://github.com/bernd/fpm-cookery/wiki/wishlist
09:04 beddari michchap_: I've given up Puppet code reusage and became a happier person
09:05 beddari michchap_: once you embrace config hashes as the main interface to feed data to modules everything becomes that much easier imho
09:06 michchap_ beddari: pretty much what I've been doing. Logstash and Elasticsearch this week. I had a bit of fun writing a function that output logstash DSL from a hash.
09:06 beddari michchap_: I shouldn't say this but again, strong voices inside Puppet Labs even don't really see much point in working so hard for those common abstractions
09:07 beddari michchap_: I quote "sysadmins want to admin"
09:07 beddari michchap_: so 2014 for me is the year of relief :
09:07 beddari hehe
09:09 michchap_ beddari: There's definitely some modules that are worth working with despite their complexity, like the mysql one, or the openstack ones, but a lot of them seem to be complex because they support stupid procedures, like installing from repo/package dl/source
09:10 michchap_ beddari: I am yet to see a puppet module that does install from source in a reasonable way. Vcsrepo is not good =/
09:11 beddari michchap_: indeed
09:12 beddari michchap_: imagine a Puppet world where there were no modules, just classes, and a working mechanism for requires, provides, etc
09:13 beddari michchap_: people think in modules which is not good when what you need to build is a resource graph
09:13 michchap_ beddari: people would put everything in init instead of breaking it up properly, just like they do now :)
09:14 beddari michchap_: I guess
09:14 beddari michchap_: in this classes-only world the forge would be just for sharing providers
09:14 beddari michchap_: and environments would be instances and not horribly broken :)
09:15 beddari dreams, now back to work :P
09:15 michchap_ beddari: one can but dream :)
09:18 openstackgerrit Mehdi Abaakouk proposed a change to stackforge/puppet-neutron: Fix the subnet dhcp enable/disable param  https://review.openstack.org/116212
09:23 openstackgerrit Mehdi Abaakouk proposed a change to stackforge/puppet-neutron: Fix the subnet dhcp enable/disable param  https://review.openstack.org/116212
09:33 openstackgerrit Mehdi Abaakouk proposed a change to stackforge/puppet-neutron: Fix the subnet dhcp enable/disable param  https://review.openstack.org/116212
10:02 mmaglana joined #puppet-openstack
10:08 shakamunyi joined #puppet-openstack
10:28 sputnik13 joined #puppet-openstack
10:58 sputnik13 joined #puppet-openstack
11:01 mmaglana joined #puppet-openstack
12:19 dprince joined #puppet-openstack
12:40 openstackgerrit Risto Laurikainen proposed a change to stackforge/puppet-nova: Make user creation optional when creating service.  https://review.openstack.org/116262
12:40 openstackgerrit Risto Laurikainen proposed a change to stackforge/puppet-glance: Make user creation optional when creating service.  https://review.openstack.org/116263
12:40 openstackgerrit Risto Laurikainen proposed a change to stackforge/puppet-cinder: Make user creation optional when creating service.  https://review.openstack.org/116264
12:40 openstackgerrit Risto Laurikainen proposed a change to stackforge/puppet-neutron: Make user creation optional when creating service.  https://review.openstack.org/116265
12:56 qba73 joined #puppet-openstack
13:01 mmaglana joined #puppet-openstack
13:08 morazi joined #puppet-openstack
13:10 richm joined #puppet-openstack
13:33 rharrison joined #puppet-openstack
14:01 mmaglana joined #puppet-openstack
14:04 daneyon joined #puppet-openstack
14:13 mmaglana joined #puppet-openstack
14:22 mmaglana joined #puppet-openstack
14:28 topshare joined #puppet-openstack
14:30 shakamunyi joined #puppet-openstack
14:35 prad joined #puppet-openstack
14:41 sputnik13 joined #puppet-openstack
14:45 mmaglana joined #puppet-openstack
14:57 topshare joined #puppet-openstack
14:57 openstackgerrit Ilya Shakhat proposed a change to stackforge/puppet-tempest: Add parameter for identity V3 API  https://review.openstack.org/115914
15:19 morazi joined #puppet-openstack
15:24 cjellick_ joined #puppet-openstack
15:24 cjellick joined #puppet-openstack
15:26 xarses joined #puppet-openstack
15:28 sanjayu joined #puppet-openstack
16:30 rmoe joined #puppet-openstack
16:32 richm hello - I'm looking for some feedback about https://review.openstack.org/109676 and https://review.openstack.org/109678 - is this the right approach?
