Perl 6 - the future is here, just unevenly distributed

IRC log for #puppet-openstack, 2014-10-14

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

All times shown according to UTC.

Time Nick Message
00:00 hogepodge I'm probably going to be online hacking on devstack, sahara, and trove all night, so I'm around to answer questions.
00:01 LeslieCarr thanks, are you in paris for openstack ?
00:01 LeslieCarr sadly I’m not
00:01 xarses Looks like I will be
00:06 britthouser joined #puppet-openstack
00:22 rmoe joined #puppet-openstack
00:54 thumpba joined #puppet-openstack
00:54 imcsk8 joined #puppet-openstack
01:04 shakayumi joined #puppet-openstack
01:10 xarses joined #puppet-openstack
01:30 sputnik13 joined #puppet-openstack
01:32 rcallawa joined #puppet-openstack
01:41 crinkle LeslieCarr: o/  <-- @pdx_krinkle
01:42 crinkle hogepodge: is still the expert on that module, I'm still getting use to it :)
01:45 hogepodge crinkle Leslie it's a bit weird. Can we just say that mistakes were made in good intentions? Feel free to kill my darlings, every single one of them.
01:45 crinkle lol
01:45 hogepodge http://www.urbandictionary.com/define.php?term=Kill+your+darlings
01:48 emccormickva Hey guys, if I'm starting from scratch on an HA deployment, what's the best config module to use? I'm looking to break down the services beyond all-in-one controller boxen (ie. separate galera, neutron, mongo)
01:48 emccormickva is it worth starting with puppetlabs-openstack given the deprecation warnings on it? Or something like builder?
01:49 crinkle puppetlabs-openstack isn't deprecated (I hope?) but it doesn't have any HA functionality built-in
01:50 crinkle puppet-openstack != puppetlabs-openstack
01:50 emccormickva sorry, it was puppet-openstack. My bad
01:51 emccormickva so am I better off hacking in the HA bits separately for puppetlabs-openstack? Or use puppet_openstack_builder? Looks like that's making use of all the new openstacklib swankiness.
01:53 emccormickva Also, at some point I'd like to get up to speed enough with what the infra team is doing to pitch in a bit, so it might be good to stay in that neighborhood if it's not too crazy ;)
01:56 shakamunyi joined #puppet-openstack
02:02 crinkle emccormickva: HA stuff is sort of on the roadmap for puppet-openstack_extras (I think, still seems kind of up in the air) but there's nothing really in place yet
02:02 crinkle I'm not familiar with puppet_openstack_builder
02:02 crinkle I use puppetlabs-openstack so I'd recommend it :)
02:04 emccormickva ok, hackathon it is ;)
02:07 emccormickva any major complication with it to breaking out things like mysql / rabbit / mongo to separate hosts? I can do all the galera, haproxy and whatnot outside if this and just dump in the IPs hopefully?
02:07 xingchao joined #puppet-openstack
02:07 crinkle there is an issue where the rabbit host isn't very configurable https://github.com/puppetlabs/puppetlabs-openstack/issues/93
02:08 crinkle I'm not sure how it does mysql and mongo
02:08 emccormickva and I think puppet_openstack_builder is Michael Chapman's thing.
02:08 michchap crinkle: hey, daylight savings has pushed the meetings to 2am-3am for me so I can't really make it anymore
02:09 crinkle michchap: check the mailing list, we're thinking of rescheduling it
02:09 crinkle please chime in with preferred times
02:09 emccormickva oh speaking of Michael Chapman *heh* ;)
02:10 michchap emccormickva: hey, so the builder approach is a bit flawed and I want to redo it but haven't had time. It's a pure hiera design so roles and profiles are defined in hiera instead of as classes
02:11 michchap emccormickva: this makes plugin choices really messy, but on the other hand it makes picking apart the builder quite easy, the HA bits and bobs are in here: https://github.com/michaeltchapman/puppet-stacktira
02:12 michchap emccormickva: this is missing pieces that Fuel has, like pacemaker-izing every service for fencing, but it pushes everything through Haproxy except rabbit, which does its own clustering thing.
02:13 emccormickva So despite it's not quite where you want it state, do you think that's probably the closest thing to ready for doing an HA deploy?
02:13 michchap emccormickva: the galera bits are handled by https://github.com/michaeltchapman/puppet-galera, which works by overriding a few bits in puppetlabs-mysql
02:15 emccormickva Yeah the Galera one looked nice. I was gonna use that with whatever I came up with.
02:16 emccormickva I'm not overly worried about it being perfect, just close enough to where I can hack in the bits if need be. Somehow someone thinks I'm going to get this stood up before I leave for Paris ;).
