| harlowja | 'The server name validator intercepts server instance create/update calls to validate whether the name that is given to the server adheres to GD standards, is constrained by length and characters, and is unique amongst all servers registered within Active Directory.' | 00:00 |
|---|---|---|
| harlowja | thats one example | 00:00 |
| wirehead_ | Rackspace’s open source version: http://www.openrepose.org/ | 00:00 |
| harlowja | nothing crazy (and yes it should be upstreamed, and yes, i'm working on that issue, haha) | 00:00 |
| sdake | repose | 00:00 |
| sdake | i am familiar with that | 00:00 |
| wirehead_ | Same ideas as Netflix Zuul. | 00:00 |
| sdake | java | 00:00 |
| * sdake groans | 00:00 | |
| *** absubram has joined #openstack-kolla | 00:00 | |
| harlowja | https://pypi.python.org/pypi/wafflehaus | 00:00 |
| harlowja | 'Cloud providers of all sizes have problems dealing with business-requirements of their cloud and diverging from upstream. By putting business logic into middleware providers can create special cases without changing upstream code.' | 00:00 |
| harlowja | which i call BS | 00:00 |
| harlowja | but ok, whatever, it pre-exists me :-P | 00:01 |
| sdake | ok so how does this middleware get loaded into a typical openstack daemon | 00:01 |
| sdake | config option? | 00:01 |
| harlowja | paste.ini files (the normal ones openstack rest endpoints have) | 00:01 |
| *** absubram_ has joined #openstack-kolla | 00:01 | |
| wirehead_ | Here’s another one: https://getkong.org/ | 00:01 |
| harlowja | https://github.com/openstack/nova/blob/master/etc/nova/api-paste.ini#L17 | 00:01 |
| wirehead_ | The overall API networked thingie community is quite taken by things such as Kong and Zuul and there are commercial management frameworks like 3scale. | 00:02 |
| harlowja | ya, i mean i get the desire to have that | 00:02 |
| harlowja | it just feels up like a copout | 00:02 |
| harlowja | which is why i'm pushing back on any new ones | 00:02 |
| sdake | harlowja totally agree on both parts | 00:02 |
| harlowja | but ya | 00:02 |
| sdake | anyway | 00:03 |
| sdake | so some special code needs to get loaded in each container? | 00:03 |
| harlowja | so i think (or the thinking is) that our waffles we will turn just into another pypi package | 00:03 |
| *** ravig has joined #openstack-kolla | 00:03 | |
| harlowja | that we'll just have to put as a requirement of these projects when building them (so that the waffles gets pulled down) | 00:03 |
| *** MarMat has quit IRC | 00:03 | |
| wirehead_ | Well, in Kolla-K8s, you can just run it as a sidecar container. :) | 00:03 |
| harlowja | (and that pypi project would be in our artifactory pypi repo where it can be accessed during container build time) | 00:03 |
| sdake | harlowja another option is to use the customization work we are planning or newton and carry backports | 00:03 |
| sdake | however modifying an internal git repo's requirements.txt would also work | 00:04 |
| sdake | and probably lower level of pain involved | 00:04 |
| harlowja | ya, either might be ok till i can get rid of this junk | 00:04 |
| harlowja | (which probably will take a while...) | 00:04 |
| sdake | what about the bare metal deployment part | 00:04 |
| harlowja | ya, deployment, working on exploring that :-P | 00:05 |
| harlowja | phase-next | 00:05 |
| -openstackstatus- NOTICE: Gerrit is offline for maintenance until 01:45 UTC (new ETA) | 00:05 | |
| *** absubram has quit IRC | 00:05 | |
| *** absubram_ is now known as absubram | 00:05 | |
| *** ChanServ changes topic to "Gerrit is offline for maintenance until 01:45 UTC (new ETA)" | 00:05 | |
| sdake | you already have some bare metals deployed | 00:05 |
| sdake | I assume yu plan to keep them | 00:05 |
| harlowja | yup yup | 00:05 |
| Daviey | sdake: I was just asking about your upgrade documentation... should it include how to take a backup snapshot of the control plane data and possible rollback guide? | 00:05 |
| harlowja | soooooo what we have right now is a bunch of puppet stuff that deploys out rpms and such | 00:06 |
| sdake | Daviey right - there is no such ability in kolla | 00:06 |
| sdake | Daviey go big or go home ;) | 00:06 |
| Daviey | sdake: Well there is... it just might not be pretty. :) | 00:06 |
| harlowja | but ya, the dropping puppet --> kolla not yet sure how it will go | 00:06 |
| harlowja | but thats ok | 00:06 |
| sdake | Daviey yes your right it is possible in the current implementation | 00:06 |
| harlowja | (for deployment) | 00:07 |
| sdake | Daviey but i think its too advanced to document | 00:07 |
| Daviey | harlowja: Are you lookin to inplace migrate or create a new cloud? | 00:07 |
| harlowja | none of that create new cloud juk | 00:07 |
| Daviey | sdake: Yeah.. poop. | 00:07 |
| harlowja | not expecting this to be easy, but then the worthwhile things are never easy :-P | 00:07 |
| harlowja | tis life | 00:07 |
| harlowja | ha | 00:07 |
| sdake | kolla is extremely easy | 00:08 |
| harlowja | :) | 00:08 |
| sdake | but the data backup and loading into your existing deploy will be a challenge to sort out | 00:08 |
| harlowja | yup | 00:08 |
| sdake | mind I ask how big the team is? | 00:08 |
| Daviey | harlowja: I'm looking to create a new cloud and migrate instances into it.. growing the new region and shrink the old one, server by server. (with some automation) | 00:08 |
| harlowja | 4 that could be say involved in this :-P | 00:08 |
| harlowja | maybe more, unsure yet | 00:08 |
| sdake | harlowja sounds like a good headcount | 00:08 |
| harlowja | ya, depends on how much i can divert them all to help with this | 00:09 |
| harlowja | but its a pain we have to take imho | 00:09 |
| Daviey | (but i also want to move from linuxbridge to ovs, cinder+lvm to ceph and use differnet IP allocation.. so doing inplace will suck) | 00:09 |
| harlowja | so ya, i and another guy are messing around with the deploy stuff (among the other things i do) | 00:09 |
| harlowja | so its slow going :-P | 00:10 |
| harlowja | the kolla-deploy | 00:10 |
| harlowja | and i think exploring the stuffs and thinking about how we go to the next steps | 00:10 |
| harlowja | (also btw this is just for the private GD cloud) | 00:10 |
| harlowja | idk what the public GD cloud will do here (afaik they use the same puppet stuff) | 00:10 |
| harlowja | which if we also stop using puppet stuff, then will they (i should get this answered, ha) | 00:10 |
| *** godleon has joined #openstack-kolla | 00:11 | |
| sdake | since your self supported, I don't see any real problems | 00:11 |
| sdake | in terms of implementation, kolla needs cells and regions then | 00:11 |