16:50 imcsk8 joined #puppet-openstack
16:50 sputnik13 joined #puppet-openstack
16:53 aimon joined #puppet-openstack
16:58 saju_m joined #puppet-openstack
17:03 saju_m joined #puppet-openstack
17:14 tfz joined #puppet-openstack
17:17 sanjayu joined #puppet-openstack
17:27 hogepodge michchap_ my solution to the installation problem was fairly crude but effective. I just built out a canonical set of sources and dependencies using pip and github sources, then hammered in any system dependencies like users and sudoers files with Puppet. Again, crude but effective.
17:28 michchap_ hogepodge: I fully intend to fix this problem properly.
17:29 michchap_ hogepodge: One of the Devstack devs asked on the list why we don't officially support master, and lack of packages is really the only thing stopping us.
17:29 hogepodge michchap_ yes, I was looking at the modules yesterday and thinking about how puppet-based packaging would work when we have Package[‘foo’] dependencies strewn throughout all of the modules.
17:30 michchap_ hogepodge: https://github.com/michaeltchapman/puppet-partial
17:30 michchap_ pull them out and funnel them into something else with a face.
17:30 hogepodge And it seems like deployment bugs that pop up are linked to packaging.
17:30 michchap_ in this case, I'm doing a straight catalog apply with the repos+packages, and over the weekend I made one that does yumdownload to build a local mirror.
17:31 openstackgerrit joined #puppet-openstack
17:32 michchap_ hogepodge: if it's a day of the week ending in 'y', I probably complained about packaging at some point :)
17:33 hogepodge Yeah. I have this ceilometer patch that tries to make sure nova-common is installed before the ceilometer user is touched.
17:33 hogepodge Makes my skin crawl.
17:33 michchap_ hogepodge: I have that bug!
17:33 hogepodge I want it backported to Icehouse, then I want to rip the user/group management out.
17:33 michchap_ my coworker has been whinging at me for weeks
17:33 hogepodge Because packaging should just handle it.
17:34 hogepodge Yeah, right now we’re waiting on the gate fix bug to pass infra.
17:34 michchap_ btw are you ok with the hash based repo thing?
17:34 hogepodge Definitely.
17:34 hogepodge I’m happy to see it go in instead of my patch.
17:34 saju_m joined #puppet-openstack
17:34 michchap_ I'm really sick atm so if you want to take it over then go ahead
17:34 hogepodge ok, I’ll try
17:35 michchap_ It's missing a 'manage_epel' convenience parameter that creates that, and probably some validation of the hash, though that might take some ruby-fu.
17:35 hogepodge ok
17:35 michchap_ Other than that I got my integration tests passing using it, so I think it's pretty good.
17:36 hogepodge Nice. I have to make a new release of the icehouse modules and tie of bow on mine, then I’ll migrate to upstream as Juno so I can test appripriately.
17:45 hogepodge EmilienM See above. I want to fix the ceilometer dependency bug on nova group being present and backport it to Icehouse, then I want to rip out Ceilometer user management like we did with nova. It’s eNovance code, though, and I wanted to check with you to see if that user/group management in Ceilometer was critical of if it’s something that would be managed by packaging anyway.
17:47 hogepodge FYI, this just went up. http://www.openstack.org/blog/2014/08/c​all-for-proposals-open-source-ecosystem​-design-sessions-at-the-paris-summit/
17:52 saju_m joined #puppet-openstack
17:58 xarses joined #puppet-openstack
18:00 hogepodge https://twitter.com/hogepodg​e/status/502877960399904768
18:19 aimon any reason noone has create a type and provider for nova host aggregates?
18:20 hogepodge aimon not that I know of
18:20 aimon I have been told I can upstream one I am about to create
18:21 saju_m joined #puppet-openstack
18:21 aimon so if all goes well we will have one soon for everyone
18:22 aimon most likely will create one for cinder host keys also
18:24 hogepodge aimon +1
18:24 hogepodge (x100)
18:32 EmilienM hogepodge: ok so. User/Group is managed by packaging and I'm not for managing it in Puppet.
18:38 hogepodge EmilienM yeah. there’s a bit of code in Ceilometer that adds ceilometer user to the nova and libvirt groups. I was wondering if it’s strictly necessary.
18:38 hogepodge It’s causing breakages because nova no longer manages groups or users by default.