02:16 michchap emccormickva: Best option today if your goal is to get an HA deployment that works reliably is Fuel. The downside is that it's a fork, and it's pretty much only a Mirantis project as far as I know. If you want to use the community modules then there's the eNovance toolset https://github.com/enovance/puppet-openstack-cloud and the doc for the builder http://stacktira.aptira.com/
02:18 emccormickva yeah definitely wanna go community. I like drinking with the Mirantis guys, but since it's all in the web thingy now it's got the same issue that Staypuft does. Everything on all control nodes.
02:19 michchap emccormickva: the builder also has everything on all control nodes, but I know that Cisco have a very heavily modified version that splits things out. It's certainly possible and probably not too much work.
02:19 emccormickva and no good way to break it out from what I saw. Plus I'd like to start pitching in here once I'm competent enough at it from the puppet side of things. Thanks for the tips. I'll look at the STAcktira / eNovance stuff too
02:21 michchap There will be plenty of us in Paris, so come along to the puppet session and there will almost certainly be discussion around this topic. We are in a very strange state right now where we don't have a very good entry point for newcomers to the project.
02:21 emccormickva thought Cisco were behind the builder or did they fork that?
02:21 michchap I wrote the builder with bodepd while at Cisco as part of a product, it was forked by an internal team for an internal deploy
02:22 michchap so it's a bit messy. spredzy has something quite neat here: https://github.com/Mylezeem/puppeels
02:22 michchap which I'd like to push forward, along with example hiera data for sample deployments.
02:23 emccormickva Got it. Hard to keep track of all of these forks of forks
02:23 michchap yep. It's a mess. We haven't updated the builder in a month or so because we're not sure what to do with it right now.
02:24 michchap there is a need for a set of sample data, and that data should feed into CI, but progress is very slow because everyone has actual clouds they have to manage.
02:25 emccormickva Always the curse of these things. That's part of why I am trying to avoid going down a dead end. If I can go with the thing that's got some support then I can contribute back to make it better.
02:26 emccormickva also definitely want community modules as I'm trying to go straight to Juno with this one.
02:30 benh57 joined #puppet-openstack
02:34 shakamunyi joined #puppet-openstack
02:35 sputnik13 joined #puppet-openstack
02:45 thumpba_ joined #puppet-openstack
02:45 shakayumi joined #puppet-openstack
02:47 britthou_ joined #puppet-openstack
02:48 f10d4_ joined #puppet-openstack
02:51 xavpaice_ joined #puppet-openstack
02:51 abramley_ joined #puppet-openstack
02:52 mwoodson_ joined #puppet-openstack
02:52 beddari_ joined #puppet-openstack
02:54 dmsimardz joined #puppet-openstack
02:54 ken_barber joined #puppet-openstack
02:54 ashp joined #puppet-openstack
02:54 holser joined #puppet-openstack
02:55 xarses joined #puppet-openstack
02:58 dkaigarodsev joined #puppet-openstack
03:02 LeslieCarr joined #puppet-openstack
03:02 britthouser joined #puppet-openstack
03:05 TorLaedre joined #puppet-openstack
03:06 vissborg joined #puppet-openstack
03:08 sputnik13 joined #puppet-openstack
03:20 rlaurika joined #puppet-openstack
03:21 mgagne joined #puppet-openstack
03:27 brianvans joined #puppet-openstack
03:43 mschmitt joined #puppet-openstack
04:13 sputnik13 joined #puppet-openstack
04:16 daneyon joined #puppet-openstack
04:37 sanjayu joined #puppet-openstack
05:20 mfisch joined #puppet-openstack
05:46 Guest18588 joined #puppet-openstack
05:46 mschmitt joined #puppet-openstack
05:46 social joined #puppet-openstack
05:46 linux-uzer joined #puppet-openstack
05:46 dachary joined #puppet-openstack
05:46 sbadia joined #puppet-openstack
05:46 tlbr joined #puppet-openstack
06:18 sputnik13 joined #puppet-openstack
06:38 mmaglana joined #puppet-openstack
06:56 dgurtner joined #puppet-openstack
07:23 linux-uzer joined #puppet-openstack
07:37 saju_m joined #puppet-openstack
07:39 arnaud_orange joined #puppet-openstack
07:46 stamak joined #puppet-openstack
07:52 beddari hmmm all the pacemaker talk and discussion kind of seem ... not really where you want to be, once you got more programmatic control over your L3 routing
07:54 beddari LeslieCarr: you would be in a position to talk about L3 and BFD for resilience/HA right? :)
08:23 qba73 joined #puppet-openstack
08:27 shakamunyi joined #puppet-openstack
09:14 rcallawa joined #puppet-openstack
09:15 rcallawa joined #puppet-openstack
09:23 openstackgerrit Stepan Rogov proposed a change to stackforge/puppet-neutron: Adding l3_ext network type support for neutron VMWare NSX plugin  https://review.openstack.org/128217
09:27 shakamunyi joined #puppet-openstack
09:46 openstackgerrit Oliver Böttcher proposed a change to stackforge/puppet-cinder: correct package name for cinder backup  https://review.openstack.org/128223
09:48 michchap beddari: according to the fuel devs it's still handy to have for fencing even in active/active situations.
