Wednesday, 2015-04-22

*** jpeeler has quit IRC00:00
*** jpeeler has joined #tripleo00:12
*** alop has quit IRC00:21
*** david-lyle has quit IRC00:24
*** david-lyle has joined #tripleo00:27
*** Marga_ has quit IRC00:32
*** Marga_ has joined #tripleo00:46
*** thrash is now known as thrash|g0ne00:52
*** prad has quit IRC00:53
*** Marga_ has quit IRC00:57
*** prad has joined #tripleo01:08
*** Marga_ has joined #tripleo01:09
*** eghobo has joined #tripleo01:18
*** eghobo_ has joined #tripleo01:20
*** eghobo has quit IRC01:23
*** wfoster has quit IRC01:24
*** wfoster has joined #tripleo01:25
*** julim has joined #tripleo01:46
*** BadCub_Away is now known as BadCub01:56
*** ChuckC has joined #tripleo02:29
*** Marga_ has quit IRC02:32
*** eghobo_ has quit IRC02:46
*** untriaged-bot has joined #tripleo03:00
untriaged-botUntriaged bugs so far:03:00
openstackLaunchpad bug 1446882 in tripleo "cryptography is failing to install" [Undecided,New]03:00
*** untriaged-bot has quit IRC03:00
*** jerryz has joined #tripleo03:32
*** tzumainn has quit IRC03:55
*** eghobo has joined #tripleo03:58
*** yog__ has joined #tripleo04:24
*** masco has joined #tripleo04:27
*** sdake has joined #tripleo04:29
*** eghobo has quit IRC04:31
*** sdake_ has joined #tripleo04:32
*** sdake has quit IRC04:33
*** eghobo has joined #tripleo04:42
*** prad has quit IRC04:45
*** ukalifon has joined #tripleo05:01
*** prad has joined #tripleo05:04
*** penick has joined #tripleo05:10
*** Marga_ has joined #tripleo05:29
*** Marga_ has quit IRC05:29
*** yamahata has joined #tripleo05:29
*** Marga_ has joined #tripleo05:30
openstackgerritMerged openstack/diskimage-builder: Support VHD output format
*** penick has quit IRC05:36
*** sdake_ has quit IRC05:42
*** eghobo has quit IRC06:33
*** sdake has joined #tripleo06:35
*** Marga_ has quit IRC06:38
*** dasm|afk is now known as dasm06:44
*** bvandenh has joined #tripleo06:52
*** dtantsur|afk is now known as dtantsur06:56
openstackgerritDan Sneddon proposed openstack/tripleo-heat-templates: Separate the network config by family and add a "node_index" for each group.
*** shardy has joined #tripleo07:04
*** funzo has joined #tripleo07:09
*** weshay has joined #tripleo07:10
*** rdopiera has joined #tripleo07:16
*** regebro has joined #tripleo07:16
*** aufi has joined #tripleo07:25
*** Goneri has joined #tripleo07:25
*** athomas has joined #tripleo07:25
*** mmagr has joined #tripleo07:25
*** bvandenh has quit IRC07:30
*** morazi has joined #tripleo07:34
*** sdake has quit IRC07:35
*** ifarkas has joined #tripleo07:40
*** rdopiera has quit IRC07:42
*** jcoufal has joined #tripleo07:43
*** sdake has joined #tripleo07:48
*** rdopiera has joined #tripleo07:49
*** jistr has joined #tripleo07:50
*** lucasagomes has joined #tripleo08:00
openstackgerritMartin Mágr proposed openstack/tripleo-heat-templates: Keystone domain for Heat
*** sdake has quit IRC08:00
*** rdopiera has quit IRC08:04
*** yamahata has quit IRC08:04
jistrhi folks, i submitted changes to rights on puppet related repositories:08:09
jistr* tripleo-puppet-elements to move from puppet-manager-core to tripleo-core (it doesn't contain any puppet)08:09
jistr* puppet-tripleo to have shared rights between puppet-manager-core and tripleo-core (it's solely used in tripleo and sometimes it's not easy to find interested reviewers in puppet-manager-core, so we'd expand the list of folks by adding tripleo-core while keeping puppet-manager-core)08:09
jistrhopefully this will allow faster progress on tripleo08:09
jprovaznshardy: good morning08:12
*** mmagr is now known as mmagr|afk08:12
spredzyjistr, hello. Thanks for taking care of that :)08:12
jistrspredzy: morning :)08:12
shardyjprovazn: Hi!08:13
jprovaznshardy: I'm looking at your 2 patches related to hooks - and ...08:14
shardyjprovazn: yup08:14
shardyjprovazn: the hook-poll one still needs some work but should basically be usable08:15
shardyI've got blocked on a heat event API bug, which I'm now looking into08:15
jprovaznshardy: to find the best way if/how to reuse this in a workflow lib so the same polling/hook removal mechanism could be used both for steps during creation or steps during package update08:15
shardybut that only affects the situation where you do multiple updates and let one time out08:15
shardyjprovazn: Ok, yeah perhaps we need to provide a python-heatclient python API for it08:16
shardyI can rework to do that08:16
jprovaznshardy: would it make sense to move the logic of polling/nested events into an utils module so I can import it?08:16
jprovaznand handle just params in the modules08:17
shardyjprovazn: Yes, absolutely, I can rework to do that08:17
jprovaznshardy: awesome, thanks08:17
shardyjprovazn: I was mostly focussed on the devtest use-case initially, which is all shell obviously08:17
shardyjprovazn: adding such an abstraction is a good idea, because i anticipate some API additions during Liberty which will make this much more efficient08:17
shardywe'll need a tranparent way to use that, while still falling back to this slow-but-kilo-compatible code08:18
jprovaznshardy: +1, adding support for nested listing on server side would be really nice to have soon - quick check on Overcloud stack tells me that "heat event-list -n 2 overcloud" calls 142 API calls08:19
shardyjprovazn: yeah, it's horrible :(08:20
shardyYou actually need -n5 I think08:20
jprovaznshardy: I started from the other side with package updates - - in this user story I know exactly where to look for hooks, so I didn't have to go though the whole stack08:20
shardyjprovazn: Ah, yeah that makes it much easier08:20
shardyjprovazn: I've been severely limited by not adding anything to the heat API for this08:21
shardyas I wanted it to work with kilo08:21
