Perl 6 - the future is here, just unevenly distributed

IRC log for #puppet-openstack, 2014-11-19

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

All times shown according to UTC.

Time Nick Message
00:12 topshare joined #puppet-openstack
00:28 topshare joined #puppet-openstack
00:37 topshare joined #puppet-openstack
00:41 mmaglana joined #puppet-openstack
00:42 topshare joined #puppet-openstack
01:17 EmilienM mfisch: thanks!
01:23 henditan joined #puppet-openstack
01:32 mmagr joined #puppet-openstack
01:37 openstackgerrit Merged stackforge/puppet-sahara: metadata: Bump major version (for juno)  https://review.openstack.org/130227
01:42 mmaglana joined #puppet-openstack
01:48 rmoe joined #puppet-openstack
01:52 guimaluf joined #puppet-openstack
01:52 xarses joined #puppet-openstack
02:11 xingchao joined #puppet-openstack
02:17 openstackgerrit Emilien Macchi proposed stackforge/puppet-keystone: Add keystone::resource::service_identity  https://review.openstack.org/134325
02:32 crinkle who has core on puppet-n1k-vsm? https://review.openstack.org/#/c/130226/ has been there for a while
02:39 EmilienM markvoelker: any idea^ ?
02:41 openstackgerrit Merged stackforge/puppet-neutron: Adding l3_ext network type support for neutron VMWare NSX plugin  https://review.openstack.org/128217
02:45 mmaglana joined #puppet-openstack
02:55 badiane_ka joined #puppet-openstack
03:31 topshare joined #puppet-openstack
03:37 henditan joined #puppet-openstack
03:38 badiane_ka joined #puppet-openstack
03:43 mmaglana joined #puppet-openstack
04:04 xingchao joined #puppet-openstack
04:19 topshare joined #puppet-openstack
04:29 mmaglana joined #puppet-openstack
04:49 vinsh joined #puppet-openstack
04:49 xingchao_ joined #puppet-openstack
05:07 gildub joined #puppet-openstack
05:11 topshare joined #puppet-openstack
05:26 sputnik13 joined #puppet-openstack
05:49 fedexo joined #puppet-openstack
05:50 dtalton joined #puppet-openstack
05:54 subscope joined #puppet-openstack
05:55 subscope joined #puppet-openstack
06:06 sputnik13 joined #puppet-openstack
06:17 rcallawa joined #puppet-openstack
06:19 rcallawa joined #puppet-openstack
07:00 tfz joined #puppet-openstack
07:27 rcallawa joined #puppet-openstack
08:05 arnaud_orange joined #puppet-openstack
08:18 qba73 joined #puppet-openstack
08:31 jawed joined #puppet-openstack
08:37 giorgiodinapoli joined #puppet-openstack
08:40 dgurtner joined #puppet-openstack
08:42 henditan joined #puppet-openstack
08:49 dgurtner joined #puppet-openstack
08:49 dgurtner joined #puppet-openstack
08:51 openstackgerrit Dan Bode proposed stackforge/puppet-keystone: Add native types for keystone paste configuration  https://review.openstack.org/135534
08:56 xingchao_ joined #puppet-openstack
09:01 jawed joined #puppet-openstack
09:05 openstackgerrit Dan Bode proposed stackforge/puppet-keystone: Add native types for keystone paste configuration  https://review.openstack.org/135534
09:10 openstackgerrit Dan Bode proposed stackforge/puppet-keystone: Add native types for keystone paste configuration  https://review.openstack.org/135539
09:29 henditan joined #puppet-openstack
09:33 stamak joined #puppet-openstack
09:36 henditan joined #puppet-openstack
09:44 henditan joined #puppet-openstack
09:47 igajsin left #puppet-openstack
09:51 igajsin joined #puppet-openstack
09:52 admin0 joined #puppet-openstack
10:17 rcallawa joined #puppet-openstack
10:46 jawed joined #puppet-openstack
11:04 rcallawa joined #puppet-openstack
11:56 xingchao joined #puppet-openstack
12:01 walkleyn joined #puppet-openstack
12:15 EmilienM good morning!
12:22 dmsimard EmilienM: good morning
12:22 dmsimard dgurtner: found anything ?