18:38 EmilienM hogepodge: ceilometer agent compute needs to access to libvirt api calls
18:39 EmilienM from a system point of view, yes, it needs user/group
18:39 EmilienM but not in puppet IMO
18:39 hogepodge Cool. Thanks. So I’ll patch the bug so I can back port it, then I’ll deprecate the code.
18:40 EmilienM ok
18:44 saju_m joined #puppet-openstack
18:46 saju_m joined #puppet-openstack
19:07 michchap_ aimon: one was just abandoned
19:08 michchap_ aimon: https://review.openstack.org/#/c/70344/
19:08 aimon michchap_ any idea why?
19:08 michchap_ aimon: I couldn't get it to work, neither could hogepodge.
19:14 aimon interesting...
19:15 aimon it looks like he tries to add metadata in same operation as add...
19:15 aimon s/add/create/
19:16 aimon I was thinking a more procedural approach, e.g. create, then add host, then add metadata.. as you would on cli
19:17 aimon othwerwise you end up in exists? hell
19:18 aimon I can see that adding backwards compatability may cause pain though
19:19 aimon .. pondering whether I should just use the new API code I have access to
19:20 aimon is there a hard requirement for backwards compatability for versions of openstack?
19:21 aimon .. seems to me one could just use a previous version of the manifests.. e.g. 3.x instead of 4.x
19:34 aimon the lags between replies/comments on that submission is not encouraging.. 8 Months from start to abandon.. that is quite a barrier for anyone to contribute
19:36 aimon especially since in the end there was no error that the author could reproduce
19:46 aimon .. I’ll see if I can pick up his where he left off and maybe we can get it in
19:46 aimon seems like good code at first blush
19:46 saju_m joined #puppet-openstack
20:23 sanjayu joined #puppet-openstack
20:27 aimon sahajakoorapati https://github.com/0xdata/h2o/wiki/Install​ing-python-2.7-on-centos-6.3.-Follow-this-​sequence-exactly-for-centos-machine-only
20:31 aimon lol wrong channel on that
20:31 aimon michchap hogepodge : see my comment on https://review.openstack.org/#/c/70344/
20:32 aimon I found the exact scenario that fails
20:32 aimon it was not obvious
20:32 hogepodge +1 aimon
20:52 snk joined #puppet-openstack
20:57 aimon ok I found where int he code has the issue .. tryinig to call .keys on null data.. working through it now
20:59 hogepodge_ joined #puppet-openstack
21:03 aimon ok its fixed
21:03 aimon poor nil handling basicallyu
21:04 aimon submitting path
21:04 aimon *patch
21:14 aimon hogepodge can you reopen that commit?
21:15 aimon or di I need to resubmit ?
21:15 aimon *do
21:15 hogepodge The message called you simon.
21:15 hogepodge I’ll blame autocorrect.
21:15 aimon lol yeah
21:15 aimon though a is near s on keyboard ;)
21:16 openstackgerrit Aimon Bustardo proposed a change to stackforge/puppet-nova: Add Nova Aggregate support  https://review.openstack.org/70344
21:17 hogepodge It’s easier  to use something other than carlessness as a scapegoat.
21:18 aimon haha I am a butcher ont he keyboard.. so I never criticize others
21:18 aimon ^ case in point
21:50 aimon hogepodge do you know the cause of the error in 2.7 on that change ?
21:50 aimon ah!
21:50 aimon Could not autoload package: Could not autoload /usr/lib/ruby/vendor_ruby/puppe​t/provider/package/windows.rb: cannot load such file -- windows/error at
21:51 aimon I got that on another change I submitted
21:52 aimon is that a systemic issue that recently cropped up.. my other change is in puppet-cinder
21:52 aimon ?
21:56 britthouser joined #puppet-openstack
22:57 hogepodge aimon the test boxes were updated to trusty
22:57 hogepodge Which has puppet 3, which causes conflicts with the testing puppet 2
22:57 hogepodge there’s a patch in flight to fix it in infra
22:57 openstackgerrit Aimon Bustardo proposed a change to stackforge/puppet-nova: Add Nova Aggregate support  https://review.openstack.org/70344
22:58 aimon ^ removed extra whitespace on aggregate create which put a space at beggining of AZ names
23:29 topshare joined #puppet-openstack
23:33 hogepodge joined #puppet-openstack
23:44 aimon joined #puppet-openstack

| Channels | #puppet-openstack index | Today | | Search | Google Search | Plain-Text | summary