| harlowja | maybe monday i can get mdorman (one of the other guys whose involved) to get in here also | 00:11 |
| harlowja | and can u add a kolla-move-from-puppet-to-kolla command | 00:11 |
| harlowja | lol | 00:11 |
| harlowja | thxxx | 00:12 |
| sdake | harlowja ha ha ha | 00:12 |
| sdake | ok | 00:12 |
| sdake | i dont think that is even possible to invent ;) | 00:12 |
| harlowja | lol | 00:12 |
| harlowja | but ya, i think we can take certain hosts out of rotation (api servers); tear them down, then bring them back up with kolla | 00:12 |
| sdake | having a systemtized way of upgrading a brownfield to a kolla deploy would rock | 00:12 |
| harlowja | that will probably be part of the plan | 00:12 |
| sdake | (in terms of documentation) | 00:12 |
| sdake | how many contorl nodes do you have per region | 00:13 |
| harlowja | let me see what i can find | 00:13 |
| Daviey | sdake: I should JFDI to see.. but to be lazy.. Are there any issues with adding and removing servers from different roles.. and leaving mess on each server? Ie, server A /was/ a control and becomes a compute? or server b was storage and becomes control? | 00:13 |
| sdake | Daviey nothing I'e personally tested | 00:14 |
| sdake | I just got a 10 node clsuter last week | 00:14 |
| sdake | and its still not deployed (because of hardware issues) | 00:14 |
| harlowja | so i think 3 api servers per region, then a couple cells servers (for nova) per region, gotta get to u on the db server count, 2 glance servers (from the example i am looking at right now) and compute (and then some logging server?) | 00:14 |
| sdake | Daviey that said, I'd try not to leave a mess ont eh server ;) | 00:14 |
| harlowja | the 3 api servers run a bunch of all the various oepnstack apis | 00:14 |
| Daviey | sdake: As i'm looking to create a small cluster of about 6 and grow it to about 200.. the personality of the config should change as it grows... but i'm worried about leaving mess on the server | 00:15 |
| harlowja | but i think thats the common pattern of what i've seen | 00:15 |
| sdake | well cells and regions together | 00:15 |
| sdake | Daviey i hear ya, we are adding things like add-osd remove-osd add-compute remove-compute in newton | 00:15 |
| sdake | and add-controller remove-controller | 00:16 |
| harlowja | ya when i can get mike in here sdake he can make sure i'm correct on all these numbers :-P | 00:16 |
| harlowja | but its never like 1 of something | 00:16 |
| harlowja | which should make tearing down things easier and replacing them with containers now under kolla | 00:16 |
| Daviey | sdake: Ah cool. Well, i'm currently planning to deploy mitaka (and migrate instances from juno) | 00:16 |
| sdake | Daviey one word of cuation | 00:17 |
| sdake | I have heard nova compute nodes can only communicate one level forward with nova conductor nodes | 00:17 |
| harlowja | i think most regions have celiometer on a 2-4 nodes also | 00:17 |
| sdake | harlowja we lack a ceilometer implementation | 00:17 |
| harlowja | k | 00:17 |
| harlowja | note to self | 00:17 |
| sdake | did you say you are doing an in-place replacement | 00:18 |
| Daviey | sdake: yeah.. i'm looking to experiment with dd'ing from lvm to ceph... scripted.. but still ugly | 00:18 |
| sdake | i was asking harlowja | 00:18 |
| harlowja | :) | 00:18 |
| harlowja | yes, thats the goal | 00:18 |
| sdake | sorry Daviey for confusion | 00:18 |
| Daviey | (sorry, that was a response about one level forward with nova) | 00:19 |
| harlowja | i only do hard things, lol | 00:19 |
| sdake | Daviey oh got it | 00:19 |
| sdake | harlowja I dont think it will be super hard to do an in place replacement | 00:19 |
| Daviey | sdake: I can see you having to setup a counseling clinic for us to discuss issues. :) | 00:20 |
| sdake | but it will take some time nd probablybe entirely manual | 00:20 |
| harlowja | yup | 00:20 |
| sdake | basically you have to get whatever persistent data you have stored on the machines | 00:21 |
| sdake | and stoe them in docker volumes instead | 00:21 |
| harlowja | ya, i'm hoping there isn't to much of that (outside of nova-compute) | 00:21 |
| *** ssurana has quit IRC | 00:21 | |
| sdake | mariadb | 00:21 |
| sdake | elasticsearch | 00:21 |
| *** ssurana has joined #openstack-kolla | 00:21 | |
| sdake | glance | 00:21 |
| Daviey | sdake: So I have a few deployments stuck on icehouse due to the underlying OS being Centos6 (which has py26), and Juno doesn't support py26. I personally feel that upgrading either py or the base OS, then jumping through the openstack releases to get to mitaka to do an inplace migration is more work than setting up a parallel cloud. | 00:22 |
| harlowja | ah, good ole centos6 | 00:22 |
| sdake | Daviey makes sense | 00:22 |
| sdake | harlowja i'd make a record of all the data you ahve in the system | 00:23 |
| sdake | harlowja whta i want to get at though is what you need in terms of implementation | 00:23 |
| Daviey | in /tmp/ | 00:23 |
| harlowja | yup, afaik thats done (or should be already) | 00:23 |
| sdake | so cells/regions | 00:23 |
| harlowja | sdake ya, i think that would be cool | 00:23 |
| sdake | anything else? | 00:23 |
| harlowja | unsure about keepalived, afaik we just use load-balancers but a guy that would know better can probably chime in on monday | 00:24 |
| sdake | harlowja you can turn off keepalived and haproxy and just use your hardware load balancers if you like | 00:24 |
| harlowja | kk | 00:25 |
| harlowja | the other thing that would be nice, is to have a kolla-deploy convert-compute or something | 00:25 |
| harlowja | cause i can imagine that will be the painful one | 00:25 |
| Daviey | sdake: Do you think it is likely that kolla-kubernetes will take over from kolla-ansible? | 00:25 |
| *** ssurana has quit IRC | 00:26 | |
| sdake | Daviey you mean ddisplace ansible ? | 00:26 |
| Daviey | yeah | 00:26 |
| sdake | Daviey unlikely any time soon | 00:26 |
| Daviey | I mean.. am i going to bet on the wrong horse now and regret it in 6 months? | 00:26 |
| sdake | Daviey but if it does, the migraiton path will be smooth | 00:26 |
| Daviey | super | 00:26 |
| harlowja | (i'm thinking that would say take a existing compute installation/hypervisor and without taking it offline ? move it to kolla based? ) | 00:26 |
| sdake | kolla-ansible will take atleast 12 months to make stable | 00:26 |
| sdake | harlowja zero downtime vm migration? | 00:27 |
| harlowja | i may be dreaming, idk | 00:27 |
| harlowja | lol | 00:27 |
| Daviey | That shouldn't be hard if you plug in your new kolla stuff to your old control plane? | 00:27 |