*** bvandenh has joined #tripleo08:21
shardyhence all the crazy recursion08:21
jprovaznI think that re-using your code in the shared workflow would be a way to go, but I would add an optional param to allow specify where in the stack to check for hooks - full stack search would be done only if no more specific resource path would be set08:23
shardyjprovazn: +108:23
jprovaznshardy: cool, I will now just copy/paste your code, then replace with import later08:24
*** derekh has joined #tripleo08:25
*** ccrouch has quit IRC08:31
*** pppingme has quit IRC08:31
*** prad has quit IRC08:32
*** rbrady has quit IRC08:32
*** prad has joined #tripleo08:34
*** rbrady has joined #tripleo08:34
*** pppingme has joined #tripleo08:35
*** jprovazn has quit IRC08:37
*** mcornea has joined #tripleo08:43
*** al has quit IRC08:46
*** al has joined #tripleo08:47
*** mmagr|afk is now known as mmagr08:47
*** prad has quit IRC08:53
*** adrianopetrich has quit IRC08:54
*** regebro has quit IRC08:57
*** prad has joined #tripleo08:59
*** untriaged-bot has joined #tripleo09:00
untriaged-botNo untriaged bugs so far! \o/09:00
*** untriaged-bot has quit IRC09:00
*** pelix has joined #tripleo09:01
*** mmagr has quit IRC09:05
*** mmagr has joined #tripleo09:08
*** akrivoka has joined #tripleo09:17
*** masco has quit IRC09:22
*** masco has joined #tripleo09:25
*** adrianopetrich has joined #tripleo09:26
*** dtantsur is now known as dtantsur|brb09:29
*** jtomasek has joined #tripleo09:29
*** jprovazn has joined #tripleo09:35
*** jprovazn has quit IRC09:35
*** jprovazn has joined #tripleo09:35
*** gfidente has joined #tripleo09:45
*** gfidente has joined #tripleo09:45
*** shardy_ has joined #tripleo09:46
*** shardy has quit IRC09:48
*** shardy_ has quit IRC09:52
*** shardy has joined #tripleo09:52
*** ifarkas has quit IRC10:28
*** ifarkas has joined #tripleo10:30
openstackgerritMartin Mágr proposed openstack/tripleo-heat-templates: Keystone domain for Heat
*** Marga_ has joined #tripleo10:46
*** masco has quit IRC10:52
*** dtantsur|brb is now known as dtantsur10:58
*** yuanying has quit IRC11:18
*** adrianopetrich has quit IRC11:25
*** shardy has quit IRC11:28
*** trown|outttypeww is now known as trown11:33
*** thrash|g0ne is now known as thrash11:43
*** wuhg has joined #tripleo11:43
openstackgerritGiulio Fidente proposed openstack/tripleo-heat-templates: Add RabbitMQ as a Pacemaker resource when EnablePacemaker
gfidentejistr, spredzy so my current approach for haproxy and rabbitmq submissions12:02
gfidenteis that we add them as resources to pacemaker in a single node scenario as well12:02
gfidenteas long as EnablePacemaker is true12:02
gfidenteso that it will monitor the service and restart, as if there were multiple nodes12:02
gfidentesounds valid?12:02
*** lucasagomes is now known as lucas-hungry12:07
jistrgfidente: yeah it does (btw that's the way astapor works... everything is set up HA by default. If you just want 1 controller, you still deploy the HA setup, just disable fencing. Seems like a good idea to me, just one config to care about.)12:12
* gfidente sends much ♥ to jistr 12:12
gfidenteso that means we might want a keepalived based job though12:12
*** dprince has joined #tripleo12:13
gfidenteto make sure we don't break things given we run with pacemaker by default12:13
gfidentedprince, right about time12:13
dprincegfidente: hey12:14
spredzyagree with what jistr said. pacemaker configuration for all scenarios (w/ and w/o ha)12:14
spredzymakes it easier to manage12:14
gfidentespredzy, yeah that is what is happening already indeed12:14
gfidentedprince, so if we enable pacemaker by default like we did, we're using in the single controller scenario as well ... and that means we're not testing the keepalived version anymore12:15
gfidentejistr, spredzy, dprince I think it is safe to run the HA job with EnablePacemaker=true and the nonha job with EnablePacemaker=false?12:15
gfidenteunless we want to add an additional job ...12:16
spredzygfidente, agree, but we need to make the ha job green to validate12:16
*** jayg|g0n3 is now known as jayg12:16
gfidentespredzy, so ha job stays on pacemaker12:16
spredzythat EnablePacemaker=true does what it is supposed to do12:16
dprincegfidente: I personally would like the option not to use pacemaker12:16
gfidentedprince, not in the single node scenario?12:17
gfidentethat seems good to me as well12:17
dprincegfidente: My experience w/ HA has been that of a Low availability. For example with our CI could I would like to respin it and I probably don't want to mess w/ HA12:17
* dprince thinks it is too time consuming to troubleshoot all these things ad-hoc12:17
dprinceso while I'd like to support the options to use them I'd like to support the simple case for those of us who prefer not to12:18
gfidentedprince, that means you would run CI for the pacemaker version in the ha job only, am I reading it right?12:20
*** ukalifon has quit IRC12:20
dprincegfidente: sorry, I worded that in a confusing way12:20
dprincegfidente: I was talking about deploying our CI cloud, not running the CI jobs12:21
dprincegfidente: is this what you are talking about
gfidentedprince, not only12:23
dprincegfidente: Having not used pacemaker a lot myself I'm not sure what to expect. I've heard there can be issues with it causing DHCP issues for example. And on a flat network like we use as a TripleO default I would like the option to avoid it12:23
gfidentedprince, for instance here:,cm12:24
gfidentewe add haproxy as a resource to pacemaker for it to monitor the process12:24
gfidenteregardless of the number of nodes12:24
gfidenteas long as enable_pacemaker is true12:24
gfidentethat means we can deploy a single node configuration with enable_pacemaker true or enable_pacemaker false12:24
