Perl 6 - the future is here, just unevenly distributed

IRC log for #puppet-openstack, 2013-11-13

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

All times shown according to UTC.

Time Nick Message
00:32 rongze joined #puppet-openstack
00:56 tnoor joined #puppet-openstack
01:08 tnoor1 joined #puppet-openstack
01:10 tnoor1 joined #puppet-openstack
01:50 rongze joined #puppet-openstack
01:54 xarses joined #puppet-openstack
02:23 rongze joined #puppet-openstack
02:26 xingchao joined #puppet-openstack
02:29 rongze joined #puppet-openstack
03:33 e1mer joined #puppet-openstack
03:35 rongze joined #puppet-openstack
03:38 rongze joined #puppet-openstack
04:52 rongze joined #puppet-openstack
05:53 rongze joined #puppet-openstack
06:47 rongze joined #puppet-openstack
07:29 morazi joined #puppet-openstack
08:05 mmagr joined #puppet-openstack
08:21 Krast joined #puppet-openstack
08:26 Krast left #puppet-openstack
08:32 Krast joined #puppet-openstack
09:12 dvanders joined #puppet-openstack
09:12 dvanders left #puppet-openstack
09:28 derekh joined #puppet-openstack
09:30 EmilienM good morning here
09:30 EmilienM i would need review on patches, is somebody available ?
10:07 Krast joined #puppet-openstack
10:34 marun joined #puppet-openstack
10:34 dvanders joined #puppet-openstack
10:35 openstackgerrit A change was merged to stackforge/puppet-ceilometer: Alarm: record history support  https://review.openstack.org/55002
10:36 EmilienM fc__: i have other patchs to review if you could
10:37 fc__ EmilienM: will do
10:37 EmilienM fc__: thank you
10:48 rongze joined #puppet-openstack
10:49 openstackgerrit A change was merged to stackforge/puppet-ceilometer: changed rabbit_virtualhost(wrong) to rabbit_virtual_host  https://review.openstack.org/55366
11:58 e1mer joined #puppet-openstack
12:20 rongze joined #puppet-openstack
13:03 openstackgerrit Flavio Percoco proposed a change to stackforge/puppet-glance: Replace qpid_host with qpid_hostname  https://review.openstack.org/56208
13:08 dprince joined #puppet-openstack
13:10 dmsimard joined #puppet-openstack
13:13 morazi joined #puppet-openstack
14:01 dmsimard1 joined #puppet-openstack
14:05 dmsimard joined #puppet-openstack
14:13 dvanders_ joined #puppet-openstack
14:15 mjblack joined #puppet-openstack
14:20 dvanders_ joined #puppet-openstack
14:23 morazi joined #puppet-openstack
14:44 rongze joined #puppet-openstack
14:52 rongze joined #puppet-openstack
14:58 e1mer joined #puppet-openstack
15:01 rmoe joined #puppet-openstack
15:09 mjblack joined #puppet-openstack
15:10 badiane_ka joined #puppet-openstack
15:18 dmsimard dachary: Blocked in my current projects, I'll see what I can do about ceph::conf
15:18 dachary dmsimard: ok :-)
15:19 dmsimard dachary: I'm thinking it might not be entirely justified
15:19 dmsimard dachary: Seeing as there's now a provider
15:20 dmsimard dachary: ceph_config
15:20 dachary indeed
15:20 dmsimard dachary: maybe for the global configuration, but otherwise ceph::mon, ceph::osd should take care of their own config ?
15:21 dmsimard xarses: What do you think ?
15:21 dachary the simpler the better
15:21 dmsimard Okay, I'll ask the question the other way around
15:22 dmsimard Is ceph::conf necessary (if it is, it would be for the global config only?) or could that perhaps go in ceph::init along with the package installation
15:22 dmsimard Should probably go in init.
15:23 dmsimard Since init is meant to be the common base, dependencies across the cluster
15:26 dmsimard dachary, xarses: Need you guys' input to make sure we're going in the same direction ^
15:29 dachary dmsimard: I agree with you : ceph::conf does not seem to be necessary anymore
15:30 dachary and I'm still unconvinced that ceph::ini is necessary ;-)
15:30 dachary ceph::init
15:31 mgagne joined #puppet-openstack
15:37 dmsimard long story short, ceph::init is a place where all the common code/dependency/base would go so you don't write the same code in different classes
15:37 dmsimard to improve maintainability and usability
15:37 mjblack joined #puppet-openstack
15:39 dmsimard e.g, you don't put ensure_packages('ceph') in ceph::osd, you put require ceph::init
15:39 otherwiseguy joined #puppet-openstack
15:40 dmsimard that's my vision of how things should go, at least
15:42 xarses dmsimard: i would expect global things like fsid, the two monitor vars whose names escape me, public_network, cluster_network, and auth_supported to be maintained by something like ceph::config
15:43 xarses hmm, didn't read far enough, you want to put that in ceph::init?