09:50 rushiagr_away joined #puppet-openstack
09:52 openstackgerrit Stepan Rogov proposed a change to stackforge/puppet-neutron: Adding l3_ext network type support for neutron VMWare NSX plugin  https://review.openstack.org/128217
09:54 gildub joined #puppet-openstack
10:16 openstackgerrit Oliver Böttcher proposed a change to stackforge/puppet-cinder: correct package name for cinder backup  https://review.openstack.org/128223
10:18 beddari michchap: hmm well fencing what then, db access? that would be handled by the db cluster itself as far as I understand
10:18 beddari michchap: if a frontend is on a partition with no write access then don't route traffic there
10:18 beddari michchap: I might not understand the full picture
10:20 beddari michchap: I might be wrong but to me it looks more like a "who owns what at what level" debate most times pacemaker vs routing comes up
10:24 michchap beddari: I think the fencing is to prevent flapping nodes from entering the cluster, for example
10:25 michchap beddari: but you can ask the Mirantis folks in Paris. They will probably have a better answer for you.
10:26 mmagr joined #puppet-openstack
10:30 beddari michchap: yeah I will, think they'll have a hard time convincing me that pacemaker is worth it vs alternatives :P
10:33 michchap beddari: it will go into openstack_extras so you can always not use it.
10:33 beddari michchap: and well L3 routed HA does not need to be active/active, basically the difference is that you route instead of moving an IP over L2, which is A LOT better :P
10:34 michchap beddari: I don't think it's really meant for l3 agent - DVR solves this much more neatly.
10:34 beddari michchap: yeah I know, I was just commenting on the general "pacemaker it" trend
10:36 beddari michchap: like Rackspace RPC9 hehe, where they even specify two lb frontend nodes ... they could have used OSPF+BFD from thre to the controllers :)
10:37 michchap beddari: I agree that native clustering that actually works is preferable - nobody is going to use pacemaker on etcd or consul for example, but if people find it useful because some parts of the behemoth that is the openstack software stack are shit at dealing with clustering, then I'm fine with it
10:37 beddari "solutions first" elegance next year
10:37 beddari or millenium
10:38 michchap beddari: exactly. bigger fish to fry right now for just about everyone
10:39 michchap beddari: I'll take a look at the giftwrap thing. I made a bit of progress writing a little tool that could take rpms from a distro and replace the source inside them by taking the metadata and a git repo and then building a package with fpm.
10:39 beddari well imho if you build something from scratch with openstack and don't integrate routing or even think about stuff like ipv6 ... then ... you are missing out :P
10:41 michchap beddari: so I have a way to replace distro packages with my own source, without messing up dependencies and other stuff, but it's still a long way from being useful.
10:41 beddari michchap: this slide pretty much sums it up http://www.slideshare.net/randybias/open-stack-designsummithapairsarenottheonlyanswer-copy/20
10:42 beddari michchap: I'm really sad that session didn't resound more hehe
10:42 beddari michchap: I think I could blame Red Hat even haha
10:43 beddari michchap: re the omnibus approach, yes, I saw the ribbon stuff yesteday
10:43 beddari michchap: didn't really get the model where you download-unpack existing packages
10:44 michchap beddari: I made a google doc with a diagram for what I wanted and did a quick bit of work to see if I liked it, but there's a bunch of CI/CD talks in Paris which will involve packaging so I might see what comes up there.
10:45 beddari I can tell you none of them would understand the idea from the same angle
10:45 michchap beddari: well the idea is that you take things from the distro and replace the source without messing up the dependency tree
10:45 michchap beddari: then at the end, you can make an omnibus, or you can use the package as normal
10:46 michchap beddari: and you can override any of the tags in the rpm, so if you need to modify any dependencies, that's possible, and if you need to change the init scripts, there's a wrapper for pleaserun.
10:47 beddari so the point of replacing the package source is just to keep the deps chain intact?
10:48 michchap beddari: and init scripts and created directory metadata and pre/post un/install scripts and all that stuff
10:49 beddari in my experience I just don't want to bother with that, I look at the application, then decide at what level I want distro packages (for maintenance)
10:49 michchap beddari: I guess I just look at the sheer size of the work that is already done by the distro and I think 'Hopefully I can take advantage of that'
10:49 beddari so for Python I'd want the runtime and most likely not much else, typically I'd package what goes into a virtualenv
10:50 beddari no I want packagers to move upstream :)
10:50 beddari .. and it is not just me :)
10:51 beddari Docker, all these things, they rely on upstream more than ever
10:51 beddari I think the role of the distros is keeping the base secure and stable
10:51 beddari openstack is a typical project where omnibus+upstream stable maintenance would be of enormous profit
10:52 michchap beddari: it would, but there is no team doing that.
10:52 beddari I know, and read the thread
10:52 beddari and it is bad
10:52 michchap beddari: if there were, I agree! I'd just use that instead of the distro!