gfidenteso to make keepalived based job sane, I was thinking to pass EnablePacemaker=false to the nonha job12:25
gfidentewe won't test the nonha pacemaker-based scenario, unless we add another nonha job for it12:25
gfidentebut we would still have pacemaker-based scenario tested in the ha job12:25
dprincegfidente: okay. So long as EnablePacemaker has to be true for these (it looks like that is the case) I think it is fine either way.12:26
*** jdob has joined #tripleo12:27
gfidentedprince, maybe this is a valid conversation to have on the list?12:28
dprincegfidente: so... I think that means do what is easiest, probably enable it even on a single host12:28
gfidentedprince, ack stays on for now is good to me12:28
spredzyI think that rather than having nonha-pacemaker nonha-keepalived ha-pacemaker and ha-keepalived jobs, we could manage the VIP in keepalived for the nonha job and with pacemaker for the ha job, it would save time, resource and avoid long CI cycle12:29
*** rlandy has joined #tripleo12:29
spredzyyet most scenarios remains covered12:29
gfidentespredzy, how about we use keepalived in the precise job?12:30
gfidentethat is obvious there is no puppet there :*12:30
dprincespredzy: but that requires pacemaker everywhere right (which is what I would prefer not to have)12:30
*** rhallisey has joined #tripleo12:30
spredzywaexactly for puppet and precise12:30
gfidentedprince, current status is we do have pacemaker everywhere12:30
dprincespredzy: now I'm confused. I thought EnablePacemaker had to be enabled for it?12:31
spredzydprince, pacemaker is already on the image themselves12:31
dprincespredzy: I don't care if it is in the image. I'm talking about configuring it and starting it.12:31
gfidentedprince, pacemaker enabled here:
gfidentedprince, true by default, so existing jobs are already using pacemaker12:32
gfidentethis was intentional to test it12:32
dprincegfidente: okay, well I would like to suggest we keep all that code, but flip the default12:32
gfidentedprince, well we want it for the ha job12:32
gfidentewe flip the default and enable it only for ha?12:33
gfidente(that means testing keepalived in single node scenario)12:33
dprincegfidente: We don't have to change our defaults to test this for HA12:33
* jistr kinda confused :)12:34
jistrso the current state is we have EnablePacemaker=true for all jobs, right?12:34
gfidentejistr, right12:34
dprincegfidente: we are moving too fast, This patch was posted on Friday and landed on Monday?12:34
jistrand gfidente proposes to switch EnablePacemaker=false *only for the puppet-nonha job* right?12:34
gfidentejistr, right12:35
spredzydprince, yes it needs to be enabled (sorry I misunderstood your original message)12:35
gfidentedprince, I raised it for this very same reason12:35
gfidentedprince, but I think we definitely wanted it in the HA job12:36
gfidentedprince, jistr, spredzy so the main point is about having it only in the HA job or in the nonha job as well12:36
dprincegfidente: having it in the HA job is fine. We've got a different mechanism for that12:36
jistrgfidente, dprince: ok, that sounds sensible to me. I'd definitely like to have EnablePacemaker=true for the HA job. I'd also like to have tested the non-ha variant with EnablePacemaker=true, just to make sure we can deploy also one controller with pacemaker enabled, but i realize we should also test the keepalived scenario somewhere, so i'd be ok with having EnablePacemaker=false in the non-ha job.12:37
jistrgfidente, dprince: i think the scenarios when deploying 3 controllers with Pacemaker works and deploying 1 controller doesn't work should be quite rare, so i don't consider testing 1 controller with Pacemaker as a must. (I guess we don't want to put more load on the CI cloud by creating another job unless necessary.)12:40
dprincejistr: we could probably mix it in somewhere I think if we need to (another job)12:41
dprincejistr: but I'm happy with just pacemaker for the HA job initially I think12:41
jistrdprince: ack, sounds good to me12:42
spredzydprince, jistr something I lost in the midle, when you say 'just pacemaker' for the HA job initialy. Are we talking about the VIP management? or all the services management also ?12:43
gfidenteall services12:43
gfidenteI consider VIP just one of the services in charge to pacemaker12:43
dprincespredzy: everything I think12:44
spredzydprince, gfidente ack that12:44
*** sseago has joined #tripleo12:47
openstackgerritDan Prince proposed openstack-infra/tripleo-ci: Set EnablePacemaker for the HA job
gfidentedprince, will you flip the default in the template as well?12:48
openstackgerritDan Prince proposed openstack/tripleo-heat-templates: Set EnablePacemaker == false be default
dprincegfidente: done12:48
dprincegfidente: I think those two patches are all we need here and everything else looks great12:49
dprincegfidente: the tripleo-ci patch is new territory... (we just added that option last week) so lets see how it goes12:49
*** radez_g0n3 is now known as radez12:50
spredzyjistr, dprince gfidente jsut so we all agree, some from now on for every service we bring into tripleo-heat-templates, we need to provide a way that is is handled with and without pacemaker, correct ?12:51
dprincespredzy: I think so. how painful will this be?12:51
gfidentecurrently we do have all working without pacemaker12:52
gfidenteI think we just need to make sure we don't break them when pacemaker is enabled12:52
spredzydprince, it might look ugly :) For the pain, I need to setup the galera thing and will be able to tell12:52
spredzybut manifests might get slightly more complex to read12:52
gfidentespredzy, well galera for now won't make sense in the non-pacemaker version though12:52