| sdake | i'll have a think but my inital response is dreaming :) | 00:27 |
| sdake | I don't think its possible because of kernel reasons | 00:27 |
| harlowja | ya, not unexpected :-P | 00:28 |
| sdake | harlowja there is always live migration ;-) | 00:28 |
| harlowja | lol | 00:28 |
| Daviey | huh? surely a nova-compute via puppet or nova-compute via kolla should be treated as identical for scheduling / migration reasons? | 00:28 |
| sdake | i know what a charlie foxtrot of a featuer that is | 00:28 |
| Daviey | providing you have the same cpu etc? | 00:28 |
| harlowja | Daviey i think its more about the VMs that are actively running | 00:28 |
| harlowja | the bunches of qemu-kvm stuff running | 00:29 |
| sdake | you can shut down nova-compute and the vms will remin running | 00:29 |
| Daviey | well they are just libvirt kvm machines? | 00:29 |
| sdake | qemu-kvm will be adopted by systemd | 00:29 |
| Daviey | surely it doesn't matter if on the metal or via docker? | 00:29 |
| harlowja | unsure, u guys let me know :-P | 00:29 |
| sdake | the problem is the file that qemu-kvm loads | 00:29 |
| sdake | it loads some files from disk and once they are loaded, they are locked | 00:29 |
| Daviey | i would be really surprised if it was a major headache | 00:29 |
| Daviey | hmm | 00:30 |
| sdake | well the persistent storage for nova compute goes in a named volume | 00:30 |
| sdake | perhaps a mv operation would work | 00:30 |
| Daviey | they aren't really locked.. you can do $ virsh edit instance-0010101 . and change the running vm without telling nova | 00:30 |
| sdake | I am not certain | 00:30 |
| sdake | Daviey i am talking about the files loaded by qemu-kvm | 00:30 |
| harlowja | in the kolla nova-compute, just want to check (since i'm not sure and haven't tried it yet) do the qemu-kvm processes run inside the nova-compute container | 00:30 |
| sdake | the kernel has references to them open | 00:30 |
| harlowja | or do they run outside the container | 00:31 |
| sdake | harlowja we use pid=host | 00:31 |
| sdake | so they run outside the container | 00:31 |
| harlowja | k | 00:31 |
| sdake | but inside a docker named voluem | 00:31 |
| Daviey | hell, you could even just rsync.. run the vm and add a db entry.. but that way leads to a bad day out. | 00:31 |
| sdake | still with all that cgroups black magic | 00:31 |
| sdake | docker cant even stop a fork bomb ;) | 00:31 |
| sdake | seriously though I'm not sure if cgroups will prevent access to moving the file into the odcker named volume | 00:32 |
| sdake | where it needs to go | 00:32 |
| *** ravig has quit IRC | 00:32 | |
| harlowja | i think folks here might be ok with nova-compute not responding to say a terminate request for a little, but don't think people would be happy if we had VM downtime (although there might be a way to make this ok, since I do get emails saying my vms might be inaccessible for a period) | 00:32 |
| sdake | if that part works, that could be entirely automated in ansible in about 1-2 hours by the right person | 00:32 |
| harlowja | ya | 00:32 |
| harlowja | just need to figure out the appropriate steps | 00:32 |
| Daviey | sdake: if you can live mirgate from a docker nova-libvirt to a docker nova-libvirt, i can't see why you couldn't do it from bare metal to docker | 00:33 |
| sdake | Daviey you can do live migration | 00:33 |
| sdake | I think harlowja doesn't want to live migrate | 00:33 |
| Daviey | what a loser. | 00:33 |
| harlowja | just to clarify, this is the nova concept of live-migrate right? | 00:33 |
| sdake | harlowja right | 00:33 |
| harlowja | k | 00:33 |
| sdake | with qemu integration | 00:34 |
| harlowja | right | 00:34 |
| harlowja | ya, don't think that can happen here | 00:34 |
| sdake | i dont eve nwant to know why :) | 00:34 |
| harlowja | not everything running off of ceph .... | 00:34 |
| harlowja | ya | 00:34 |
| harlowja | and various other things :-P | 00:34 |
| sdake | ceph | 00:34 |
| harlowja | that i also don't want to know why, lol | 00:34 |
| sdake | you are using ceph? | 00:34 |
| Daviey | harlowja: why not bring up a kolla control plane... add a kolla compute node, migrate instances from a legacy compute node into it.. and then whn evac'd.. reprovision the old server as a kolla compute then GOTO 10 | 00:35 |
| Daviey | ? | 00:35 |
| harlowja | sdake well ummmm sort of afaik lol | 00:35 |
| sdake | harlowja i highly recommend ceph for all your storage | 00:35 |
| sdake | but you definately can't migrate your nova compute to ceph | 00:35 |
| harlowja | i know i know, i tried talking to the guys here about that .... | 00:35 |
| harlowja | they got ceph-nightmares or something, idk | 00:36 |
| sdake | well if it craters | 00:36 |
| harlowja | Daviey that might work, unsure if the network here will work for that | 00:36 |
| sdake | your fucked | 00:36 |
| sdake | wherea if ext4 craters | 00:36 |
| sdake | you lose one node | 00:36 |
| harlowja | ya, i think they had it crater once, lol | 00:36 |
| harlowja | thats what i hear | 00:36 |
| sdake | lol | 00:36 |
| harlowja | and then they ran away from ceph for all the storage | 00:37 |
| harlowja | so hypervisors afaik use local storage for most things | 00:37 |
| harlowja | which will, really kills live-migration (among other things, ha) | 00:37 |
| Daviey | currently we are using cinder+lvm with the nova compute always using localhost for storage (which gives no migration ability)... i need to prove that ceph will do what we need | 00:37 |
| harlowja | Daviey let me know, maybe i can convince the ops guys here to not have nightmares | 00:37 |
| sdake | harlowja in a kolla model hyeprvisors use named volumes | 00:37 |
| harlowja | lol | 00:37 |
| Daviey | (i know it will... but i need to prove it) | 00:37 |
| harlowja | sdake i guess then named volumes can ---> ceph or ... | 00:38 |
| sdake | harlowja here let me erlog so i can access my serers | 00:38 |
| Daviey | harlowja: If my legacy compute nodes were running a newer version of openstack, that is how i'd do it... but i suspect i'll need to dd disks from lvm to ceph as part of my by-hand migration | 00:38 |
| harlowja | :-/ | 00:39 |
| harlowja | fun fun | 00:39 |
| harlowja | ha | 00:39 |
| Daviey | i'll script it.. but it will still suck | 00:39 |
| harlowja | ya, no doubt | 00:39 |
| Daviey | harlowja: Do you have a timeline you are working towards? | 00:40 |
| harlowja | not so clear yet | 00:40 |
| harlowja | soon? | 00:40 |
| harlowja | lol | 00:40 |
| harlowja | like this year? | 00:40 |
| harlowja | in the next 4 months ? | 00:40 |
| harlowja | lol | 00:40 |
| harlowja | thats about as much as i got | 00:40 |