12:29 openstackgerrit Emilien Macchi proposed stackforge/puppet-keystone: Add keystone::resource::service_identity  https://review.openstack.org/134325
12:29 EmilienM spredzy: ^
12:29 spredzy EmilienM, acj
12:29 spredzy ack*
12:39 openstackgerrit Richard Megginson proposed stackforge/puppet-keystone: support the ldap user_enabled_invert parameter  https://review.openstack.org/129951
12:42 dgurtner dmsimard: yes. I actually have a working version locally :)
12:43 dmsimard dgurtner: SWEET !
12:43 dgurtner dmsimard: local CI run finally finished :D
12:43 dmsimard dgurtner: What was the issue ?
12:43 dgurtner so basically the ini_settings ensure => absent didn't remove anything anymore
12:44 dmsimard was pretty sure it was that
12:44 dgurtner not sure if we changed the version, or if it's because the scenario_node_terminus config is interfering
12:44 dgurtner anyways I replaced it with a shell/sed :)
12:45 dgurtner which is cleaner anyways, because the s_n_t was having side-effects
12:45 dmsimard cool
12:45 dmsimard so do all tests pass now or what ?
12:45 dmsimard or we still have some things to iron out ?
12:45 dgurtner i only run the ubuntu 12.02 ones, but i got all 250 tests to pass
12:46 dgurtner Finished in 62 minutes 58 seconds
12:46 dgurtner 250 examples, 0 failures
12:57 dmsimard dgurtner: nice. I saw some failures in the CentOS ones.. something about meta-link with the epel repo ?
12:58 francois1 joined #puppet-openstack
13:01 openstackgerrit David Gurtner proposed stackforge/puppet-ceph: Fixing the setup and teardown of SNT in mons and osds module  https://review.openstack.org/135590
13:03 dmsimard dgurtner: I'll run a CI test with that patch manually
13:03 dmsimard from the CI server
13:04 dgurtner dmsimard: keeping my fingers crossed :)
13:04 dmsimard dgurtner: I have faith in you :D
13:09 dmsimard Is dumpling still supported ? Starting to be old..
13:10 openstackgerrit David Gurtner proposed stackforge/puppet-ceph: Fixing the setup and teardown of SNT in mons and osds module  https://review.openstack.org/135590
13:10 dgurtner dmsimard: a cosmetic change ^
13:10 rcallawa joined #puppet-openstack
13:12 dgurtner dmsimard: afaik it's the last 3 version, so currently we should probably check emperor, firefly and giant only
13:14 dmsimard dgurtner: Asked #ceph-devel.. i'll add a commit on top of yours (once it's merged) to add giant
13:15 dgurtner dmsimard: also what's the deprecation status on 2.7? we might want to drop compatibility (as it's still blocking some tests from being enabled)
13:15 dmsimard I think we wanted to officially deprecate at the juno cycle and remove at kilo
13:16 dmsimard EmilienM: ^ ?
13:16 dmsimard EmilienM: Are you still gating against 2.7 ?
13:17 dmsimard dgurtner: Hey, just noticed. What's the deal with jenkins ? It's not running spec tests at all right now.. just lint and syntax
13:17 dmsimard I'll check that out.
13:18 dgurtner yup, was investigating as well
13:18 dgurtner saw some previous reviews only gating against 3.1 though:
13:18 dgurtner https://review.openstack.org/#/c/125248/
13:23 richm joined #puppet-openstack
13:25 dmsimard dgurtner: So far so good
13:25 dmsimard man, these tests take a while
13:25 dmsimard and they're not running on a bad machine either
13:26 dmsimard doh.. the VMs only have one core
13:31 dgurtner dmsimard: feel free to change the VM config in a future patch ;)
13:33 ibravo joined #puppet-openstack
13:34 dmsimard dgurtner: Yeah, the VMs are single core and 512MB RAM. I want to bump that but I don't see (quickly) how to do it in the nodeset file. Doc isn't exactly awesome.
13:34 dmsimard dgurtner: The VMs are actually swapping :P
13:35 dmsimard it looks like it's just taking the prefab and that's it
13:37 dmsimard dachary: Saw you reviewed the patchset. Around yet ?
13:38 dmsimard oooh.. nevermind, I figured it out
13:39 openstackgerrit Arnoud de Jonge proposed stackforge/puppet-nova: Configuration for spice now has a seperate section in nova.conf, moved config options from [DEFAULT] to [spice].  https://review.openstack.org/135600
13:40 EmilienM dmsimard: no
13:40 dmsimard EmilienM: No you don't gate against 2.7 anymore ?