dprincespredzy: lets see an example or two. If it is going to trash our manifests perhaps we debate this on the list12:53
gfidentespredzy, dprince here is an example
gfidentethere I try to setup the cluster with pacemaker if enabled12:53
gfidenteor with puppet if pacemaker is not enabled12:53
*** adrianopetrich has joined #tripleo12:54
gfidenteso this change ideally allows for all four combinations: single/multiple with/without pacemaker12:54
*** spzala has joined #tripleo12:56
jistrgfidente, spredzy, dprince: would it be possible to care for the non-pacemaker case only about 1 node initially? I think that might simplify things a bit. E.g. setting up multi-node Galera without Pacemaker is not very clean, so if we only care about single-node when EnablePacemaker=false, it might make things considerably easier.12:59
jistrHA setup without Pacemaker could be added later12:59
gfidentejistr, yes I think this is the goal13:00
gfidenteexcept for rabbitmq apparently wasn't so terrible, but I don't think we expect non-pacemaker multiple-node to work anytime soon13:00
gfidentejistr, and btw the rabbitmq non-pacemaker multiple-node doesn't work as is, as we know, you had a patch ... so it's definitely 2nd citizen at least13:01
* jistr still thinking that at some point we might need to leave the single-manifest approach to keep sanity, especially if we want to support various alternatives like we're doing for pacemaker vs. keepalived13:01
openstackgerritDan Prince proposed openstack-infra/tripleo-ci: Add a ceph job
openstackgerritDan Prince proposed openstack-infra/tripleo-ci: Set EnablePacemaker for the HA job
*** lucas-hungry is now known as lucasagomes13:05
spredzyjistr, would it be possible today to generate the manifest.pp ? I mean if we create an overcloud_rabbitmq.pp where we do all the work about rabbitmq  but I want the final pp file to be the concatenation of overcloud_controller.p and overcloud_rabbitmq.pp is it feasible ?13:06
dprincejistr: perhaps we can move some of the messy stuff into puppet-tripleo to help keep sanity then13:08
dprincejistr: I still think there is value in the single manifest per role approach (even though it duplicates some things)13:09
dprincejistr: it makes them very hackable for example... having almost everything identified in a single file13:09
jistrspredzy: i'd probably like "include <something>" more than generating it though...13:10
*** lblanchard has joined #tripleo13:10
jistrdprince: yeah, moving parts to puppet-tripleo would be a solution imho13:11
jistrdprince: i think what makes the puppet-tripleo not-so-hackable is that we need to fiddle with the images. If we had a way to write content into the machine after it's been started by heat, but before Puppet starts running, puppet-tripleo (and the rest of puppet modules) would suddendly become much more hackable :)13:13
jistri wanted to explore that a bit but didn't have enough bandwidth13:13
dprincejistr: right, which is why we've tried to avoid puppet much into puppet-tripleo13:13
gfidentejprovazn, jpeeler do you guy know if we can give multiple args to get_file or repeat it?13:14
*** tzumainn has joined #tripleo13:15
gfidentedprince, on the ceph job13:16
gfidentewhy ControllerEnableCephStorage: true then?13:16
dprincegfidente: because without that the job would fail13:17
spredzyjistr, gfidente dprince so should we start thinking in terms of tripleo::rabbitmq, tripleo::mariadb etc... from now on ?13:17
dprincegfidente: because cinder scheduler has a race... it has to run a periodic cycle or two to find all its storage13:17
dprincegfidente: sometimes it may pass, sometimes not.13:17
gfidenteyeah I remember that now13:18
dprincegfidente: by having local storage on the controller we make sure it always has some space13:18
*** yamahata has joined #tripleo13:19
dprincespredzy: not sure yet. That may clean up some things... but we lose the hackability13:19
slaglejistr: EmilienM : what about also  moving puppet-tripleo to openstack/13:19
EmilienMslagle: it's WIP Sir :-)13:19
dprinceslagle: it will follow the stackforge modules I think13:20
dprinceslagle: which are all moving there...13:20
EmilienMslagle: we are in the process to move under the big tent. TC agreed (yesterday), now I have to figure out technical details13:20
slagleah, ok13:20
gfidentedprince, but I don't think it's worth signalling the controllers about the ceph completion, sounds like overengineering for a pretty peculiar case where you want to use cinder in the first minute of deployment13:20
spredzydprince, as we need to remount the overcloud_controller image, bring the changes and unmount it for each change we wanna try ?13:21
dprincegfidente: the right solution for this is to use post deployment validations13:21
jpeelergfidente: i don't - would try repeating it first13:21
gfidentedprince, ack13:21
dprincespredzy: yes, going through that process (rebuilding images or editing them manually) really slows down the dev cycle13:22
jprovazngfidente: sorry, no idea13:23
dprincegfidente: I'm not trying to over-engineer ceph. Just make it work in our CI13:23
dprincegfidente: FWIW, *this* problem is what I was talking about a week or so ago when I mentioned Ceph wasn't working13:23
gfidentedprince, the over-engineering was re-configuring cinder after ceph completed instead of doing it in parallel13:23
gfidentedprince, yeah and indeed I still don't consider it an issue if not for the peculiar case of CI13:24
gfidentedprince, validation will fix it implicitly13:24
gfidentedprince, on the keepalived / pacemaker thing,
gfidentedprince, the idea is to pass manage_vip false when enablepacemaker13:32
*** jpeeler has quit IRC13:32
*** rhallisey has quit IRC13:36
openstackgerritDerek Higgins proposed openstack/tripleo-image-elements: Revert "An element to force install Qemu 2.1.3-3 on F21"