| harlowja | to many variables to accurately predict timeline :-P | 00:41 |
| Daviey | hah. Well i only started looking at kolla as part of the threat analysis stuff at the summit.. and playing with it 2 weeks ago. I hope to start bringing dev/qa nodes into it in ~6 weeks. | 00:41 |
| Daviey | But i suspect we are at about the same stage right now. | 00:41 |
| harlowja | possibly | 00:41 |
| *** sdake_ has joined #openstack-kolla | 00:41 | |
| harlowja | u might have more free resources that can dedicate to this than i do :-P | 00:41 |
| harlowja | i'm like a fraction of a person, lol | 00:42 |
| harlowja | split across many things, ha | 00:42 |
| sdake_ | who does for what | 00:42 |
| Daviey | I have 42 hour days. | 00:42 |
| harlowja | lol | 00:42 |
| sdake_ | i have 84 hour days | 00:42 |
| *** sdake has quit IRC | 00:42 | |
| sdake_ | harlowja i'mg goint o try to prototype a migration from one continaer to another of nova compute | 00:43 |
| sdake_ | and see if it works in that regard | 00:43 |
| sdake_ | if you can hang for 15 mins i willl show you results | 00:43 |
| Daviey | sdake_: you mean a live migrate of a kolla compute node to another kolla node? | 00:43 |
| sdake_ | Daviey no I am going to use the unix mv ommand to move the files around | 00:44 |
| harlowja | cool | 00:44 |
| sdake_ | and see if that works | 00:44 |
| harlowja | i was just talking to my manager who is still around about the ceph 'problem' | 00:44 |
| harlowja | and why there seems to be 2 camps around it | 00:44 |
| sdake_ | 'problem | 00:44 |
| sdake_ | lol | 00:44 |
| harlowja | the 'its the best thing ever camp' | 00:44 |
| harlowja | and then the 'nightmare camp' | 00:44 |
| harlowja | i'm in no-thoughts-camp | 00:45 |
| harlowja | and we were starting to wonder why this is | 00:45 |
| sdake_ | harlowja my idea wont work | 00:45 |
| harlowja | (cause i don't really know) | 00:45 |
| Daviey | sdake_: well the concern that some people have is latency and flodding the network.. but this is a general shared storage issue and not specific to ceph | 00:45 |
| sdake_ | harlowja you can't use mv across filesystems | 00:45 |
| Daviey | and "it's new, don't know how to debug".. which was the same as ovs vs linuxbridge | 00:46 |
| sdake_ | unix = fired | 00:46 |
| harlowja | Daviey i think it may also be the 'i had it crater on me once and now i have PTSD from that event' | 00:46 |
| Daviey | harlowja: ha | 00:46 |
| harlowja | sdake_ is there anyway to say have the named volume just point to the existing 'folder' that the existing computes already run from | 00:47 |
| harlowja | and then have a few minutes of the libvirt daemon being down? | 00:47 |
| Daviey | sdake_: is this to mv it from the host (from legacy libvirt on host) to one within a kolla/docker container? | 00:47 |
| sdake_ | Daviey actually it gets moved to /var/lib/docker | 00:47 |
| sdake_ | but yup | 00:47 |
| sdake_ | harlowja named volumes are stoerd in /var/lib/docker | 00:48 |
| Daviey | yeah, you can bind mount (not sure what it is called in docker), but you can mirror a folder like that | 00:48 |
| harlowja | is there a way to migrate without nova | 00:48 |
| harlowja | i think so | 00:48 |
| harlowja | like local-to-the-machine-migrate, ha | 00:48 |
| harlowja | from one libvirt to the other | 00:48 |
| harlowja | that should be ok right | 00:48 |
| harlowja | (although seems sorta crappy, ha) | 00:48 |
| sdake_ | ya that might work - depends what version of qemu you have | 00:48 |
| harlowja | centos7 stuffs | 00:49 |
| sdake_ | the prbolem with live migration is lack of a standard binary format for qemu live migration | 00:49 |
| sdake_ | harlowja we use the centos virt sig repos for qemu | 00:49 |
| sdake_ | harlowja which is the latest cutting edge stuff from red hat with some features removed on purpose | 00:49 |
| Daviey | docker run -v /var/lib/nova/instances:/var/lib/docker/nova/instances ... | 00:49 |
| sdake_ | harlowja centos7 typically ships with a really old(tm) version of qemu | 00:50 |
| Daviey | sdake_: AIUI you can only run one instance of libvirt on the entire machine? | 00:50 |
| harlowja | :-p | 00:50 |
| Daviey | I don't think this is a qemu problem, is it? | 00:50 |
| sdake_ | Daviey nah I htink you can run multipel copies of libvirt | 00:50 |
| sdake_ | if you use networking | 00:51 |
| Daviey | sdake_: Oh, i thought i read in the kolla docs that it didn't work | 00:51 |
| sdake_ | what doesn't work | 00:51 |
| sdake_ | oh multiple copies | 00:51 |
| Daviey | libvirt on the metal and libvirt in docker | 00:51 |
| sdake_ | what I mean is it can be done | 00:51 |
| sdake_ | if onfigured properly | 00:51 |
| Daviey | ok | 00:51 |
| harlowja | what about not having the libvirt part running in a container? | 00:51 |
| sdake_ | killing libvirt does not kill qemu-kvm processes | 00:51 |
| sdake_ | harlowja probably your best option is to bindmount libvirt data rather then use a named volume | 00:52 |
| sdake_ | we use a named volume because thats how you do htings in docker now | 00:52 |
| sdake_ | bindmounts are untidy | 00:52 |
| Daviey | yeah, using docker -v HOST:MOUNT, or literally bind mounting outside of docker? | 00:53 |
| sdake_ | but your system will already be untidy since you wont be doing a ffresh install | 00:53 |
| sdake_ | Daviey ya our docker module has an option for that I think - if not its easy enough to add | 00:53 |
| Daviey | (i still think host evacuation from legacy to kolla is cleaner and easier) | 00:53 |
| *** salv-orl_ has joined #openstack-kolla | 00:56 | |
| *** salv-orlando has quit IRC | 00:56 | |
| sdake_ | harlowja changing the ansible like this should work: http://paste.fedoraproject.org/374277/1758146/ -> http://paste.fedoraproject.org/374277/01758146 | 00:56 |
| sdake_ | Daviey i tend to agree - you can do a completely fresh host OS installation | 00:56 |
| harlowja | kk | 00:57 |
| sdake_ | without all the crap on there that was there from openstack 3 versions ago :) | 00:57 |
| sdake_ | harlowja in fact you could get rid of all the named volumes we use and just use bindmounts | 00:57 |
| sdake_ | not all the named volumes | 00:57 |
| harlowja | whats the benefit of doing that? | 00:57 |
| sdake_ | but the onees related to persistent | 00:57 |
| sdake_ | the benefit is you wont have ot migrate your data into the named volume | 00:58 |
| harlowja | k | 00:58 |
| harlowja | ya, for libvirt stuff that might be the way to go | 00:58 |
| sdake_ | ya - only option | 00:58 |
| sdake_ | i'd recommend nvme ssds for /var/lib/docker as well | 00:59 |
| sdake_ | ;) | 00:59 |
| harlowja | might already be that, i'm not sure :-P | 00:59 |