13:41 EmilienM dmsimard: true
13:41 dmsimard EmilienM: Ok ty
13:45 dmsimard dgurtner: Found why spec tests aren't running anymore
13:45 dmsimard dgurtner: Working for a fix
13:47 dgurtner and we can drop 2.7 as well then :)
13:47 rcallawa joined #puppet-openstack
13:50 rcallawa joined #puppet-openstack
13:53 dmsimard https://review.openstack.org/#/c/135602/
13:55 dalgaaf__ joined #puppet-openstack
14:00 dachary dmsimard: yes
14:01 dmsimard dachary: Great news. dgurtner fixed the issue (you know that) and I'm running CI against the patch right now. So far so good.
14:01 dmsimard dachary: I found out that we didn't have spec tests run by jenkins in puppet-ceph anymore, submitted https://review.openstack.org/#/c/135602/ to fix it
14:02 dmsimard dachary: And I'll submit a patch that'll add another nodeset with bumped-up specifications (cores, ram) that'll be used by CI so tests complete faster. VMs only have 1 core and 512MB RAM (and they swap)
14:03 dachary that sounds great, puppet-ceph is back in business :-)
14:03 dmsimard dachary: Isn't that just awesome.
14:05 dmsimard dachary: Also. I'll add giant to the spec tests. Should we drop dumpling ?
14:05 dachary dumpling should stay because it's long term support
14:06 dmsimard Testing dumpling, emperor, firefly AND giant will start to take a long time.
14:06 dachary it's well maintained and won't cause trouble
14:06 dmsimard drop emperor then ?
14:06 dachary if we must emperor is the better candidate
14:06 dmsimard like, test against LTS releases + latest stable
14:06 dmsimard so, when hammer is out, drop giant and add hammer
14:06 dachary that's sensible
14:06 dmsimard When is support for dumpling ending ?
14:08 topshare joined #puppet-openstack
14:08 openstackgerrit Merged stackforge/puppet-ceph: Fixing the setup and teardown of SNT in mons and osds module  https://review.openstack.org/135590
14:09 topshare joined #puppet-openstack
14:10 markvoelker crinkle: EmilienM: cwchang or nagchund at cisco dot com can help with the puppet-n1k-vsm review.  I'll give them a shout.
14:10 dmsimard dachary, dgurtner: Should the default release offered by puppet-ceph be the latest stable ? So giant now ?
14:11 EmilienM markvoelker: thanks
14:12 topshare joined #puppet-openstack
14:13 admin0 joined #puppet-openstack
14:13 mmagr joined #puppet-openstack
14:14 richm joined #puppet-openstack
14:14 richm left #puppet-openstack
14:14 richm joined #puppet-openstack
14:15 dachary dmsimard: yes
14:16 dachary dmsimard: giant should be the default
14:16 dachary IMO ;-)
14:18 dgurtner giant as default +1
14:19 dmsimard k
14:28 topshare joined #puppet-openstack
14:34 openstackgerrit joined #puppet-openstack
14:34 derekh joined #puppet-openstack
14:35 morazi joined #puppet-openstack
14:50 openstackgerrit joined #puppet-openstack
14:50 pradk joined #puppet-openstack
15:09 ericpeterson joined #puppet-openstack
15:27 xingchao joined #puppet-openstack
15:33 dkaigarodsev joined #puppet-openstack
15:39 ericpeterson joined #puppet-openstack
15:56 mdorman joined #puppet-openstack
15:57 aarefiev joined #puppet-openstack
16:05 mgagne xarses: ping
16:13 xarses mgagne: hi
16:13 mgagne xarses: question related to rabbitmq HA
16:13 mgagne xarses: I saw https://bugs.launchpad.net/fuel/+bug/1276770
16:14 mgagne xarses: what's your opinion about rabbit_hosts with N rabbitmq vs HAProxy to load-balance rabbitmq?
16:15 mgagne xarses: why would someone use haproxy instead of rabbitmq clustering?
16:15 xarses we have both, we prefer the service to support rabbit_hosts though
16:16 xarses we have the cluster for service that don't support it, although I think murano has that fixed now
16:18 mgagne xarses: ok so it's mainly for services without rabbit_hosts support
16:19 xarses we used to use it for all services and found it an issue in ha testing
16:19 mgagne xarses: good to know I guess =)
16:20 EmilienM mgagne: we have both also
16:21 xarses IIRC one of the other installers juju / packstack / suse does only haproxy
16:21 xarses it was in the state of ha talk
16:21 EmilienM xarses: mgagne: I don't know if it's the right way, but we define the RabbitMQ VIP in rabbit_hosts parameter
16:21 EmilienM and enable HA queues
16:22 mgagne EmilienM: what I read is that if a rabbitmq node fails and come back, it might not have time to resync all queues back
16:22 mdorman +1 on haproxy+rmq clustering.   i think we moved to haproxy to help more evenly distribute the connections across all rmq nodes
16:23 EmilienM mdorman: do you use the VIP in rabbitmq_hosts?