*** jpeeler has joined #tripleo13:44
*** radez is now known as radez_g0n313:49
*** rlandy has quit IRC13:49
*** rhallisey has joined #tripleo13:49
*** rlandy has joined #tripleo13:50
*** yog__ has quit IRC13:52
*** mcornea has quit IRC13:53
*** mcornea has joined #tripleo13:58
*** jcoufal has quit IRC13:59
*** Marga_ has quit IRC14:02
openstackgerritDan Prince proposed openstack-infra/tripleo-ci: Set root disk == node disk for puppet jobs
openstackgerritDan Prince proposed openstack-infra/tripleo-ci: Add a ceph job
gfidentedprince, can you help me figure what is the distinguo between the regular ephemeral storage of a vm and the ephmeral storage defined as --epehmeral by the flavor?14:19
*** ChuckC has quit IRC14:19
gfidentedprince, I am not sure what implications it has because I see some 'ephemeral' storage defined for the flavor14:27
gfidentebut my VMs just have the sum of Disk+Ephrmeral as a single partition14:27
*** Marga_ has joined #tripleo14:29
gfidentedprince, I see the ephemeral is in /mnt14:29
dprincegfidente: in TripleO if you set ROOT_DISK == to the NODE_DISK there will be no ephemeral14:29
*** wuhg has quit IRC14:29
gfidentebut at the nova level, what is the difference in between the two?14:29
dprincegfidente: I'm pushing us towards doing this as a default
gfidenteshall ceph prefer the other?14:29
gfidentedprince, I noticed, that's why I asked14:29
*** mmagr is now known as mmagr|afk14:30
dprincegfidente: we aren't making use of /mnt with puppet14:30
dprincegfidente: it is a waste of space to have it14:30
gfidentedprince, ack, for nova I mean14:30
dprincegfidente: furthermore your job would actually fail because the root disk won't have enough space14:30
gfidenteis there a difference in between the two?14:30
*** radez_g0n3 is now known as radez14:31
dprincegfidente: the difference is / is the root partition, and /mnt is not (if using ephemeral)14:31
gfidentebut they are both deleted when vm is terminated14:31
gfidentewe might want ceph to use /mnt instead of /srv/data?14:32
dprincegfidente: Nova/Ironic is really great... but the one abstraction that doesn't really fit is the name "ephemeral"14:32
*** mmagr|afk is now known as mmagr14:32
dprincegfidente: If /mnt is mounted on the root partition it works fine.14:32
dprincegfidente: turns out using non-default locations for things like the tripleo-image-elements do is a real pain.14:33
dprincegfidente: why not use the project default instead... be it /srv/data... or /var/lib/nova...14:34
dprincegfidente: We can map more complex partitioning strategies into our deployments perhaps eventually too. But using the default configuration location and not hard coding /mnt/state everywhere is easier IMO14:35
dprincegfidente: for now... set NODE_DISK == ROOT_DISK in your environments, enjoy the extra space, and be happy14:36
*** jerryz has quit IRC14:36
gfidenteso you think /srv/data for OSDs is fine?14:36
dprincegfidente: sure14:37
gfidentewe should at least move it into the static ceph.yaml though14:38
gfidentewill do it14:38
*** akrivoka has quit IRC14:38
*** ifarkas has quit IRC14:39
*** akrivoka has joined #tripleo14:40
openstackgerritDan Prince proposed openstack/tripleo-incubator: Update puppet docs to set ROOT_DISK
dprincegfidente: added a note to clarify this in the docs too ^14:42
*** sdake has joined #tripleo14:43
*** yamahata has quit IRC14:43
*** yamahata has joined #tripleo14:44
*** sdake_ has joined #tripleo14:44
openstackgerritGiulio Fidente proposed openstack/tripleo-heat-templates: Move Ceph OSD data path configuration into static ceph.yaml file
*** mcornea has quit IRC14:46
*** mcornea has joined #tripleo14:47
*** sdake has quit IRC14:48
*** jerryz has joined #tripleo14:48
*** ukalifon has joined #tripleo14:51
*** jprovazn has quit IRC14:52
*** untriaged-bot has joined #tripleo15:00
untriaged-botNo untriaged bugs so far! \o/15:00
*** untriaged-bot has quit IRC15:00
*** akrivoka has quit IRC15:03
*** dsneddon has quit IRC15:03
*** dsneddon has joined #tripleo15:04
*** aufi has quit IRC15:14
*** Marga_ has quit IRC15:17
*** rlandy has quit IRC15:21
*** rlandy has joined #tripleo15:23
openstackgerritYanis Guenane proposed openstack/tripleo-incubator: Allow a dev to specify puppet-modules override
spredzydprince, jistr gfidente would you buy an approach that look like this ^^ ?15:27
dprincespredzy: so the goal here is to inject custom puppet modules into your image?15:29
dprincespredzy: I think we already support that w/ source repositories15:29
spredzydprince, thanks. reading15:30
dprincespredzy: so like take puppet-neutron for example. I could set export DIB_REPOREF_puppet_neutron=<refspec>15:30
*** jprovazn has joined #tripleo15:30
dprincespredzy: and then when I build images (with disk-image-create) it will include that puppet module from Git15:31
spredzydprince, same would work w/ non stackforge/openstack related modules ?15:31
spredzylike puppetlabs-rabbitmq ?15:31
dprincespredzy: it works for any of the modules we have here:
dprincespredzy: so long as you have either DIB_INSTALLTYPE_puppet_modules=source or DIB_DEFAULT_INSTALLTYPE=source (you probably won't have the second set if you are using packages though)15:32
dprincespredzy: so for example mysql would use #export DIB_REPOREF_puppetlabs_mysql=....15:32
dprincespredzy: for mysql (since it isn't on Gerrit) you could use a stable release15:33
dprincespredzy: which reminds me... I need to push an element to get us using stable deps for rabbit, mysql, etc. (but not the stackforge/puppet* modules which we want to use master)15:33
spredzyok. is it possible to get a specific branch ? Or shoulld I push the change to my fork master ?15:34
spredzydprince, ^15:34