10:52 beddari so then we should do it, no ? :)
10:53 michchap beddari: Anyone running OpenStack in prod knows how pointless it is staying on stable branches.
10:53 michchap beddari: so many bugfixes don't get backported, and tons of them are things like DVR or heat multi-engine where it can't really be backported
10:53 beddari well but at least if we got the packages produced in a simple way (with as simple as possible, but sensible, deps towards the distro level, like Craig explained) then that is a way forward
10:54 michchap beddari: so the end result is that stable branches are just buggy and terrible compared to master, within a pretty short period
10:54 beddari culture
10:54 michchap beddari: is there a session on it? Zigo has been making lots of noise about the debian packages on the list
10:55 michchap beddari: I did take a look at his team's build system. It was very difficult for me to figure out.
10:55 beddari I don't know, I just read the thread on openstack-dev about stable maintenance and then there was some nodding that yes this is bad
10:55 beddari there was some guy from Red Hat that said "oh well we might as well close it completely" as in fork and maintain internally
10:55 michchap beddari: well hopefully there is a session and you can go rant and maybe something will happen.
10:55 beddari which is just "#¤"#
10:56 beddari :P
10:56 michchap yeah but you realise many places do that already, right?
10:56 beddari sure
10:56 beddari but I'm a Red Hat customer
10:56 michchap and upstreaming code is such a massive pain in the ass who can blame them
10:56 beddari so I'm allowed to make them what I want them to be :P
10:56 michchap well...you can try
10:56 beddari hehe
10:57 beddari well, see, so if we made stable consumable then that might be a push in a positive direction
10:58 beddari which was Craig Traceys idea originally I'd say
10:58 igajsin1 joined #puppet-openstack
10:58 beddari all the other problems I can't fix but this is doable
10:59 beddari I'll try reading the stuff about the Debian side
11:00 beddari michchap: so, kind of interesting, are they working within a vendor for the Debian packages or just voulenteering?
11:01 michchap beddari: I think Thomas Goirand might be eNovance?
11:01 michchap beddari: but I'm not sure what that means post acquisition longer term
11:01 beddari well that is now Red Hat hehe
11:01 michchap beddari: not really my place to say
11:02 beddari Ubuntu packaging is separate?
11:03 beddari (remark: oh I cringe each time I think of all the duplicated work)
11:03 michchap beddari: I looked at their build logs and got scared and ran away
11:03 beddari haha yeah
11:03 michchap beddari: the Ubuntu stuff was totally unintelligable
11:03 michchap beddari: I guess they have their own tools. I couldn't really figure out what was going on
11:03 beddari launchpad most likely
11:04 michchap beddari: there seemed to be a worrying amount of bzr
11:04 beddari so making "pure" openstack consumable for users should be VERY HIGH pri
11:05 beddari just from our little chat here it is kind of obvious
11:05 michchap beddari: doesn't anvil do that?
11:05 beddari haven't looked at it as I saw Craig did and tossed it
11:06 michchap beddari: the part it was missing for me was being able to also build non-python things that I need. If I'm going in on packaging I want to be able to build other stuff that I need as well.
11:06 beddari yup
11:06 derekh joined #puppet-openstack
11:06 beddari 90% of all tools are non-abstracted
11:06 michchap *sigh*
11:06 beddari single purpose
11:06 beddari hehe I
11:06 michchap except the distro ones, and they're not very easy to work with.
11:06 beddari I've been thinking about this for years
11:07 michchap I had a look at koji briefly.
11:07 michchap the lower level bits are all OK - like rpmbuild and stuff.
11:07 beddari I'm probably going to become a much better programmer just to be able to solve it for the general case
11:08 beddari yes I've been using it at various jobs but well only from the outside
11:09 beddari someone had a nice idea once I discussed this, gentoo portage seems to be the best fit for the build system :)
11:09 beddari then just tack on fpm and go
11:10 michchap yeah I had a similar thought. Portage is raved about by all the gentoo users
11:10 beddari I don't know if it can be scaled down enough to not take too much mindshare, if it can't be simple then it can't be used
11:11 beddari nixpkgs is another
11:11 beddari they even almost started packaging openstack (but gave in to the time required it seemed)
11:11 michchap right
11:12 michchap it did occur to me that building packages for gentoo might actually be a cool place to start
11:12 michchap 'packages'
11:12 beddari it blurs the distro-package vs omnibus line though
11:13 beddari but if I could get portage on centos
11:13 beddari :)
11:13 michchap anyway, nice chatting to you. I gotta take a break. Let me know if you hear of a packaging session in Paris and I'll be there.
11:13 beddari sure, thanks
11:31 beddari michchap: if you havent looked at nix yet, this is informative https://www.youtube.com/watch?v=Eis-WqHda20
11:42 EmilienM Good morning
12:16 social oh hy, did anyone try debugging the puppet parser?