16:23 mdorman not sure if we put it in rabbitmq_hosts or rabbitmq_host, let me look
16:23 xarses ya, thats what we had, and it broke down in ha testing. however it might be better now that we have implemented heartbeat
16:24 mgagne I guess my troubles are around the consistency or state of the rabbitmq node when it goes back in the load-balancing
16:24 mdorman yeah we do put it on rabbitmq_hosts
16:24 EmilienM mdorman: one single VIP + port in this param, and it works well?
16:25 mdorman and i have seen weird behaviors when rmq node failures/rejoins happen.  never taken the time to really dig into it.  but definitely there can be something odd there
16:25 mdorman EmilienM:  yeah.  for the most part i think it works fine.
16:25 mdorman but we also have not done extensive rmq failover testing, so i can’t say that with 100% confidence
16:25 xarses mdorman: mgagne, yep, thats what caused us to move away from using the vip, but with heartbeat it might not be a problem now
16:26 mgagne xarses: cool thanks for the info
16:26 xarses we also put the vip into a net namespace
16:26 mgagne xarses: what about firewall? do you run rabbitmq through a firewall? what about those friendly firewalls that silently drops your connection/session. How do you deal with it on the client side?
16:26 EmilienM xarses: so you use rabbitmq_host ?
16:27 xarses EmilienM: rabbit_hosts with each rabbit node in the list
16:27 mgagne EmilienM: He uses rabbitmq_hosts wherever possible and rabbit_host for project without support for it I guess
16:27 EmilienM xarses: ok, and you gave up the haproxy thing, right?
16:27 mdorman we had the silent session drop issue, but via a HW load balancer that was timing out idle connections after a couple minutes.  we just increased the idle timeout on the LB to something really long, because we couldn’t get the clients to reliably recognize their connection was gone
16:28 xarses EmilienM: we stopped using it thinking it was part of the problem, but it is likely fixed with the use of rabbitmq keepalive
16:28 mgagne mdorman: what happens when it's gone? have you dealt with it?
16:28 xarses havent retested yet
16:28 xarses mgagne: other than the linux tables, no
16:28 EmilienM mgagne: xarses: ok, thx for feedback.
16:28 * mgagne thinks it would have been a great discussion on #openstack-operator
16:28 EmilienM xarses: what is this rabbitmq keepalive?
16:29 xarses https://etherpad.openstack.org/p/kilo-oslo-oslo.messaging
16:29 EmilienM ah true
16:29 EmilienM xarses: thx :)
16:29 xarses bottom
16:29 mgagne mdorman: we use tcp keepalive at the kernel level to avoid session dropping: net.ipv4.tcp_keepalive_time, net.ipv4.tcp_keepalive_intvl, net.ipv4.tcp_keepalive_probes
16:31 xarses mgagne: ya, IIRC that still resulted in some issues when the nodes went away
16:31 mgagne mdorman: but still didn't figure out how to deal with firewall that forcefully terminates/drops your session during maintenance. Clients are still not able to detect session dropping.
16:31 xarses 1 of the 3 connections in nova-compute would still hang
16:31 xarses mgagne: rabbitmq keepalive =)
16:31 mgagne xarses: yep, saw that one recently
16:32 mgagne xarses: it does fix it?
16:32 mgagne xarses: is it new in oslo.messaging?
16:33 xarses it's not merged yet, I expect it to be in kilo. We've been using it with havana / icehouse
16:33 xarses and juno
16:33 mgagne xarses: cool, will look at it
16:33 mgagne xarses: also found https://bugs.launchpad.net/nova/+bug/856764
16:33 mdorman mgagne:  yeah, we tried keepalive in kernel,too, but weren’t successful.  like i said we just dealt with it by increasing the timeout time on the LB to like an hour or a day or something.  but, yeah, had the exact same problem with clients not reconnecting.   i’m hopeful this keepalive thing in oslo wil solve that
16:33 xarses also, here is a list of everything we've done in the name of ha https://etherpad.openstack.org/p/fuel-ha-fixes-catalogue
16:34 vinsh joined #puppet-openstack
16:34 mgagne xarses: thanks a lot, will look at it
16:40 aarefiev joined #puppet-openstack
16:41 rcallawa joined #puppet-openstack
16:52 mschmitt joined #puppet-openstack
16:56 mgagne xarses: I don't have all the details but it looks like TCP keepalive works (for us) but nova-compute service is still see as down by nova service-list and we have to restart nova-compute to be up again. A coworker allegedly said that tcp connection to rabbitmq was fine, connection to queue was fine from nova-compute.