*** rwsu has joined #tripleo15:34
dprincespredzy: I think source repositories works with tags, branches, and refspecs for Git15:35
*** pdb has quit IRC15:35
spredzyok awesome. Just abandoner my review then15:35
*** pdb has joined #tripleo15:35
spredzywasn't aware of this feature and how it worked.15:35
dprincespredzy: yeah, it gives quite a bit of flexability in terms of how we build images and construct the CI jobs15:36
spredzydprince, So about what we were talking ealier and the 'hackability' of putting everything in puppet-tripleo (and other puppet modules), if I want to run it with the set of module I need I simply need to export the proper env variables, correct ?15:37
spredzyrun a devtest_opvercloud15:37
dprincespredzy: correct, and then rebuild your images.15:37
dprincespredzy: it is the rebuilding of images that is the real pain though15:37
dprincespredzy: doesn't take that long... but it does slow down the dev cycle a bit. Always has for TripleO.15:38
dprincespredzy: the puppet integration (with heat) has actually sped up the dev iteration time for my quite a bit15:38
dprincespredzy: for example I often just do a 'heat stack-update' to test new changes....15:38
spredzydprince, ack. The earlier review, wasn't rebuilding them simply mounting putting the neede change unmount15:39
dprincespredzy: but... the more we put into puppet-tripleo the less we'll be able to do this15:39
spredzybut its a duplication of a feature that already exists :/15:39
dprincespredzy: yeah, I'm not sure adding that to devtest is ideal15:39
spredzyidea was to keep using -c (for cache) and add a 40sec overhead to have the right module in15:39
*** mmagr has quit IRC15:40
dprincespredzy: I'd rather have a solution that worked in production15:40
spredzydprince, yep totally understand. Will do the DIB_ export from now on.15:40
dprincespredzy: so rather than this... what if we simply had a Heat software deployment that would allow you to push the same updates15:40
jistrdprince: +1 on that15:41
dprincespredzy: we don't have anything to do this yet... but it would speed up the dev iteration quite a bit15:41
spredzythat would run before Step1 ?15:41
jistr(i'd like some solution which doesn't require rebuilding the images)15:41
dprincespredzy, jistr lets put some thought into this15:41
spredzydprince, jistr sounds like a good idea to me15:41
jistrdprince: the part i wasn't able to come up with when thinking about this previously is how to push/pull the data there... not sure if they should go through Heat, maybe not?15:42
dprinceI think it is a gap between how tripleo-heat-templates works today with its module deployment and how for example a puppet-master setup would work (where it allows you to push out modules...)15:42
dprincejistr: Yeah, It is a bit tricky. A bit beyond the scope of what a get_file could do I think15:42
jistrdprince: yeah because the data can be potentially quite large15:43
*** yamahata has quit IRC15:43
*** ukalifon has quit IRC15:43
dprincejistr: perhaps a mechanism to integrate with an external sync somewhere (rsync). And then the deployment just runs it.15:43
dprincejistr: that would be lighter weight and could work quite nicely15:43
dprincejistr: so first step would be: rsync the latest modules.15:44
*** akrivoka has joined #tripleo15:44
jistryeah sounds good. I thought of NFS previously but that's kinda heavier than rsync. Rsync sounds good15:44
dprincejistr: then the deployment proceeds as normal15:44
dprinceRight, so we we have these things you'd even have the option not to pre-install any puppet modules in the image itself. You still could... for tried and true tested images. But for hackability you could just build a really light base image and go at it15:45
dprincemuch faster dev cycle...15:45
spredzydprince, just so I make sure I get it. on some rsync server you will have a version of the /opt/stack/puppet-modules with the one you want, and heat will go and run rsync to get it on each server it spawns on the overcloud ?15:45
dprincespredzy: yep, I think that is the idea on the table here15:46
jistri wonder if we want to make this even more generic though...15:47
*** weshay has quit IRC15:47
spredzydprince, ok15:47
spredzyjistr, how so ?15:47
* jistr searches15:47
dprincejistr: oh, do keep wondering :)15:47
*** sdake_ has quit IRC15:48
* spredzy have to leave15:48
*** eghobo has joined #tripleo15:49
dprincespredzy: see ya15:49
jistrsee ya15:49
*** spredzy is now known as spredzy|afk15:49
gfidenteradez, you around?15:50
jistrwhat if we used heat-config-script for that, and gave the machines a bash script. You want to rsync modules, you write that. You want to rsync something else, you write rsync something else. You want some completely different customization, you write whatever :)15:50
radezgfidente: yea what's up?15:50
gfidenteradez, you the owner of puppet-pacemaker?15:50
jistrdprince: ^15:51
gfidenteradez, I couldn't find an easy way to set the no-quorum-policy , is there any? or can we add it?15:51
radezgfidente: I started it some time ago but I haven't worked on it in ages15:51
gfidenteradez, but you're here so you're called into action!15:52
* gfidente hopes15:52
radezI believe patches are welcome, jayg are you the right person to verify that?15:52
gfidentelooks like it's a generic version of stonith.pp ?15:52
radezgfidente: haha that's how it works sometimes :)15:52
jistrgfidente: maybe we could even use just another puppet run before our current puppet runs, and we'd do the rsync via puppet. It might be a bit more complex to get running than just writing a bash script, but Puppet somewhat guarantees idempotency, which means we wouldn't have to be worried about running the same thing on stack-update too.15:53
jistrdprince: ^15:53
jistr(sorry gfidente :) )15:53