15:43 dmsimard I don't know, questioning wether or not a ceph::config would be relevant (or confusing) given we have ceph_config
15:44 xarses anyway, I would expect some DRY, in that the global stuff that you would set the same for both OSD and MON, should be in a common place
15:44 dmsimard puppet-openstack modules don't seem to be using a ::conf class
15:45 xarses such as the vars i stated above, that will be in each the config for mon, osd, radosgw...
15:46 dmsimard I'll build on what I submitted last week with ceph::params, see what I can suggest and we can discuss further (it won't require much work)
15:53 xarses dmsimard: fyi https://github.com/xarses/fuel/blob/b6fc802​f7f09fe80f760afed132e6a60db4ec1a8/deploymen​t/puppet/ceph/manifests/radosgw.pp#L54-60
15:53 dmsimard xarses: ok - I rarely use upstart, even though I know it's there - old Debian habits of mine
15:54 dmsimard xarses: So I didn't notice
15:54 xarses I just added it
15:54 dmsimard xarses: In my recipes, I feed the patched radosgw init script I pushed upstream since it's not in the packages yet
15:55 xarses but aparentlpy puppet cant handle that the upstart script wan'ts id=radosgw.gateway, because if the service is not on, it returns STDERR
15:55 xarses ya, we'll probably roll that patch soon too
15:56 xarses aparently puppet freaks out if STDERR is returned besides 0, and non zero ret value
15:56 xarses anyways just a note if you have manifests, you'll probably want to do the same
16:07 francois1 joined #puppet-openstack
16:08 hogepodge joined #puppet-openstack
16:36 xarses joined #puppet-openstack
16:36 e1mer joined #puppet-openstack
16:39 rongze joined #puppet-openstack
16:48 rongze joined #puppet-openstack
16:53 hogepodge joined #puppet-openstack
16:54 dmsimard dachary, xarses: Were we using underscores or spaces as separators in the config ?
16:55 xarses dmsimard: currently, ceph-deploy uses underscores so we should use what ever it does, if we want to use spaces, ceph-deploy should be made to match
16:56 xarses I actually had to fix it so it didn't use both
16:56 dmsimard ok
16:56 dmsimard i'll set underscores.
16:57 mjblack joined #puppet-openstack
16:58 hogepodge joined #puppet-openstack
17:00 mjblack joined #puppet-openstack
17:03 rmoe joined #puppet-openstack
17:18 dmsimard Think I got something that makes sense, I'll try it out.
17:22 angdraug joined #puppet-openstack
17:22 dmsimard writing the spec tests for this will be painful :(
17:24 hogepodge joined #puppet-openstack
17:26 xarses I know the feeling
17:43 dmsimard I present to you the first ceph.conf built with ceph_config and ceph::init evar: http://paste.openstack.org/show/52468/
17:43 hogepodge joined #puppet-openstack
17:45 dmsimard So much for ordering my stuff right, it places the config values in an apparent random order within the namespace
17:47 dmsimard If there is one thing I dislike about puppet it's the randomness in execution if you don't hardcode things to happen in a specific order :(
17:49 dachary dmsimard: https://wiki.openstack.org/wiki/Pu​ppet-openstack/ceph-blueprint#conf says "space" but I'm ok with _ ;-) Would you like me to update ?
17:50 dmsimard dachary: I would go with the ceph-deploy choices
17:50 dachary sure :-)
17:50 dmsimard dachary: So we stay in line with the "standard"
17:52 dachary updated
17:56 dmsimard Looks like what I got is working well, now on to the spec tests :(
18:01 mmagr joined #puppet-openstack
18:01 rongze joined #puppet-openstack
18:28 blentz joined #puppet-openstack
18:43 xarses joined #puppet-openstack
19:03 hogepodge ping bodepd
19:15 bauzas joined #puppet-openstack
19:25 dmsimard dachary, xarses: Committing my work, go easy on me - this is something to at least generate discussion, at best make it through :)
19:26 xarses so we should just blindly merge it ;D
19:26 dmsimard nononononono
19:27 dachary :-)
19:27 dachary xarses: +1
19:27 dachary ahaha
19:27 dachary bbl ( diner time here )
19:27 dmsimard I like what I did though
19:27 xarses alternately, we can just wait until revision 17 before we review it
19:30 xarses dmsimard: I don't see a review
19:31 dmsimard sending it through now, was writing commits
19:31 dmsimard the bot will pick it up :)
19:31 openstackgerrit David Moreau Simard proposed a change to stackforge/puppet-ceph: Introduction of ceph::init, ceph::params  https://review.openstack.org/56270
19:44 xarses Nice work overall
19:45 xarses dmsimard: I'd prefer to see the osd/mon/mds sections in their own claseses, even if the only thing you implement is the ceph_config parts
19:59 dachary xarses: +1
20:00 tnoor1 joined #puppet-openstack
20:01 dachary dmsimard: ceph::init looks very much like what ceph::conf could be, don't you think ?