12:17 social I have regsubst which is inside of complex define and takes string from ${name}
12:17 social and it just returns ${name} :(
12:28 jproulx1 left #puppet-openstack
12:34 mmagr joined #puppet-openstack
12:35 shakayumi joined #puppet-openstack
12:37 arnaud_orange joined #puppet-openstack
12:56 beddari social: pastebin something (prefererrably runnable)
12:57 social beddari: simple case it works :)
12:58 social anyway /usr/share/ruby/vendor_ruby/puppet/parser/functions/regsubst.rb < result = target.send(operation, re, replacement)
12:58 social output of working and than non working cases < http://paste.fedoraproject.org/141732/41329129
13:00 morazi joined #puppet-openstack
13:00 aarefiev__ joined #puppet-openstack
13:00 social aaaa
13:00 social regexp
13:00 social /o\ it's just wrong
13:01 aarefiev_ joined #puppet-openstack
13:02 mschmitt joined #puppet-openstack
13:13 stamak joined #puppet-openstack
13:14 emccormick joined #puppet-openstack
13:29 mschmitt1 joined #puppet-openstack
13:32 arnaud_orange joined #puppet-openstack
13:41 openstackgerrit Jiri Stransky proposed a change to stackforge/puppet-cinder: Set volume_driver for iscsi backend explicitly  https://review.openstack.org/128283
13:49 mschmitt joined #puppet-openstack
13:53 dgurtner joined #puppet-openstack
13:54 dgurtner joined #puppet-openstack
14:22 ericpeterson joined #puppet-openstack
14:29 saneax_-_ joined #puppet-openstack
15:03 britthou_ joined #puppet-openstack
15:14 mdorman joined #puppet-openstack
15:15 pradk joined #puppet-openstack
15:24 doo joined #puppet-openstack
15:44 aimon joined #puppet-openstack
15:52 brianvans joined #puppet-openstack
15:53 hogepodge EmilienM or sbadia
15:53 EmilienM hogepodge: ?
15:54 hogepodge Either of you running the sahara module?
15:54 EmilienM hogepodge: we don't.
15:54 hogepodge :-/
16:02 LeslieCarr beddari: hehe, i’m just trying to wrap my head around icehouse and all the pieces ;)  Last time i set up openstack was I think grizzly
16:16 mgagne joined #puppet-openstack
16:19 mschmitt joined #puppet-openstack
16:20 sputnik13 joined #puppet-openstack
16:28 rmoe joined #puppet-openstack
16:48 shakayumi joined #puppet-openstack
16:58 britthouser joined #puppet-openstack
17:33 mmaglana joined #puppet-openstack
17:34 xarses joined #puppet-openstack
17:55 doo joined #puppet-openstack
18:08 doo joined #puppet-openstack
18:13 openstackgerrit Aleksandr Didenko proposed a change to stackforge/puppet-swift: Allow to customize log_name for services  https://review.openstack.org/128385
18:14 doo joined #puppet-openstack
18:28 morazi joined #puppet-openstack
18:38 openstackgerrit Emilien Macchi proposed a change to stackforge/puppet-nova: Debian: update libvirtd init script path  https://review.openstack.org/128399
18:38 EmilienM sbadia: spredzy ^
18:39 EmilienM mgagne: if you are around too
18:40 mgagne EmilienM: wouldn't it be great if there was a place where all source packages repositories were documented? =)
18:41 EmilienM mgagne: could be
18:41 mgagne EmilienM: hehe, I wish a lot but don't do much lately :-/
18:41 mgagne EmilienM: hmm, libvirt, not an openstack package
18:42 EmilienM mgagne: packaging is a pain
18:42 mgagne EmilienM: yep :-/
18:42 EmilienM mgagne: let me give more doc please. Make sense. It was obvious to me after the investigation.
18:48 doo joined #puppet-openstack
18:50 mgagne EmilienM: this change happened in sid only, we should wrap it in a conditional to still support wheezy
18:50 EmilienM mgagne: http://anonscm.debian.org/cgit/pkg-libvirt/libvirt.git/tree/debian/libvirt-bin.maintscript
18:50 EmilienM gotcha ^^
18:50 mgagne https://packages.debian.org/wheezy/amd64/libvirt-bin/filelist vs https://packages.debian.org/sid/amd64/libvirt-daemon-system/filelist
18:51 EmilienM yep
18:51 EmilienM mgagne: http://anonscm.debian.org/cgit/pkg-libvirt/libvirt.git/commit/debian/libvirt-bin.maintscript?id=28dd3618a9a2a8f769d1fa12c28647c4599d9e8e
18:52 EmilienM 2 days ago...
18:53 mgagne EmilienM: but I'm not sure how we should support this. I mean, is our master branch supposed to only support the latest packages from the latest Debian version? Could you install Juno on Wheezy? What about people pinning packages on Wheezy? etc.
18:54 EmilienM mgagne: it's hard.