| harlowja | (ssds that is) | 00:59 |
| sdake_ | and i'd recomend btrfs for thir filesystem | 00:59 |
| harlowja | hmmmm | 00:59 |
| harlowja | ya, pretty sure its not that, lol | 00:59 |
| sdake_ | i know, babies eaton | 00:59 |
| Daviey | sdake_: Is IO going to be that much of an issue? | 00:59 |
| sdake_ | Daviey which io where | 00:59 |
| sdake_ | Daviey openstack is all about io:) | 00:59 |
| sdake_ | harlowja btrfs is needed (I thiink) for upgrades | 01:00 |
| Daviey | sdake_: why /var/lib/docker ssd? | 01:00 |
| sdake_ | (to work in the future) | 01:00 |
| harlowja | k, will have to get person who knows more about this than me for that one :-P | 01:00 |
| sdake_ | Daviey oh yes it makes a huge difference | 01:00 |
| sdake_ | my build is 15 minutes, prior to intel 750 1.2tb ssd - my build was 45 minutes - I have gige to the internet | 01:00 |
| Daviey | sdake_: for /everything?!/ | 01:00 |
| sdake_ | Daviey well here is the deal - I put /var/lib/docker on an ssd becuase I work on kolla 18 hrs a day | 01:01 |
| Daviey | well, sure for dev | 01:01 |
| sdake_ | Daviey ifor me the cost benefit analysis was a no brainer | 01:01 |
| harlowja | only 18/84 | 01:01 |
| harlowja | why u not working that much | 01:01 |
| harlowja | lol | 01:01 |
| sdake_ | gotta recreate some time :) | 01:02 |
| sdake_ | harlowja you might want to double check libvirt doesn't shut down qemu-kvm | 01:02 |
| harlowja | yup | 01:02 |
| Daviey | but mariadb, base OS and python code.. all on SSD is going to become a problem with either capacity and/or $$$ | 01:02 |
| sdake_ | you may have to teriminate it -9 | 01:02 |
| harlowja | sdake_ ya, its a good thing to figure out, the whole how is this whole process gonna work (and what are the steps) for nova-compute | 01:03 |
| sdake_ | harlowja ya - and doing the same thing for controller nodes | 01:03 |
| harlowja | ya, i think those are easier, cause those can get shot-in-the-head easier | 01:03 |
| harlowja | less pita, lol | 01:03 |
| sdake_ | harlowja it woudl be helpful for m e to ge ta copy of some configs that describe cells and regions | 01:04 |
| sdake_ | I think this is probably in teh nova.conf files | 01:04 |
| sdake_ | but i am not sure | 01:04 |
| Daviey | harlowja: planning to just add the kolla VIP to your legacy load balancer? So that it serves old and kolla control plance concurrently ? | 01:04 |
| harlowja | ya, i can locate some of it | 01:04 |
| harlowja | Daviey unsure | 01:04 |
| sdake_ | harlowja how many nodes? | 01:05 |
| harlowja | or just use the load-balancer directly? | 01:05 |
| sdake_ | per region | 01:05 |
| Daviey | harlowja: you might not use keepalived? | 01:05 |
| harlowja | sdake_ will get that from someone who knows it better than me :-P | 01:05 |
| sdake_ | cool | 01:05 |
| sdake_ | the reason I ask is because haproxy has a limied amount of capacity | 01:05 |
| sdake_ | I dont nwo what it is | 01:05 |
| sdake_ | i haven't measured it | 01:05 |
| sdake_ | haproxy is what does all the network IO | 01:06 |
| sdake_ | keepalived just places the VIP in the correct controller node | 01:06 |
| sdake_ | with 64 nodes, we had no prolems with any performance | 01:06 |
| harlowja | cool,, ya we have this 'OpenStack Regions/Environments ' confluence page, but its information is not so easy to decode | 01:06 |
| harlowja | i don't think i have the decoder ring for it, lol | 01:06 |
| sdake_ | obviously that isn't 200-500 that a region can have | 01:06 |
| sdake_ | brb relogging | 01:07 |
| Daviey | sdake_: You could look to do layered HAproxy if one doesn't cope. | 01:07 |
| Daviey | but right now... bed. Night chaps.. good talking x | 01:08 |
| harlowja | lata | 01:10 |
| *** sdake_ has quit IRC | 01:11 | |
| *** sdake has joined #openstack-kolla | 01:12 | |
| *** salv-orlando has joined #openstack-kolla | 01:14 | |
| sdake | harlowja i'd recommend setting up a 10 node cluster using your current tooling | 01:15 |
| sdake | harlowja loading it up with a workload | 01:15 |
| sdake | and try to figure out how to migrate it to kolla | 01:15 |
| harlowja | agreed | 01:15 |
| sdake | tiral and error will be best teacher | 01:15 |
| sdake | i know that seems obvious | 01:15 |
| sdake | but it seemed like a flash of birilliance 5 minutes ago ;) | 01:15 |
| harlowja | :-P | 01:15 |
| Daviey | harlowja: One more q for you, are you looking to use puppet still to bring up your hosts.. install docker.. ssh keys etc? | 01:16 |
| harlowja | unsure | 01:16 |
| harlowja | i would guess so? | 01:16 |
| Daviey | sdake: and one more from you.. have you done all-in-one (- compute) and migrated it to multi-host after deployment? | 01:17 |
| *** cu5 has joined #openstack-kolla | 01:17 | |
| sdake | Daviey i have in teh past | 01:17 |
| sdake | not recently | 01:17 |
| sdake | kolla movess fast | 01:17 |
| Daviey | sdake: any issues? | 01:17 |
| sdake | it worked at the time | 01:17 |
| *** salv-orl_ has quit IRC | 01:17 | |
| sdake | recommend validating yourself though | 01:17 |
| Daviey | right. yeah | 01:17 |
| harlowja | sdake monday i'll grab a more operator-puppet-man and hopefully can further flush out some questions/ideas here | 01:18 |
| Daviey | harlowja: I'm expecting to do my kolla-infra via chief. | 01:19 |
| Daviey | chef* | 01:19 |
| harlowja | cool | 01:19 |
| Daviey | (or maybe ansible, tbc) | 01:19 |
| harlowja | ya pick your bootstrapper | 01:19 |
| *** cu5 has quit IRC | 01:20 | |
| Daviey | Our hosts are deployed via chef, but the workload via ansible... the division is, erm, unclear. So i think kolla can sit in either for me | 01:20 |
| *** ChanServ changes topic to "Kolla IRC meetngs on Wednesdays @ 16:30 UTC even weeks, 23:00 UTC odd weeks - see agenda @ https://goo.gl/OXB0DL - IRC channel is *LOGGED* @ http://goo.gl/3mzZ7b (old logs from #kolla http://goo.gl/VKpPzA)" | 01:21 | |
| -openstackstatus- NOTICE: Gerrit maintenance concluded successfully | 01:21 | |
| sdake | fwiw , we are going to make bare metal setup a playbook in newton | 01:23 |
| harlowja | kolla folks really like 'Add a bug id or Trivial Fix' don't they :-P | 01:24 |
| sdake | well trained i'd say :) | 01:24 |
| harlowja | :-P | 01:24 |
| sdake | you can use blueprints for big bodies of work | 01:24 |
| harlowja | josh-random-stuff-blueprint | 01:24 |
| harlowja | lol | 01:24 |
| sdake | i usually rubber stamp blueprints unless they are obnoxious in some way | 01:24 |
| sdake | the reason we need bug ids is for backporting | 01:25 |
| sdake | if yoru patch isn't going to be backported to a stable branch, you can use trivialfix if its not a blueprint | 01:25 |