20:02 dmsimard dachary: yes, and it installs ceph
20:02 dmsimard dachary: In the future it could hold more things :)
20:02 mgagne dachary: ::init is usually used to do global/common configuration and a base install of a module.
20:02 dmsimard xarses: You mean like ceph::init::mon ?
20:03 mgagne dachary: nothing prevents you from splitting it in multiple parts though
20:03 xarses dmsimard: i would expect the mon parts to only be needed if i was running ceph::mon
20:04 xarses dmsimard: unless im mistaken, and every osd/mon/mds/radosgw/client needs to have the same settings
20:04 xarses in which case, it should be in ceph::init
20:04 dmsimard xarses: It's my understanding that each server should ultimately hold a copy of the same ceph.conf
20:05 dmsimard xarses: But you're probably more knowledgeable than I am
20:06 xarses dmsimard: as far as i can tell that used to be the the nomenclature back when ceph.conf requireda complete list of monitors and osd's
20:07 xarses it seems to not be necessary for items outside of the global section
20:07 xarses dont get me wrong, all osd nodes should have the same [osd] section, but a mds node shouldn't need a [osd] section
20:08 xarses unless we are adding [osd.0]...[osd.X] sections
20:08 dmsimard I guess you're right
20:08 dmsimard I was thinking the "global" osd, mon and mds sections needed to be there all the time
20:09 dachary I'm not sure I see the value of ceph::init
20:10 mgagne dachary: ::init is usually the point of entry of any puppet modules
20:11 dachary mgagne: is it the case for openstack modules ?
20:12 mgagne dachary: no, it's the case of any puppet modules
20:12 dachary I'm confused are we talking about init.pp or XXX::init ?
20:12 mgagne dachary: however it doesn't mean it should be used to install/configure 56 things
20:12 mgagne they are the same
20:13 xarses dachary: MODULE/mantifests/init.pp is MODULE:init
20:13 xarses erm
20:13 xarses dachary: MODULE/mantifests/init.pp is MODULE
20:13 mgagne ^
20:14 dmsimard That actually reminds me that I messed up and xarses missed it :D
20:14 dmsimard I named the class ceph::init instead of ceph
20:14 mgagne MODULE/manifests/init.pp = MODULE::init = include MODULE = class { 'MODULE' }
20:14 dmsimard ninja fixing it
20:15 dachary that was the root of my confusion :-)
20:15 xarses mgagne: thanks
20:16 dmsimard xarses: If they are undef, the config provider won't set them - that is what you're saying ?
20:17 dachary mgagne: forgive my ignorance but, does that mean you can write ceph::init { } instead of class { 'ceph': } and it's the same ?
20:17 mgagne dachary: classes can be included in 2 ways: class { 'ceph:' } or include ceph
20:17 xarses dmsimard: if the value for a setting is undef the ini_file provider wont write the line into the config file at all
20:18 dmsimard xarses: That's good to know.
20:18 xarses dmsimard: it wasted about a day of my life
20:18 dmsimard xarses: LOL
20:18 mgagne dachary: the first form doesn't allow you to define the class twice, especially if they don't have the same parameter
20:19 dachary mgagne: got it :-)
20:19 dmsimard xarses: Considering your comment to split sections outside of init, do you think I should add ceph::osd, ceph::mds and ceph::mon with only the configuration for the time being ?
20:19 mgagne dachary: the 2nd form would allow you to define class parameters with hiera using the data binding
20:19 xarses dmsimard: thats fine with me
20:19 mgagne dachary: and allow you to include the class as many times as you want
20:20 dachary that's what we want indeed
20:21 xarses mgagne: can you explain how to define the parameters in the second case using data binding?