18:54 mgagne EmilienM: hehe
18:54 EmilienM mgagne: and when it's hard... let's add a new param in defining the default with previous to keep backward compat !
18:54 mgagne jesus
18:55 EmilienM lol
18:55 EmilienM mgagne: you can call me like this, it's all right
18:55 EmilienM mgagne: seriously, what do you think ?
18:55 mgagne EmilienM: haha "Yes my son?"
18:55 EmilienM \o/ you crazy
18:55 EmilienM ok I push a new patchset
18:59 ianw joined #puppet-openstack
19:01 gildub joined #puppet-openstack
19:04 shakayumi joined #puppet-openstack
19:08 mschmitt joined #puppet-openstack
19:09 openstackgerrit Emilien Macchi proposed a change to stackforge/puppet-nova: Debian: update libvirtd init script path  https://review.openstack.org/128399
19:09 EmilienM mgagne: I tried something ^
19:09 openstackgerrit Emilien Macchi proposed a change to stackforge/puppet-nova: Debian: update libvirtd init script path  https://review.openstack.org/128399
19:10 mgagne EmilienM: how about a conditional in params with os version detection, inherits from params and use it to define a reasonable value?
19:10 EmilienM mgagne: as you said earlier, it's a bit more complicated with people which use to ping packages, and so.
19:11 EmilienM os detection is a bit dangerous for this use case.
19:11 mgagne EmilienM: they will still have the ability to override the file path
19:11 EmilienM ah
19:11 mgagne EmilienM: just implement a decent default value logic based on the os version
19:12 EmilienM mgagne: so you want to use ::distrelease ?
19:12 EmilienM or operatingsystemrelease
19:12 mgagne EmilienM: I'm not sure which one we should use tbh
19:13 mgagne EmilienM: lsb vs operatingsystemrelease, what are we using elsewhere?
19:13 EmilienM mgagne: and a fact which detect the libvirt release?
19:14 mgagne EmilienM: it's getting complicated lol
19:14 mgagne EmilienM: we should write a libvirt module then :p
19:14 EmilienM spredzy: ^
19:14 mgagne EmilienM: but detecting the os version should be enough for now
19:14 EmilienM I think the fact idea is good, but a bit overkill
19:15 spredzy It is overkilled, but remains the 'most accurate' way to determine the file name
19:15 mgagne EmilienM: those pinning packages know what they are doing and should be able to override the file path
19:15 EmilienM mgagne: I'm running 7.6. Can you confirm it starts by 7.6 ?
19:15 mgagne EmilienM: I don't run Debian
19:15 EmilienM see how you should \o/
19:15 mgagne EmilienM: I use package sources to verify
19:15 mfisch joined #puppet-openstack
19:18 EmilienM mgagne: there is way to detect sid.
19:18 mgagne EmilienM: is there no version number associated to sid?
19:18 EmilienM mgagne: I would stay on this design for now because 1/ it maintains backward compatibility at all, 2/ those pinning packages know what they are doing and should read the doc
19:18 EmilienM mgagne: nop
19:19 mgagne EmilienM: why not? what does facter report?
19:19 EmilienM mgagne: I'm running Wheezy with this package.
19:19 EmilienM so 7.6
19:19 mgagne EmilienM: so you are the guy pinning packages? :P
19:19 EmilienM lol
19:19 EmilienM I'm the guy who needs this paraem
19:20 EmilienM mgagne: so i think the last patchset is enough
19:24 hogepodge So who's working on puppet-sahara?
19:24 hogepodge I'm running a hacked up version of it now, but if anyone wants to spend some time with me this weekend bringing it up to speed that would be great.
19:25 hogepodge Lots of things have changed in the last few months.
19:25 EmilienM hogepodge: maybe some mirantis folks? I know they work on the upstream project
19:25 EmilienM xarses: ^
19:25 hogepodge Ok, I should be looking at their upstream code.
19:26 EmilienM hogepodge: https://github.com/stackforge/fuel-library/tree/master/deployment/puppet/sahara
19:30 EmilienM Error: /Stage[main]/Nova::Compute::Libvirt/Service[libvirt]: Could not evaluate: Could not find init script for 'libvirt-bin'
19:30 EmilienM mgagne: my patch is not finished... ^
19:31 EmilienM they don't like backward compatibility in debian
19:32 hogepodge EmilienM they never change anything, so when the do they want to make it an event.
19:33 EmilienM this event made my day
19:35 EmilienM oh actually there is a nova::compute::libvirt::libvirt_service_name
19:38 holser hogepodge, ask holser, bogdando, dilyin in #fuel-dev regarding puppet-sahara. I will help with your questions
19:39 EmilienM holser: would be great to keep the conversation here, since puppet-sahara is on stackforge :)
19:39 holser oki, doki
19:40 EmilienM holser: I haven't seen deeply the code, but fuel-library is using puppet-sahara from stackforge, right?