| sdake | that is the basic training we have in place ;) | 01:25 |
| harlowja | lol | 01:25 |
| harlowja | k | 01:25 |
| sdake | without bugids, the release team would probably scream at me wrt zstreams | 01:26 |
| harlowja | what if in the future it gets backported but i didn't expect it to be backported at the time of creation | 01:26 |
| sdake | tonyb would probably be unhappy | 01:26 |
| harlowja | cause i got no future prediction powers :-P | 01:26 |
| sdake | and i wouldn't be able to execute | 01:26 |
| sdake | harlowja usually the person that reviews it makes a call and says it shoudl be backported | 01:27 |
| sdake | but thi doesn't always happen | 01:27 |
| Daviey | sdake: a backport doesn't need a bug... but it doesn't hurt either. | 01:27 |
| sdake | Daviey it does foru s :) | 01:27 |
| harlowja | i'm just wondering how that person predicts the future | 01:27 |
| harlowja | lol | 01:27 |
| sdake | harlowja if its a bug, high or critical, it gets backported | 01:27 |
| sdake | if its a feature, it doesn't get backported | 01:27 |
| Daviey | sdake: well ok.. but i'm part of stable-maint-core and we've never pushed for it | 01:27 |
| sdake | its simple as that | 01:27 |
| harlowja | k | 01:27 |
| sdake | Daviey ok then you would know ;) | 01:28 |
| harlowja | what if its a feature that turns into a fix for a bug by accident | 01:28 |
| harlowja | lol | 01:28 |
| sdake | Daviey my bad, didn't know you had experience in the area | 01:28 |
| harlowja | like the feature turns out to fix all the bugs, via it being the best feature ever | 01:28 |
| sdake | harlowja then it becomes very difficult to track, but we can still backport it | 01:28 |
| harlowja | :) | 01:28 |
| sdake | ya we are doing that with lbaas | 01:29 |
| sdake | because magnum depends on it | 01:29 |
| sdake | adn we shipped mitaka and liberty with magnum | 01:29 |
| sdake | but it was busted | 01:29 |
| sdake | lbaas is cleary a feature | 01:29 |
| *** thumpba has joined #openstack-kolla | 01:29 | |
| sdake | but in this case its a bug because it isn't present | 01:29 |
| sdake | and magnum depends on it | 01:29 |
| sdake | and magnum is busted without it | 01:29 |
| sdake | (and without other changes as well) | 01:29 |
| harlowja | gotcha | 01:29 |
| sdake | what we should have done is removed magnum form the repo | 01:29 |
| sdake | and then reverted it after the release | 01:29 |
| sdake | but i didn't realize how big of a problem it was | 01:30 |
| sdake | but we are stuck with it now | 01:30 |
| harlowja | ya, sounds similar to the other project that i had that turned source repos -> rpms that then get deployed (via this puppet stuff) | 01:30 |
| harlowja | it became a weird tangle of figuring out, oh well now u need that other project for this project to work also | 01:30 |
| harlowja | thus kolla will now figure all that out (instead of me, lol) | 01:31 |
| sdake | well kolla relies on the upstream (in source biulds0 requirements and upper constraints to be correct | 01:31 |
| harlowja | https://github.com/openstack/anvil/tree/master/conf/personas/in-a-box (for example, these listed the pieces needed to work for a various set of things) | 01:31 |
| sdake | if they are not, all bets are off ;) | 01:31 |
| harlowja | ya, stuff just gets weird when pieces are somewhat optional | 01:32 |
| *** openstackgerrit has quit IRC | 01:33 | |
| *** openstackgerrit has joined #openstack-kolla | 01:33 | |
| harlowja | anvil also pre-existed the whole requirements and upper-constraints stuff so i had to years ago put a variant (crappy one) of a find-the-constraint set that will hopefully work in it | 01:34 |
| harlowja | u don't want to know, lol | 01:34 |
| *** thumpba has quit IRC | 01:35 | |
| sdake | havne'et heard of anvil but i looked ath the repo | 01:54 |
| sdake | 3k commits | 01:54 |
| sdake | seems pretty active | 01:54 |
| *** Jeffrey4l_ has quit IRC | 02:01 | |
| sdake | harlowja re cells | 02:05 |
| sdake | harlowja i spoke with russellb today about cells and regions | 02:05 |
| sdake | (as scalability options) | 02:05 |
| sdake | he told me that you install amqp+db for each cell | 02:06 |
| sdake | he also told me you install all controllers per region | 02:06 |
| sdake | and then make one database for keystone, and use fernet tokens | 02:06 |
| sdake | with regions, we should be able to scale to several thousand nodes | 02:07 |
| sdake | not sure how cells is helpful but if thats wht you have today then I dont see harm in implementing it and other will likely want it as well | 02:07 |
| sdake | is this your understanding of how these deployments are structured? | 02:07 |
| sdake | cells will be harder then regions | 02:08 |
| sdake | since it sound slie the only new dev you reallly require is cells plus regions, we should be able to get this done this cycle | 02:09 |
| sdake | we have 130 node OSIC cluster coming online | 02:09 |
| sdake | we are going to deploy kolla on it | 02:09 |
| sdake | then we are going to deploy kolla on top of vms produced by that kolla | 02:09 |
| sdake | to give us a couple thousand nodes to work with | 02:09 |
| sdake | and do the development that way | 02:09 |
| sdake | note this layering approach is JUST for dev | 02:10 |
| sdake | so we can scale out our 130 nodes to something that is real | 02:11 |
| *** papacz has quit IRC | 02:11 | |
| *** Jeffrey4l_ has joined #openstack-kolla | 02:13 | |
| sdake | Jeffrey4l_ ping | 02:18 |
| Jeffrey4l_ | sdake, pong | 02:19 |
| sdake | Jeffrey4l_ did you see the discussion i just hd with harlowja | 02:19 |
| sdake | Jeffrey4l_ if not check the logs plz | 02:19 |
| sdake | it involved godaddy's requirements for kolla deployment | 02:20 |
| sdake | godaddy is larger then our typical deployment by several orders of magnitude | 02:20 |
| Jeffrey4l_ | just saw this. I will check the whole log. | 02:21 |
| sdake | Jeffrey4l_ you probably wont be able to make the midcycle in july 12-13th right? | 02:21 |
| Jeffrey4l_ | sdake, yes. :( | 02:21 |
| sdake | Jeffrey4l_ can you make the main summit? | 02:21 |
| sdake | it is in barcelona iirc | 02:22 |
| Jeffrey4l_ | possible, if i have no passport issue. | 02:22 |
| sdake | can you get cracking on the paperwork | 02:22 |
| sdake | if you need a letter from the foundation for a visa they will provide iirc | 02:23 |
| sdake | that way the govt knows your doing your day job and not moving countries :) | 02:23 |
| Jeffrey4l_ | yep. I applied the letter. | 02:23 |
| sdake | nice already on the ball | 02:23 |
| sdake | chinese government aware of openstack as a thing | 02:24 |
| Jeffrey4l_ | but still failed when applying the US visa. :( | 02:24 |