gfidentejistr, I don't want to read about the mess there15:53
gfidentejistr, I trust you15:53
dprincejistr: re, heat-config-script. I think that could be useful. FWIW we are already heading towards using it for validations as well:, and15:54
dprincejistr: oops, those came out in the wrong order15:54
gfidentejayg, ping when have a few mins15:55
jistrdprince: ok... would we run the same script on stack-update too, and trust the writer to write it in an idempotent manner?15:55
jistrdprince: maybe using puppet for it is overkill (referencing what i wrote above)15:56
*** CheKoLyN has joined #tripleo15:56
*** funzo has quit IRC15:57
*** funzo has joined #tripleo15:58
*** Goneri has quit IRC15:59
*** Marga_ has joined #tripleo16:00
*** funzo has quit IRC16:02
*** dprince has quit IRC16:07
*** jistr has quit IRC16:12
*** radez is now known as radez_afk16:21
*** morazi has quit IRC16:21
*** athomas has quit IRC16:22
*** dprince has joined #tripleo16:25
*** akrivoka has quit IRC16:28
*** Marga_ has quit IRC16:29
*** Marga_ has joined #tripleo16:30
*** Marga_ has quit IRC16:33
*** Marga_ has joined #tripleo16:34
*** dtantsur is now known as dtantsur|afk16:38
*** alop has joined #tripleo16:44
*** subscope_ has joined #tripleo16:48
*** jtomasek has quit IRC16:49
*** penick has joined #tripleo16:53
*** jdob has quit IRC17:00
*** derekh has quit IRC17:00
*** subscope_ has quit IRC17:07
*** subscope_ has joined #tripleo17:07
*** trown is now known as trown|lunch17:11
*** therve has joined #tripleo17:13
therveDoes someone know how to use devpi to include custom packages in disk-image-builder?17:15
therveTrying to follow some emails and README, but I'm not succeeding17:16
dprincedsneddon: hi. So I've got a function here which should allow us to obtain the correct bind IP for a subnet:
dprincedsneddon: working on a patch now that allows us to apply that to a service... starting w/ keystone17:20
*** sdake has joined #tripleo17:24
*** ChuckC has joined #tripleo17:26
*** lucasagomes is now known as lucas-dinner17:28
*** sdake has quit IRC17:29
*** sdake has joined #tripleo17:30
-openstackstatus- NOTICE: gerrit is restarting to clear hung stream-events tasks. any review events between 16:48 and 17:32 utc will need to be rechecked or have their approval votes reapplied to trigger testing in zuul17:33
*** Shrews has joined #tripleo17:35
*** pelix has quit IRC17:38
*** Shrews has left #tripleo17:38
*** david-lyle has quit IRC17:42
*** radez_afk is now known as radez17:45
*** rbrady has quit IRC17:45
*** alop has quit IRC17:47
jayggfidente: sorry I missed your ping before, what's up?17:50
*** yamahata has joined #tripleo17:50
gfidentejayg, I'm not around, sent you an email though17:52
*** mcornea has quit IRC17:53
*** jerryz has quit IRC17:53
jayghaha, ok, I'll check17:54
*** rbrady has joined #tripleo17:59
*** rbrady has joined #tripleo17:59
dsneddondprince: Thanks! I'll have a look.17:59
dprincedsneddon: just thinking about this a few minutes I think I'm going to push a new patch which handles a "default" value parameter. I think we'll need that to wire in support to use the control plane network as a default18:01
dprincedsneddon: as for the Heat templates, I'm wondering what a good default value for these parameters might be. If any?18:01
dsneddondprince, I see, so that will take effect if the subnet is empty, or when the subnet is not matched?18:01
dprincedsneddon: exactly18:01
*** alop has joined #tripleo18:02
dprincedsneddon: for Heat template defaults... we could have it be blank by default. Or we could take an alternate approach and actually recommend some default networks. If no IP gets found it could just use the provided default (control plane)18:03
*** yamahata has quit IRC18:03
*** jdob has joined #tripleo18:04
dsneddondprince, I think a good default is blank, and then the services default to controllerhost. Does this function need to return a default, or should the function only get called if subnet is not blank?18:04
dprincedsneddon: The cleanest way to wire it in is probably to always use the function, and just have the find_bind_ip address default to control plane if no subnet IP is found18:05
dsneddondprince: Assuming we can figure out which subnet is the control plane. Do we even have a way to figure out which is the IP on the first interface? Like eth0, em1, etc. Because that's the provisioning network.18:06
dprincedsneddon: no need to figure it out. I'll just pass it in because we already know that via Heat18:06
dsneddondprince, OK, that's better.18:06
dprincedsneddon: turns out Heat knows this one because it is assigned via Nova ;)18:06
dprincedsneddon: so as a default this works quite well18:07
openstackgerritEmilien Macchi proposed openstack/tripleo-heat-templates: puppet/controller: validate RabbitMQ with serverspec
EmilienMdprince: very WIP for now, but this is my start18:23
*** trown|lunch is now known as trown18:25
*** Goneri has joined #tripleo18:29
*** yamahata has joined #tripleo18:33
*** yamahata has quit IRC18:40
dprinceEmilienM: cool. looks good18:50
*** alop has quit IRC18:56
*** pblaho_ has joined #tripleo19:02
*** lsmola__ has joined #tripleo19:02
*** dtantsur|afk has quit IRC19:04
*** pblaho has quit IRC19:05
openstackgerritMerged openstack/diskimage-builder: Append full path to img-functions:run_in_target
*** lsmola_ has quit IRC19:06
*** hewbrocca has quit IRC19:06
*** lsmola__ has quit IRC19:10
*** pblaho__ has joined #tripleo19:10
*** pblaho_ has quit IRC19:10
*** dtantsur has joined #tripleo19:10
*** hewbrocca has joined #tripleo19:12
*** MasterPiece has joined #tripleo19:12
*** lsmola__ has joined #tripleo19:22
openstackgerritgreghaynes proposed openstack/diskimage-builder: Initial element tests