19:40 holser I think, it is not
19:41 holser though, we are in process of merging with upstream
19:41 EmilienM oki, doki
19:41 dilyin hogepodge, your libvirt service is not started becayse there is no init script for it. is it actually installed? anyway it's not related to sahara
19:42 EmilienM dilyin: you may talk about my issue in fact
19:42 hogepodge dilyin, yup, I think that we have wires crossed.
19:42 stamak joined #puppet-openstack
19:42 dilyin EmilienM, i haven't read the thread. what issue?
19:43 EmilienM dilyin: sid
19:43 EmilienM oops
19:43 EmilienM https://review.openstack.org/#/c/128399/
19:44 dilyin EmilienM, init file name is different from service name? looks reasonable
19:44 EmilienM dilyin: my original issue is this change: http://anonscm.debian.org/cgit/pkg-libvirt/libvirt.git/commit/debian/libvirt-bin.maintscript?id=28dd3618a9a2a8f769d1fa12c28647c4599d9e8e
19:44 EmilienM in SID
19:44 EmilienM mv /etc/default/libvirt-bin /etc/default/libvirtd
19:44 EmilienM very nice!
19:45 dilyin EmilienM, yes /etc/init.d/libvirtd looks better. why should it be different from service name
19:45 EmilienM dilyin: my concern is more backward compatibility management here.
19:46 EmilienM dilyin: I 100% agree with the change, but not the way to do it.
19:46 EmilienM anyway, let's fix it in puppet
19:48 dilyin EmilienM, maybe you can juse set service name in params to libvirt-bin
19:48 EmilienM dilyin: the service name is changeable in nova::compute::libvirt
19:49 EmilienM my issue is with the init script path
19:49 dilyin EmilienM, init script name == service name as far as Puppet is consirned
19:50 dilyin EmilienM, and actually I don't quite understand your pull request, what do you want to do
19:51 holser dilyin Debian just renamed
19:51 holser libvirt-bin to libvirtd
19:51 EmilienM dilyin: actually, I could use libvirt_service_name, indeed
19:51 holser and it breaks backward compatibility
19:51 EmilienM dilyin: thanks you have me an idea
19:51 EmilienM gave*
19:51 holser libvirt_service_name*
19:52 dilyin EmilienM, $libvirt_init_file = '/etc/default/libvirt-bin' this is defaults file not init file
19:53 holser dilyin, look at change set
19:53 * holser mv /etc/default/libvirt-bin /etc/default/libvirtd
19:53 holser mv /etcinit.d/libvirt-bin /etc/init.d/libvirtd
19:54 dilyin EmilienM, yes, your are right to rename this file, but variable name is misleading
19:54 EmilienM dilyin: let me product a new patchset, you will see how I fix it, maybe i'm wrong but let me 2 min
19:54 dilyin EmilienM, and you should also rename libvirt_service_name perhaps in params.pp
19:56 dilyin EmilienM, there is also another path: create symlinks from new file names to old file names
19:56 EmilienM nop
19:56 EmilienM no symlink
19:56 EmilienM some people tried to kill me when I did that in puppet-neutron
19:57 dilyin EmilienM, yes, symlinks are usually the last resort
20:04 shakayumi joined #puppet-openstack
20:05 morazi joined #puppet-openstack
20:14 mgagne Anyone knows OpenSUSE? My question is: how do you update it from 13.0 to 13.1? Do you just have to update to the latest packages and you are done or do you need to do a "dist-ugrade" ?
20:16 EmilienM mgagne: not sure you will find many support here. AJaeger mays know though
20:16 mgagne https://en.opensuse.org/Upgrade "sed -i 's/12\.3/13.1/g' /etc/zypp/repos.d/*" well this lets me suggest you have to update a version somewhere to upgrade to the next version
20:20 shakayumi joined #puppet-openstack
20:39 xarses EmilienM: hogepodge reading the scrollback
20:41 xarses looks like you guys found some help from the other guys. Let me know if you still need a hand and I'll look to see who can help with sahara
20:43 EmilienM xarses: thanks!
20:56 EmilienM spredzy: mgagne dilyin holser : could you review https://review.openstack.org/#/c/128399/ ?
20:56 holser EmilienM, let me have a look
20:58 holser Package['libvirt'] -> File_line<| path == "/etc/default/${::nova::compute::libvirt::libvirt_service_name}" |> is too long
20:59 holser require => Package['libvirt'] seems to be better
20:59 holser before*
20:59 holser sorry
21:01 EmilienM holser: updating right now
21:01 holser and this https://review.openstack.org/#/c/124705/ caused a regression
21:02 spredzy EmilienM, +1 after mgagne comments are taken in account
21:02 spredzy holser, or he could create a tmp variable $libvirt_default_filepath = longstring and use it instead
21:02 spredzy what do you guys think ?
21:03 spredzy it will look shorter, but won't change the logic already in place
21:03 holser yeah, that’s good approach
21:05 EmilienM holser: damn
21:05 holser ?