| sdake | so if a letter comes from the foundation it should help your chances | 02:24 |
| sdake | europe isn't as rigid on visas as the US | 02:24 |
| sdake | the letter helps | 02:25 |
| sdake | i needed one to get into china | 02:25 |
| sdake | i needed two in fact | 02:26 |
| sdake | and every time i walked into the us they looked at me funny when they looked through my passport :) | 02:26 |
| Jeffrey4l_ | lol | 02:26 |
| Jeffrey4l_ | i will well prepared for the newton summit. | 02:27 |
| sdake | fortunately I have a fres hclean passport to damage ;) | 02:27 |
| sdake | Jeffrey4l_ we dont really prepare all that much except as a group for planning sessions | 02:27 |
| sdake | Jeffrey4l_ but do whatever makes your comfortable :) | 02:28 |
| Jeffrey4l_ | OK. | 02:28 |
| *** openstackgerrit has quit IRC | 02:32 | |
| *** openstackgerrit has joined #openstack-kolla | 02:33 | |
| *** daneyon_ has joined #openstack-kolla | 02:34 | |
| openstackgerrit | Merged openstack/kolla: adding else statement with empty string starting ansible 2.X it require a else statement. https://review.openstack.org/325483 | 02:39 |
| *** daneyon_ has quit IRC | 02:40 | |
| *** ravig has joined #openstack-kolla | 02:58 | |
| *** ravig has quit IRC | 03:00 | |
| *** eaguilar has joined #openstack-kolla | 03:04 | |
| *** MarMat has joined #openstack-kolla | 03:08 | |
| *** eaguilar has quit IRC | 03:11 | |
| *** ravig has joined #openstack-kolla | 03:11 | |
| *** ravig has quit IRC | 03:19 | |
| *** ravig has joined #openstack-kolla | 03:20 | |
| *** sacharya has joined #openstack-kolla | 03:22 | |
| *** sacharya_ has quit IRC | 03:26 | |
| *** ravig has quit IRC | 03:31 | |
| *** ravig has joined #openstack-kolla | 03:31 | |
| *** ravig has quit IRC | 03:35 | |
| openstackgerrit | Swapnil Kulkarni (coolsvap) proposed openstack/kolla: Make container kolla-toolbox build with debian https://review.openstack.org/297104 | 03:44 |
| openstackgerrit | Swapnil Kulkarni (coolsvap) proposed openstack/kolla: Make swift-base container build with debian https://review.openstack.org/297121 | 03:44 |
| *** sacharya_ has joined #openstack-kolla | 03:48 | |
| *** sacharya has quit IRC | 03:51 | |
| *** ravig has joined #openstack-kolla | 04:24 | |
| openstackgerrit | Joshua Harlow proposed openstack/kolla: Delay opening a log file until first use https://review.openstack.org/322342 | 04:29 |
| *** sacharya has joined #openstack-kolla | 04:32 | |
| *** sacharya_ has quit IRC | 04:36 | |
| *** thumpba has joined #openstack-kolla | 05:03 | |
| openstackgerrit | Swapnil Kulkarni (coolsvap) proposed openstack/kolla: Make heka container build with debian https://review.openstack.org/297737 | 05:04 |
| openstackgerrit | Merged openstack/kolla: Make container kolla-toolbox build with debian https://review.openstack.org/297104 | 05:07 |
| openstackgerrit | Merged openstack/kolla: Make swift-base container build with debian https://review.openstack.org/297121 | 05:07 |
| *** thumpba has quit IRC | 05:08 | |
| openstackgerrit | Joshua Harlow proposed openstack/kolla: Have more useful stdout with a progressbar (as well) https://review.openstack.org/325529 | 05:08 |
| *** sacharya has quit IRC | 05:13 | |
| *** sacharya has joined #openstack-kolla | 05:15 | |
| *** daneyon_ has joined #openstack-kolla | 05:16 | |
| *** sacharya has quit IRC | 05:19 | |
| *** daneyon_ has quit IRC | 05:22 | |
| *** ravig has quit IRC | 05:36 | |
| *** sacharya has joined #openstack-kolla | 06:16 | |
| *** sacharya has quit IRC | 06:20 | |
| *** thumpba has joined #openstack-kolla | 07:00 | |
| *** thumpba has quit IRC | 07:05 | |
| *** salv-orl_ has joined #openstack-kolla | 07:15 | |
| *** salv-orlando has quit IRC | 07:18 | |
| *** fragatina has quit IRC | 07:26 | |
| *** MarMat has quit IRC | 07:35 | |
| *** fragatina has joined #openstack-kolla | 07:39 | |
| *** fragatina has quit IRC | 07:43 | |
| *** fragatina has joined #openstack-kolla | 07:45 | |
| *** fragatin_ has joined #openstack-kolla | 07:46 | |
| *** fragatina has quit IRC | 07:49 | |
| *** sdake has quit IRC | 08:07 | |
| openstackgerrit | Benedikt Trefzer proposed openstack/kolla: Make container dind unpin old docker relase https://review.openstack.org/297434 | 08:10 |
| openstackgerrit | Benedikt Trefzer proposed openstack/kolla: Make heka container build with debian https://review.openstack.org/297737 | 08:12 |
| *** fragatin_ has quit IRC | 08:12 | |
| *** sacharya has joined #openstack-kolla | 08:16 | |
| openstackgerrit | Benedikt Trefzer proposed openstack/kolla: Make nova-libvirt container build with debian https://review.openstack.org/297169 | 08:16 |
| openstackgerrit | Benedikt Trefzer proposed openstack/kolla: make rabbitmq container build with debian https://review.openstack.org/297728 | 08:17 |
| *** sacharya has quit IRC | 08:21 | |
| *** b_bezak has joined #openstack-kolla | 08:22 | |
| *** salv-orl_ has quit IRC | 08:23 | |
| *** salv-orlando has joined #openstack-kolla | 08:23 | |
| *** daneyon_ has joined #openstack-kolla | 08:53 | |
| *** salv-orlando has quit IRC | 08:54 | |
| *** salv-orlando has joined #openstack-kolla | 08:55 | |
| *** daneyon_ has quit IRC | 08:58 | |
| *** thumpba has joined #openstack-kolla | 09:01 | |
| *** thumpba has quit IRC | 09:06 | |
| *** dwalsh has joined #openstack-kolla | 09:11 | |
| *** dgonzalez has quit IRC | 09:37 | |
| *** mdnadeem has quit IRC | 09:48 | |
| *** mdnadeem has joined #openstack-kolla | 09:49 | |
| *** mdnadeem has quit IRC | 09:50 | |
| *** mdnadeem has joined #openstack-kolla | 09:51 | |
| *** dgonzalez has joined #openstack-kolla | 09:53 | |
| *** dwalsh has quit IRC | 10:07 | |
| *** godleon has quit IRC | 10:09 | |
| *** pbourke has quit IRC | 10:10 | |
| *** pbourke has joined #openstack-kolla | 10:11 | |
| *** SiRiuS has joined #openstack-kolla | 10:17 | |
| *** sacharya has joined #openstack-kolla | 10:17 | |
| *** sacharya has quit IRC | 10:22 | |
| *** SiRiuS has quit IRC | 10:35 | |
| *** daneyon has quit IRC | 11:03 | |
| *** SiRiuS has joined #openstack-kolla | 11:03 | |
| *** daneyon has joined #openstack-kolla | 11:36 | |
| *** daneyon has quit IRC | 11:40 | |
| *** sacharya has joined #openstack-kolla | 12:18 | |
| *** sacharya has quit IRC | 12:22 | |
| *** dave-mccowan has joined #openstack-kolla | 13:13 | |
| *** b_bezak has quit IRC | 13:14 | |
| *** salv-orl_ has joined #openstack-kolla | 13:15 | |
| *** salv-orlando has quit IRC | 13:18 | |
| openstackgerrit | Merged openstack/kolla: make rabbitmq container build with debian https://review.openstack.org/297728 | 13:36 |
| *** ravig has joined #openstack-kolla | 13:36 | |