16:56 mgagne xarses: god knows what happened
16:57 xarses mgagne: classic symptom of what we where working on fixing
16:57 mgagne xarses: which is?
16:58 xarses the connection(s) are to channels that are closed, and wont be reaped untill the tcp session keepalive (2h) kicks in and closes it from the kernel
16:58 mgagne xarses: we reduced it to like 30s
16:58 mgagne xarses: and it works for most services, we didn't have to restart neutron agent on compute nodes
16:58 xarses and then you see random offlines of the nodes too, sometimes db gone away
16:59 xarses ya, neutron agent was weird
16:59 xarses compute was the most impacted
16:59 xarses the implementation of rabbitmq keepalives ensures that the connection won't stay connected to a session that was closed because rabbit won't respond to the heatbeat if the channel is closed
17:01 mgagne xarses: oh right. I'm not a rabbitmq/amqp expert but I know some of those words =)
17:01 mgagne xarses: so connection is up but channel is down?
17:02 sputnik13 joined #puppet-openstack
17:02 xarses yes
17:02 mgagne xarses: thanks, will dig into that direction
17:05 rcallawa joined #puppet-openstack
17:07 xarses back in a bit
17:09 mmaglana joined #puppet-openstack
17:10 openstackgerrit Merged stackforge/puppet-nova: Service Validation for Nova-API  https://review.openstack.org/126458
17:16 dgurtner joined #puppet-openstack
17:16 dgurtner joined #puppet-openstack
17:22 stupidnic I have a question regarding the openstack::role::storage
17:22 stupidnic What I want to know is do I have to have cinder enabled on the controller?
17:22 stupidnic I am using Ceph, so I am not 100% certain where that role should live
17:24 openstackgerrit David Moreau Simard proposed stackforge/puppet-ceph: Bump default version to giant, test against giant and drop emperor  https://review.openstack.org/135668
17:25 JayJ joined #puppet-openstack
17:25 mmaglana joined #puppet-openstack
17:26 aimon joined #puppet-openstack
17:29 xarses joined #puppet-openstack
17:35 tfz joined #puppet-openstack
17:39 RichardRaseley joined #puppet-openstack
17:41 dmsimard dachary: testing the impact of bumping the CI VMs specifications
17:41 dtalton joined #puppet-openstack
17:41 dmsimard dachary: baseline is with the current VMs in the nodeset for two ubuntu was 53 minutes 47 seconds
17:45 dmsimard dachary: Are CI tests queued ? Or can two run simultaneously ? Like, if there is two commits in two different reviews
17:49 dachary dmsimard: gerritexec queues tests, one after the other
17:50 dachary it is not smart at all ;-)
17:51 dachary it would make a lot of sense to run two gerritexec
17:51 dachary one for ubuntu and one for centos
17:51 sputnik13 joined #puppet-openstack
17:51 dachary and launching the tests to filter only the relevant distribution
17:52 dachary the same could be done to run one gerritexec per operating system / ceph version
17:52 mmaglana joined #puppet-openstack
17:52 dachary and we already have the flags in place for this
18:00 mmagr joined #puppet-openstack
18:03 dmsimard dachary: Yeah, I was thinking about running centos and ubuntu at the same time... but not twice.. like, 4 VMs simultaneously. Not four if there happens to be another commit that comes in while the tests are already running
18:05 dmsimard dachary: I want to explore also if we can run the VMs in RAM tmpfs
18:06 vinsh joined #puppet-openstack
18:06 dachary that's probably a good idea :-)
18:12 rmoe joined #puppet-openstack
18:17 dmsimard dachary: oh, ram tmpfs is an easy one.. not going to bother trying to do it from within vagrant or virtualbox.. I'll simply remount /home/ceph/Virtualbox VMs into a tmpfs
18:17 dmsimard problem solved
18:18 aimon joined #puppet-openstack
18:21 aimon joined #puppet-openstack
18:21 dachary :-D
18:23 sputnik13 joined #puppet-openstack
18:26 EmilienM mgagne: thanks for review, what did you think about the new design?