20:22 dachary dmsimard: I renamed the init section into ceph and udpated the description at https://wiki.openstack.org/wiki/Pu​ppet-openstack/ceph-blueprint#ceph
20:23 mgagne xarses: there is a complete example on puppetlabs website: http://docs.puppetlabs.com/h​iera/1/complete_example.html
20:24 xarses mgagne: thanks
20:24 mgagne xarses: unfortunately (or not), they updated it and are now using json instead of yaml for the example.
20:26 morazi joined #puppet-openstack
20:34 tnoor1 joined #puppet-openstack
20:36 tnoor2 joined #puppet-openstack
20:37 rmoe_ joined #puppet-openstack
20:38 xarses Can a core reviewer look over https://review.openstack.org/#/c/54591/ ?
20:39 xarses puppet-cinder
20:43 openstackgerrit David Moreau Simard proposed a change to stackforge/puppet-ceph: Introduction of ceph installation and configuration management  https://review.openstack.org/56270
20:43 dmsimard xarses: patched with your input
20:43 xarses dmsimard: looking
20:44 otherwiseguy joined #puppet-openstack
20:49 dmsimard xarses: If you want to test it, I put sample (simple!) manifests here: https://gist.github.com/dmsimard/7304179
20:54 mjblack joined #puppet-openstack
20:59 xarses dmsimard: sorry, missed 2 things the first time
21:00 hogepodge joined #puppet-openstack
21:03 dmsimard mon_initial_members is in [mon], not global ?
21:03 dmsimard Not sure what mon_host is
21:04 dmsimard Oh, "mon host", the underscore somehow confused me
21:05 dmsimard xarses: I think both belong in [mon]
21:07 dmsimard xarses: Ok, maybe not - my understanding was not correct. Parameters in [mon] would override the same parameters if they were also declared in [global]. So they can be declared on global.
21:09 dmsimard xarses: I'll switch them around
21:11 xarses dmsimard: I agree that they sound like [mon] items, but as far as i could get out of #ceph, mon_host is required for clients, and mon_initial_members the jury was out on, but it is very much required for to elect quorum on cluster init, and probably restart
21:12 dmsimard is mon_host ultimately the same as a [mon.x] declaration ?
21:13 dmsimard I don't have a mon_host in my CI config and it works well with mon.x sections
21:13 xarses dmsimard: a join of all mon.X/address
21:14 xarses dmsimard: ceph-deploy uses it instead of [mon.X] sections
21:14 hogepodge joined #puppet-openstack
21:15 xarses my understanding was once a client could access any monitor that is a member of the qourm, it will discover all other mons and osds from the mon|osd map
21:15 morazi joined #puppet-openstack
21:15 dmsimard xarses: It's okay, we can support both ways
21:16 xarses so either a list of all monitors is required in mon_host or [mon.X] sections
21:16 xarses for now, lets just fix that fsid isn't actually required then
21:18 dmsimard They'll be set as optional
21:21 dmsimard xarses: about xattr, seems required for ext4 otherwise not necessary ? http://ceph.com/docs/master/rados/c​onfiguration/filestore-config-ref/
21:22 dmsimard xarses: maybe we can set it to true if the filesystem is ext4
21:22 xarses ok, so it would belong in [osd] then?
21:23 dmsimard Well, parameters can be defined both in global or lower (mon, osd, mds). If they are declared at both places, the lower has priority
21:24 xarses ok
21:24 dmsimard The monitors probably make sense in global
21:24 dmsimard xattr though, should be in osd
21:24 dmsimard Haven't mapped any and every parameter yet, though
21:24 xarses for refrence, ceph-deploy puts mon_host and mon_initial_members in [global]
21:25 dmsimard Yeah they'll be there in the next review
21:25 xarses ok
21:26 openstackgerrit David Moreau Simard proposed a change to stackforge/puppet-ceph: Introduction of ceph installation and configuration management  https://review.openstack.org/56270
21:26 dmsimard Aw crap there's a spec failure
21:30 openstackgerrit David Moreau Simard proposed a change to stackforge/puppet-ceph: Introduction of ceph installation and configuration management  https://review.openstack.org/56270
21:31 dmsimard Got to go, I'll address other reviews later tonight or tomorrow - catch you later
22:08 hogepodge joined #puppet-openstack
22:12 badiane_ka joined #puppet-openstack
22:57 rcrit I have one thing I want to do, be able to set the qpid_protocol in glance
22:57 rcrit this requires a change in two submodules in packstack. Do I need a separate bug for each?
22:57 rcrit I filed one already, https://bugs.launchpad.net/packstack/+bug/1251050

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