| openstackgerrit | Merged openstack/kolla: Make nova-libvirt container build with debian https://review.openstack.org/297169 | 13:37 |
| openstackgerrit | Merged openstack/kolla: Make heka container build with debian https://review.openstack.org/297737 | 13:39 |
| *** mdnadeem_home has joined #openstack-kolla | 14:18 | |
| *** sacharya has joined #openstack-kolla | 14:18 | |
| Mech422 | Hmm - it appears Kolla's ceph stuff gets confused if you use XFS, and tries to mount the partition with a btrfs 'compress' option? | 14:20 |
| *** rhallisey has joined #openstack-kolla | 14:21 | |
| *** sacharya has quit IRC | 14:23 | |
| Mech422 | ahh - my bad... knob in globals needed turning :-P | 14:25 |
| *** mdnadeem_home has quit IRC | 14:27 | |
| *** mdnadeem_home has joined #openstack-kolla | 14:29 | |
| *** mdnadeem_home has quit IRC | 15:03 | |
| *** thumpba has joined #openstack-kolla | 15:04 | |
| *** sacharya has joined #openstack-kolla | 15:04 | |
| *** thumpba has quit IRC | 15:09 | |
| *** daneyon has joined #openstack-kolla | 15:12 | |
| *** daneyon has quit IRC | 15:17 | |
| *** mdnadeem_home has joined #openstack-kolla | 15:19 | |
| *** inc0 has joined #openstack-kolla | 15:34 | |
| inc0 | hey | 15:34 |
| *** sacharya has quit IRC | 15:40 | |
| *** inc0 has quit IRC | 15:45 | |
| openstackgerrit | Ryan Hallisey proposed openstack/kolla: Spec: Deploy Kolla images using Kubernetes https://review.openstack.org/304182 | 15:51 |
| *** sdake has joined #openstack-kolla | 16:02 | |
| *** mdnadeem_home has quit IRC | 16:02 | |
| *** esimone has quit IRC | 16:03 | |
| *** sacharya has joined #openstack-kolla | 16:08 | |
| *** sacharya has quit IRC | 16:13 | |
| *** mdnadeem_home has joined #openstack-kolla | 16:20 | |
| *** jmccarthy has quit IRC | 16:25 | |
| *** jmccarthy has joined #openstack-kolla | 16:25 | |
| openstackgerrit | Christian Berendt proposed openstack/kolla: Fix URL to Heka documentation in README file https://review.openstack.org/325595 | 16:37 |
| *** mdnadeem_home has quit IRC | 16:38 | |
| *** rhallisey has quit IRC | 16:50 | |
| *** thumpba has joined #openstack-kolla | 17:04 | |
| *** thumpba has quit IRC | 17:09 | |
| *** sacharya has joined #openstack-kolla | 17:09 | |
| *** sacharya has quit IRC | 17:14 | |
| *** ravig has quit IRC | 17:16 | |
| *** dwalsh has joined #openstack-kolla | 17:19 | |
| openstackgerrit | Christian Berendt proposed openstack/kolla: Cleanup help string of install_type parameter https://review.openstack.org/325600 | 17:24 |
| *** dwalsh has quit IRC | 17:48 | |
| *** ravig has joined #openstack-kolla | 17:49 | |
| *** ravig has quit IRC | 17:54 | |
| *** ssurana has joined #openstack-kolla | 17:55 | |
| Mech422 | so fresh source build/deploy of stable/mitaka using PARTIONS for ceph instead whole devices 'works' modulo 2 bugs: you need to run kolla-ansible deploy twice, and the OSDs get detected twice.. | 17:55 |
| Mech422 | so with 3 osds, you end up in 'WARN' state with 3/6 OSDs up | 17:56 |
| Mech422 | investigating further now | 17:56 |
| *** Jeffrey4l_ has quit IRC | 17:58 | |
| *** sacharya has joined #openstack-kolla | 18:10 | |
| *** ssurana has quit IRC | 18:10 | |
| Mech422 | it appears ./ansible/roles/ceph/tasks/start_osds.yml is looking for a 'changed' status in name: Looking up OSDs for Ceph (first play) | 18:12 |
| *** ravig has joined #openstack-kolla | 18:12 | |
| Mech422 | can't see where its getting it from though | 18:12 |
| *** sacharya has quit IRC | 18:15 | |
| *** sdake has quit IRC | 18:31 | |
| *** thumpba has joined #openstack-kolla | 18:34 | |
| *** ssurana has joined #openstack-kolla | 18:36 | |
| *** thumpba has quit IRC | 18:38 | |
| *** daneyon has joined #openstack-kolla | 18:49 | |
| *** daneyon has quit IRC | 18:53 | |
| *** fragatina has joined #openstack-kolla | 18:57 | |
| *** ravig has quit IRC | 19:04 | |
| *** ravig has joined #openstack-kolla | 19:05 | |
| *** salv-orlando has joined #openstack-kolla | 19:15 | |
| *** ssurana has quit IRC | 19:18 | |
| *** salv-orl_ has quit IRC | 19:18 | |
| *** fragatina has quit IRC | 19:29 | |
| *** fragatina has joined #openstack-kolla | 19:29 | |
| *** sacharya has joined #openstack-kolla | 19:34 | |
| *** godleon has joined #openstack-kolla | 19:35 | |
| *** ssurana has joined #openstack-kolla | 19:39 | |
| *** sacharya has quit IRC | 19:39 | |
| *** d_code has joined #openstack-kolla | 19:40 | |
| *** d_code has quit IRC | 19:44 | |
| *** sdake has joined #openstack-kolla | 19:45 | |
| *** sdake_ has joined #openstack-kolla | 19:54 | |
| *** sdake has quit IRC | 19:55 | |
| *** ssurana has quit IRC | 19:56 | |
| *** dwalsh has joined #openstack-kolla | 20:01 | |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Convert MariaDB to work without HostNetwork=True https://review.openstack.org/325613 | 20:06 |
| openstackgerrit | Ken Wronkiewicz proposed openstack/kolla-kubernetes: Convert MariaDB to work without HostNetwork=True https://review.openstack.org/325613 | 20:07 |
| *** fragatina has quit IRC | 20:10 | |
| *** dwalsh has quit IRC | 20:11 | |
| *** ssurana has joined #openstack-kolla | 20:15 | |
| *** sdake_ has quit IRC | 20:24 | |
| openstackgerrit | Merged openstack/kolla-kubernetes: Add bootstrap method to the quickstart doc https://review.openstack.org/324113 | 20:25 |
| openstackgerrit | Merged openstack/kolla-kubernetes: Break deployment procedure into two steps https://review.openstack.org/320668 | 20:28 |
| *** fragatina has joined #openstack-kolla | 20:28 | |
| *** sacharya has joined #openstack-kolla | 20:40 | |
| *** ssurana has quit IRC | 20:41 | |
| *** cu5 has joined #openstack-kolla | 20:44 | |
| *** ssurana has joined #openstack-kolla | 20:45 | |
| *** ravig has quit IRC | 20:47 | |
| *** ssurana has quit IRC | 20:48 | |
| *** cu5_ has joined #openstack-kolla | 20:50 | |
| *** cu5_ has quit IRC | 20:52 | |
| *** ravig has joined #openstack-kolla | 20:52 | |
| *** cu5 has quit IRC | 20:55 | |
| *** cu5 has joined #openstack-kolla | 20:58 | |
| *** cu5 has quit IRC | 21:01 | |
| *** sdake has joined #openstack-kolla | 21:27 | |
| *** absubram has quit IRC | 21:38 | |
| *** sdake has quit IRC | 21:49 | |
| *** Jeffrey4l_ has joined #openstack-kolla | 22:30 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Install keepalived in neutron_l3_agent container https://review.openstack.org/325618 | 22:36 |
| *** Jeffrey4l_ has quit IRC | 22:42 | |
| openstackgerrit | Merged openstack/kolla-kubernetes: Add some hints for working with MariaDB. https://review.openstack.org/324812 | 22:48 |
| *** SiRiuS has quit IRC | 22:57 | |
| *** ravig has quit IRC | 23:06 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!