18:26 EmilienM (service validation)
18:26 mgagne EmilienM: looks much lightweight than the first implementation
18:27 mgagne EmilienM: original design was too heavy to implement without proper planning and design
18:27 EmilienM mgagne: can you see an OpenStack deployment without PuppetDB ?
18:27 EmilienM (random question)
18:27 mgagne EmilienM: for light deployment: yes. unless puppetlabs deprecates installation without puppetdb
18:28 EmilienM mgagne: for large scale I meant
18:28 EmilienM mgagne: I don't see how you can deploy Swift without it for example
18:28 mgagne EmilienM: storeconfigs? :D
18:29 EmilienM mgagne: indeed ! thanks I'll do that (joke)
18:29 nibalizer lol
18:29 mgagne EmilienM: tbh I don't what else someone could use instead of puppetdb
18:29 nibalizer i haven't seen anyone publicly talk aobut scaling puppetdb
18:29 mgagne nibalizer: what do you think?
18:29 mgagne nibalizer: fun to hear
18:29 nibalizer i think maybe one puppetdb server per avail zone?
18:30 nibalizer in theory you can make multiple puppetdb servers behind haproxy
18:30 nibalizer and then postgres behind that
18:30 nibalizer but as i understand it, there isn't something like galera for postgres
18:31 nibalizer i wonder if one puppetdb+ one postgres per openstack avail zone?
18:32 mgagne nibalizer: people using puppetmaster in production, at what scale are they?
18:33 nibalizer i think all the really high scale folks run puppet apply
18:33 * nibalizer not a lot of experience
18:34 mgagne nibalizer: if puppetdb performance falls off the table after 1000 nodes, we have a problem I guess
18:34 nibalizer no it can go higher than that i think
18:34 nibalizer its kindof a beast
18:34 nibalizer but i think the HA part hasn't been figured out yet
18:34 nibalizer along with horizontally scaling
18:34 mgagne nibalizer: you will agree that HA != scalability right? =)
18:34 mgagne right
18:35 nibalizer basically the way i understand it is the puppetdb service isn't a problem, its pretty leight weight
18:35 nibalizer the real kicker is postgres
18:35 mgagne so people don't care about HA and/or haven't reached yet scalability issues with puppetdb
18:35 nibalizer or if they have they had plabs secret agents make do and its not super public what they did
18:36 mgagne nibalizer: so it's back to pure postgres scalability issues/tools, not related to puppetdb
18:36 crinkle grim_radical in #puppet-dev is a good person to talk to about puppetdb btw
18:36 mgagne crinkle: #puppet-openstack is the new #openstack-operators channel =)
18:37 nibalizer haha
18:37 nibalizer #puppet-openstack is my new favorite channel actually
18:37 nibalizer its legit
18:37 dmsimard nibalizer: Lots of (big) operators deploy with puppet so there's kind of an overlap :p
18:38 dmsimard #openstack-operators is quieter than the mailing list
18:47 mschmitt joined #puppet-openstack
18:47 EmilienM dmsimard: people is shy :)
18:49 tchaypo does #openstack-operators have answers about the correct gel to transfer images onto wood?
18:53 EmilienM mgagne: I was wondering, does this patch deserves an unit test? https://review.openstack.org/#/c/128752
18:54 mgagne EmilienM: is the provider already unit tested?
18:55 EmilienM mgagne: very basic tests
18:56 spredzy|mtg mgagne, I'd say not enough if he had to add a specific use case
18:56 EmilienM spredzy|mtg: it fix an interesting case, I would say we may want an unit test
18:57 mgagne EmilienM: 2 things: 1) if there is no unit tests yet, it's not fair to ask someone to bootstrap a unit test for a provider. coding for providers is hard, testing is even harder. 2) operators aren't developers. We often talked about helping contributors to code tests for them if they weren't experimented with unit testing. More often that not, after asking for unit tests, change goes stalled and never get merged.
18:57 EmilienM mgagne: that sounds fair, that why I was asking you. Thanks
18:58 EmilienM so I keep my +2
18:58 mgagne EmilienM: off the top of my head, I wouldn't know how to unit test this case. seem unfair to ask someone else to do it
18:58 EmilienM mgagne: +1
18:58 mgagne EmilienM: but we should ask others if they can contribute tests though
18:59 spredzy mgagne, ack the point for future reviews
19:10 mgagne EmilienM: I have a comment about service_identity. see inline comments: https://review.openstack.org/#/c/134325/
19:12 EmilienM mgagne: looking now
19:13 EmilienM mgagne: I kind of agree with you for manifests/roles/admin.pp file, maybe I did that first to demonstrate my code is actually working
19:13 EmilienM mgagne: if I revert manifests/roles/admin.pp as it was before, is it fine?