21:05 EmilienM the regression
21:06 EmilienM spredzy: can you extend? because there is no need to use another variable, since we can consume an existing param
21:06 EmilienM I may be wrong though in the understanding
21:08 spredzy EmilienM, that was to address holser initial comment but it doesn't apply to rev 5
21:08 EmilienM spredzy: cool
21:08 spredzy rev == patchset, only to 4 so no need for it anymore
21:08 EmilienM mgagne: updated
21:13 * holser EmilienM: before => Package['libvirt'], I think it’s too strict
21:14 holser EmilienM: I would do before => Service[‘libvirt’]
21:14 EmilienM holser: it was not your suggestion?
21:15 holser Well, now I am checking and thinking that package will overwrite what pappet appled, though on next puppet run puppet will correct the settings
21:15 holser though, better to change to Service
21:16 holser Other then that I have no objections to your patch, just need to test it on debian, that’s it
21:16 EmilienM holser: I would rather use like we have in other files, notify => Service
21:17 holser yeah, notify will work also
21:17 holser everytime it changes Service will be retriggered
21:18 EmilienM holser: pushed..
21:19 holser it looks good to me
21:20 holser we just need to test on debian, and fix unit tests (I guess)
21:20 emccormick joined #puppet-openstack
21:20 EmilienM holser: I'm testing Debian.
21:20 holser oki
21:20 EmilienM holser: why do you need to fix unit tests?
21:20 holser I am just guessing
21:25 holser https://github.com/stackforge/puppet-nova/blob/master/manifests/params.pp#L33 it’s only for RedHat
21:25 holser it’s not specified for Debian:
21:25 holser sorry, it is
21:25 holser https://github.com/stackforge/puppet-nova/blob/master/manifests/params.pp#L82
21:31 EmilienM mgagne: the patch is ready for review fyi: https://review.openstack.org/#/c/128399/
21:32 holser Emilian, Debian section doesn’t have libvirt_service_name
21:33 holser read my last 5 comments
21:33 EmilienM holser: it has...
21:33 EmilienM https://github.com/stackforge/puppet-nova/blob/master/manifests/params.pp#L82
21:33 EmilienM it's for Debian
21:33 mgagne EmilienM: could you put back the File_line dependency line?
21:34 holser is that name correct?
21:34 mgagne EmilienM: requiring the other class is not enough as it could have containment issues.
21:34 * holser according to http://anonscm.debian.org/cgit/pkg-libvirt/libvirt.git/commit/debian/libvirt-bin.maintscript?id=28dd3618a9a2a8f769d1fa12c28647c4599d9e8e
21:34 holser it’s not
21:35 EmilienM mgagne: done
21:38 EmilienM mgagne: thx
21:38 holser mgagne, I am just curious why can’t we use require => [Class['nova::compute::libvirt'], Package['libvirt’]] way?
21:39 holser it’s more neaty … as for me
21:42 mgagne holser: http://puppetlabs.com/blog/class-containment-puppet
21:42 mgagne holser: but it looks like the issue is only with classes within classes
21:43 holser EmilienM, don’t you think it’s good to add something like if ($::operatingsystemmajrelease < XX) { $libvirt_service_name = ‘libvirt-bin’ } else {  $libvirt_service_name = ‘libvirtd’ }
21:43 holser to keep backward compatibility
21:44 mgagne holser: if a class (A) includes a class (B), if you require class (A), there is no guaranty that class (B) will be realized as expected.
21:44 holser that’s true for classes
21:44 * holser nods
21:47 mgagne holser: but it just goes to show that puppet-nova manages libvirt a bit too much imo
21:47 holser indeed :)
21:47 EmilienM mgagne: fyi, I'll be at openstack meetup the 27th of Oct
21:48 mgagne EmilienM: oh, that's true, you are in montreal now
21:48 EmilienM mgagne: lol
21:48 EmilienM mgagne: 1 month and a half !
21:48 mgagne EmilienM: they keep asking us to do a presentation but I unfortunately don't have much time these days =(
21:49 mgagne EmilienM: dmsimard will assist afaik
21:49 EmilienM mgagne: you should come
21:49 mgagne EmilienM: I will =)
21:50 mgagne EmilienM: unless rafael is there and ask for people to ask questions and others to answer them, I won't participate much I guess
21:51 mschmitt joined #puppet-openstack
21:51 EmilienM mgagne: he is in front of me, he says "of course I'll be there" :)
21:51 mgagne EmilienM: hehe
21:52 * mgagne currently battles flavor management in nova cells
22:23 ianw joined #puppet-openstack
23:03 openstackgerrit joined #puppet-openstack
23:22 rcallawa joined #puppet-openstack
23:23 rcallawa joined #puppet-openstack
23:34 rcallawa_ joined #puppet-openstack
23:41 michchap joined #puppet-openstack

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