*** alop has joined #tripleo19:36
openstackgerritMerged openstack/diskimage-builder: Break install-types out of base
*** shakamunyi has joined #tripleo19:40
openstackgerritEmilien Macchi proposed openstack/tripleo-heat-templates: puppet: install Horizon on overcloud-controller
*** openstackgerrit has quit IRC19:54
*** openstackgerrit has joined #tripleo19:55
openstackgerritJan Provaznik proposed openstack/tripleo-common: Add basic project structure
*** therve has left #tripleo19:56
*** Marga_ has quit IRC19:57
*** radez is now known as radez_g0n319:58
*** subscope_ has quit IRC19:59
*** alop has quit IRC20:00
openstackgerritMerged openstack/tripleo-heat-templates: Add option to enable ceph storage on controller
openstackgerritMerged openstack/tripleo-heat-templates: Parameter to enable/disable controller swift store
openstackgerritMerged openstack/tripleo-heat-templates: Set EnablePacemaker == false be default
*** alop has joined #tripleo20:07
*** jtomasek has joined #tripleo20:14
*** rlandy has quit IRC20:25
*** jdob has quit IRC20:27
*** gfidente has quit IRC20:31
*** jtomasek has quit IRC20:32
*** alop has quit IRC20:32
*** trown is now known as trown|outttypeww20:38
*** jprovazn has quit IRC20:39
*** jayg is now known as jayg|g0n320:49
*** alop has joined #tripleo20:52
openstackgerritBen Nemec proposed openstack/os-cloud-config: Enable ssl registration for Heat
openstackgerritBen Nemec proposed openstack/tripleo-incubator: Enable Heat SSL support
*** Marga_ has joined #tripleo20:59
*** untriaged-bot has joined #tripleo21:00
untriaged-botNo untriaged bugs so far! \o/21:00
*** untriaged-bot has quit IRC21:00
*** Marga_ has quit IRC21:03
*** alop has quit IRC21:04
*** jtomasek has joined #tripleo21:05
*** Marga_ has joined #tripleo21:07
*** dprince has quit IRC21:13
*** alop has joined #tripleo21:17
*** lblanchard has quit IRC21:22
*** bvandenh has quit IRC21:24
*** rhallisey has quit IRC21:25
*** thrash is now known as thrash|g0ne21:25
*** openstackgerrit has quit IRC21:29
*** openstackgerrit has joined #tripleo21:30
*** yamahata has joined #tripleo21:34
openstackgerritgreghaynes proposed openstack/diskimage-builder: Add functional smoke test for disk-image-create
*** Goneri has quit IRC21:39
*** rhallisey has joined #tripleo21:40
*** lucas-dinner has quit IRC21:41
openstackgerritgreghaynes proposed openstack/diskimage-builder: Support building ACIs
openstackgerritgreghaynes proposed openstack/diskimage-builder: Add functional smoke test for disk-image-create
*** alop has quit IRC21:43
*** sdake has quit IRC21:44
openstackgerritBen Nemec proposed stackforge/puppet-tripleo: Allow use of ssl for public api endpoints
greghaynesslagle: SpamapS ^ would really like a re-review for the functest patch, infra just merged the job21:45
*** alop has joined #tripleo21:45
*** yamahata has quit IRC21:46
openstackgerritgreghaynes proposed openstack/diskimage-builder: Initial element tests
*** dprince has joined #tripleo21:46
*** dprince has quit IRC21:58
*** rhallisey has quit IRC22:01
*** jtomasek has quit IRC22:07
*** openstackgerrit has quit IRC22:11
*** openstackgerrit has joined #tripleo22:11
openstackgerritSteve Baker proposed openstack/diskimage-builder: Mount fallback with -o nouuid for XFS base images
*** rhallisey has joined #tripleo22:16
*** spzala has quit IRC22:21
openstackgerritMerged openstack/diskimage-builder: Add element to process config-drive network info
openstackgerritMerged openstack/diskimage-builder: Add a yum-minimal element that just uses yum
*** rhallisey has quit IRC22:29
greghaynesmordred: ^22:30
greghaynessimple-init merged22:30
lifelesshmm, no dan22:31
lifelessbnemec: hi22:31
*** sdake has joined #tripleo22:31
*** BadCub is now known as BadCub_Hiding22:31
lifelessbnemec: I want to make contact with a rh'er about this thread - getting 0.11 out the door. Maybe even 1.022:32
lifeless and the concerns around pbr22:32
lifelessbnemec: if you can link me up with someone, so I can work through it and make the patching out unneeded, that would be sweet22:32
openstackgerritMerged openstack/tripleo-image-elements: Enforce owner readonly permissions on stackrc
bnemeclifeless: Sure, I'll see if I can figure out who the right person is.22:36
openstackgerritMerged openstack-infra/tripleo-ci: Set EnablePacemaker for the HA job
lifelessbnemec: wicked, thanks22:38
*** rhallisey has joined #tripleo22:40
openstackgerritMerged openstack/os-apply-config: Updated from global requirements
*** shakamunyi has quit IRC22:46
*** alop has quit IRC22:52
*** alop has joined #tripleo22:53
*** penick has quit IRC22:55
*** eghobo_ has joined #tripleo23:00
*** ChuckC has quit IRC23:00
*** eghobo has quit IRC23:02
*** yuanying has joined #tripleo23:03
*** spzala has joined #tripleo23:06
openstackgerritMerged openstack/tripleo-incubator: Remove unused URL_BASENAME from acquire-image
openstackgerritMerged openstack/tripleo-image-elements: Compile horizon translation files.
* bnemec approves all the things23:16
*** eghobo_ has quit IRC23:18
greghaynesI approve of that23:18
*** ChuckC has joined #tripleo23:21
*** david-lyle has joined #tripleo23:24
*** ChuckC has quit IRC23:25
*** ChuckC has joined #tripleo23:25
*** CheKoLyN has quit IRC23:27
*** alop has quit IRC23:29
*** eghobo has joined #tripleo23:29
*** panda has quit IRC23:31
*** panda has joined #tripleo23:31
SpamapSgreghaynes: fast tracked and +A'd :)23:36
*** Marga_ has quit IRC23:36
*** Marga_ has joined #tripleo23:37
*** sdake has quit IRC23:39
openstackgerritBen Nemec proposed openstack/os-cloud-config: Fix tuskar path
*** eghobo has quit IRC23:41
mordredgreghaynes: ZOMGZOMGZOMGZOMG23:45
* mordred hands all the people all the beers23:46
*** eghobo has joined #tripleo23:46
greghaynesyep, party time23:47

Generated by 2.14.0 by Marius Gedminas - find it at!