19:14 mschmitt1 joined #puppet-openstack
19:14 mgagne EmilienM: yep
19:17 openstackgerrit Emilien Macchi proposed stackforge/puppet-keystone: Add keystone::resource::service_identity  https://review.openstack.org/134325
19:18 mdorman re: earlier discussion around unit tests on https://review.openstack.org/#/c/128752   definitely agree with the sentiment discussed above.  i’m happy to work on tests, but i’m not great at rspec.  and on this one i wouldn’t really know where to start with it, either
19:18 mdorman but can discuss the use case/details more if people want
19:19 EmilienM mdorman: no problem
19:31 nibalizer i just need a train ride
19:31 nibalizer then tests for everything
19:31 nibalizer EVERYTHING
19:31 nibalizer muhhahahhahahaahahahhaa
19:33 dmsimard dachary: Oh, we already use a fork of rspec-system, that's right ..
19:33 nibalizer dmsimard: really? ouch
19:33 nibalizer beaker 2.0 was mentioned to day on the openstack lists
19:33 nibalizer dropping ruby1.8 support
19:34 dmsimard nibalizer: yeah, puppet-ceph has been using rspec-system since it's inception. Beaker was not yet out.
19:34 dmsimard nibalizer: We need to migrate someday ..
19:34 dmsimard that day is not today
19:34 dmsimard :D
19:34 nibalizer ya
19:34 nibalizer well
19:34 dachary dmsimard: right. We're all in agreement that beaker is the way to go but ... it's not compatible.
19:34 nibalizer i almost have beaker-rspec in the openstack ci infra
19:34 nibalizer so that may motivate you
19:35 dachary nibalizer: nice !
19:35 dmsimard dachary: I want to add prefabs to rspec-system, it's quick and easy so we can use an up-to-date 12.04 and test against 14.04 as well..
19:36 dmsimard dachary: https://github.com/puppetlabs/rspec-system/blob/master/resources/prefabs.yml
19:36 dachary ok
19:36 dmsimard dachary: https://vagrantcloud.com/puppetlabs (https://vagrantcloud.com/puppetlabs/boxes/ubuntu-12.04-64-puppet/versions and https://vagrantcloud.com/puppetlabs/boxes/ubuntu-14.04-64-puppet)
19:38 dmsimard dachary: Think you can find some time to do it ? I don't have a gitorious account :P
19:38 dachary ahaha
19:39 dachary I would not know where to start
19:50 dmsimard dachary: Why wasn't beaker compatible ? It was the disk attach thing ?
19:55 EmilienM mgagne: sorry but I still don't understand what's wrong with https://review.openstack.org/#/c/134325/7..8/manifests/roles/admin.pp,cm
19:57 nibalizer so if you look at the failing test here
19:57 nibalizer https://review.openstack.org/#/c/126086/
19:57 nibalizer you can see beaker runs
19:57 nibalizer then puppet fails
19:57 nibalizer so thats a big step!
19:58 openstackgerrit Merged stackforge/puppet-keystone: keystone_user_role fix for when user not granted roles yet  https://review.openstack.org/128752
19:59 mgagne EmilienM: ?
19:59 EmilienM mgagne: I just use some code I did to refactorize, I just need to know what's wrong with that
20:00 EmilienM it does the same thing, right?
20:00 mgagne EmilienM: admin isn't a service identity
20:00 mgagne EmilienM: we shouldn't use service_identity to manage the "admin' user.
20:00 mgagne EmilienM: it's not semantically correct
20:01 EmilienM mgagne: I just create a user and the role
20:01 mgagne EmilienM: is "admin" a service account?
20:01 EmilienM mgagne: but it works, isn't ? with less code
20:02 * EmilienM just tries to understand
20:02 mgagne EmilienM: you shouldn't use a resource designed to manage *service* identity to manage a non-service user
20:03 EmilienM mgagne: my resource is flexible enough to do it
20:04 mgagne EmilienM: if you think about it in a short-minded way, yes.
20:04 dmsimard dachary: I have to go for today but this sucks.. I bumped the specs of the VMs and put the VMs in a tmpfs and still hitting around 60 minutes for the ~250 two ubuntu tests.. Way too long.
20:04 dmsimard dachary: It's running on arguably good hardware too..
20:04 mgagne EmilienM: but don't you see that it's not the right tool/thing to use?
20:04 EmilienM mgagne: I'm ok to not using it, I just wanted to understand why not
20:05 mgagne EmilienM: a good programmer will tell you that it semantically makes no sense to use service_identity (even though it works)
20:05 mgagne EmilienM: I don't know how else I should reword it
20:06 EmilienM mgagne: fair enough
20:44 stamak joined #puppet-openstack
20:46 EmilienM crinkle: do you think https://review.openstack.org/#/c/134844 could land soon?
20:47 EmilienM crinkle: because richm is working on a bug in the same file in https://review.openstack.org/#/c/133601
21:16 JayJ joined #puppet-openstack
21:27 ibravo joined #puppet-openstack
21:45 crinkle EmilienM: we should land richm's first and i'll rebase against it
21:45 crinkle I don't have a good way to test ldap stuff so I've been avoiding reviewing those
21:46 EmilienM crinkle: ack
21:51 richm crinkle: thanks - I sympathize with the problems setting up keystone + ldap for testing . . .
21:53 richm then, as soon as the openstackclient stuff is ready, I can change the keystone code to use openstackclient instead of using the v3 api directly
22:03 mgagne anyone using puppetlabs-mongodb? if so, how are you overriding the default repo with your own private one?
22:07 EmilienM mgagne: no
22:08 EmilienM mgagne: this is how we do: https://github.com/enovance/puppet-openstack-cloud/blob/master/manifests/database/nosql.pp#L54-L71
22:08 rcallawa_ joined #puppet-openstack
22:08 mgagne EmilienM: hehe, was digging the code atm
22:09 EmilienM mgagne: tbh, it's maybe not the right way
22:09 EmilienM but it works quite well
22:09 mgagne EmilienM: indeed
22:09 EmilienM \o/
22:09 mgagne EmilienM: I'm not sure how the module handles class containment though
22:09 EmilienM mgagne: any feedback is welcome
22:09 mgagne EmilienM: looks like the repo *could* be installed after the package resource
22:10 openstackgerrit joined #puppet-openstack
22:11 EmilienM mgagne: you mean, a bug in puppetlabs-mongodb, right?
22:11 mgagne EmilienM: maybe
22:12 mgagne EmilienM: maybe I'm lucky because I have a global dependencies on Class['Apt'] -> Package <| |>
22:12 mgagne EmilienM: or recent puppet versions are executed top-bottom
22:12 EmilienM mgagne: there is something:
22:12 EmilienM Apt::Source['downloads-distro.mongodb.org']->Package<|tag == 'mongodb'|>
22:13 mgagne hooooo where?
22:13 mgagne good catch on the tag
22:13 mgagne EmilienM: https://github.com/puppetlabs/puppetlabs-mongodb/blob/master/manifests/repo/apt.pp
22:13 openstackgerrit Merged stackforge/puppet-keystone: Add keystone::resource::service_identity  https://review.openstack.org/134325
22:13 mgagne EmilienM: I should pay more attention =)
22:14 EmilienM mgagne: you should :P
22:14 mgagne EmilienM: my solution: Apt::Source <| title == 'downloads-distro.mongodb.org' |>  { ensure => absent }
22:14 mgagne EmilienM: apt::source { 'mongodb': ...} Apt::Source['mongodb'] -> Package<| tag == 'mongodb' |>
22:15 mgagne EmilienM: or Apt::Source <| title == 'downloads-distro.mongodb.org' |>  { location => "your own" }
22:18 EmilienM mgagne: smart
22:32 dachary dmsimard_away: beaker uses a different set of conventions, it only looks similar
22:57 rushiagr_away joined #puppet-openstack
23:14 EmilienM dachary: reading your comment on https://review.openstack.org/#/c/135729/
23:14 EmilienM dachary: it is of course in the roadmap
23:15 vinsh joined #puppet-openstack
23:18 openstackgerrit Emilien Macchi proposed stackforge/puppet-nova: Refactorise Keystone resources management  https://review.openstack.org/134493
23:19 EmilienM michchap: can you review again https://review.openstack.org/#/c/134493/ ? I'm not sure about your -1
23:25 ibravo joined #puppet-openstack
23:28 gildub joined #puppet-openstack
23:39 rcallawa joined #puppet-openstack

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