| javeriak | hey anyone notice that little microsoft help paperclip on the openstack pages today | 00:05 |
|---|---|---|
| javeriak | i guess its for april fools | 00:06 |
| *** hj-hp has quit IRC | 00:14 | |
| *** hj-hp has joined #openstack-ansible | 00:20 | |
| *** sdake has quit IRC | 00:32 | |
| *** Mudpuppy_ has quit IRC | 00:32 | |
| *** Mudpuppy has joined #openstack-ansible | 00:33 | |
| *** hj-hp has quit IRC | 00:33 | |
| *** Mudpuppy has quit IRC | 00:38 | |
| *** hj-hp has joined #openstack-ansible | 00:45 | |
| *** hj-hp has quit IRC | 00:47 | |
| *** nosleep77 has quit IRC | 00:50 | |
| cloudnull | Javeriak its the return of clippie. =) | 00:51 |
| stevelle | I'm quite ready for it to go away already | 00:51 |
| cloudnull | +1 | 00:55 |
| javeriak | .... why would they do that... | 01:00 |
| *** vmtrooper has joined #openstack-ansible | 01:02 | |
| *** vmtrooper has quit IRC | 01:07 | |
| *** vmtrooper has joined #openstack-ansible | 01:07 | |
| *** stevemar has joined #openstack-ansible | 01:16 | |
| *** logan2 has quit IRC | 01:19 | |
| *** logan2 has joined #openstack-ansible | 01:21 | |
| palendae | javeriak: April Fool's fun I guess | 01:27 |
| *** vmtrooper has quit IRC | 01:56 | |
| *** vmtrooper has joined #openstack-ansible | 01:57 | |
| *** vmtrooper has quit IRC | 02:01 | |
| *** sdake has joined #openstack-ansible | 02:16 | |
| *** sdake has quit IRC | 02:17 | |
| *** sdake has joined #openstack-ansible | 02:21 | |
| *** darrenc is now known as darrenc_afk | 02:42 | |
| *** darrenc_afk is now known as darrenc | 02:59 | |
| *** Mudpuppy has joined #openstack-ansible | 03:05 | |
| *** sdake has quit IRC | 03:10 | |
| *** Mudpuppy has quit IRC | 03:10 | |
| *** vmtrooper has joined #openstack-ansible | 03:45 | |
| *** vmtrooper has quit IRC | 03:50 | |
| *** mahito has quit IRC | 03:58 | |
| *** mahito has joined #openstack-ansible | 04:05 | |
| *** javeriak has quit IRC | 04:06 | |
| *** javeriak has joined #openstack-ansible | 04:07 | |
| *** javeriak has quit IRC | 04:09 | |
| *** fawadkhaliq has joined #openstack-ansible | 04:11 | |
| prometheanfire | heh, nice | 04:33 |
| *** mahito has quit IRC | 05:03 | |
| *** mahito has joined #openstack-ansible | 05:08 | |
| *** javeriak has joined #openstack-ansible | 05:10 | |
| *** vmtrooper has joined #openstack-ansible | 05:34 | |
| *** vmtrooper has quit IRC | 05:39 | |
| *** fawadkhaliq has quit IRC | 05:42 | |
| svg | Good morning; I'd like to start looking to the proper way to implement cephsupport as an optional config in rpc (only on the openstack side, not the ceph cluster implementation itself); i'd like to discuss teh different things that need to be done, and the right wat to add this, when someone has time :) | 05:56 |
| *** fawadkhaliq has joined #openstack-ansible | 06:00 | |
| *** fawadk has joined #openstack-ansible | 06:05 | |
| *** fawadkhaliq has quit IRC | 06:09 | |
| *** ishant has joined #openstack-ansible | 06:13 | |
| *** fawadkhaliq has joined #openstack-ansible | 06:16 | |
| *** fawadk has quit IRC | 06:19 | |
| *** javeriak has quit IRC | 07:02 | |
| *** javeriak has joined #openstack-ansible | 07:04 | |
| *** javeriak has quit IRC | 07:13 | |
| *** vmtrooper has joined #openstack-ansible | 07:23 | |
| *** vmtrooper has quit IRC | 07:28 | |
| *** stevemar has quit IRC | 07:30 | |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Heat config changes for kilo https://review.openstack.org/167947 | 07:32 |
| *** markvoelker has quit IRC | 08:18 | |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Fix bug in playbooks/library/neutron https://review.openstack.org/170046 | 08:45 |
| *** markvoelker has joined #openstack-ansible | 08:49 | |
| *** markvoelker has quit IRC | 08:55 | |
| *** vmtrooper has joined #openstack-ansible | 09:12 | |
| *** vmtrooper has quit IRC | 09:17 | |
| openstackgerrit | git-harry proposed stackforge/os-ansible-deployment: Add network.yml monitoring playbook https://review.openstack.org/170062 | 09:54 |
| *** ishant has quit IRC | 10:28 | |
| *** ishant has joined #openstack-ansible | 10:29 | |
| *** mahito has quit IRC | 10:30 | |
| *** fawadkhaliq has quit IRC | 10:45 | |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Fix bug in playbooks/library/neutron https://review.openstack.org/170046 | 10:48 |
| *** markvoelker has joined #openstack-ansible | 10:52 | |
| *** markvoelker has quit IRC | 10:57 | |
| *** vmtrooper has joined #openstack-ansible | 11:00 | |
| *** vmtrooper has quit IRC | 11:06 | |
| *** fawadkhaliq has joined #openstack-ansible | 11:46 | |
| *** fawadkhaliq has quit IRC | 11:51 | |
| *** markvoelker has joined #openstack-ansible | 11:53 | |
| *** galstrom_zzz is now known as galstrom | 11:58 | |
| *** markvoelker has quit IRC | 11:58 | |
| *** subscope has quit IRC | 11:59 | |
| svg | ping ^^ | 12:09 |
| andymccr | hey svg | 12:13 |
| svg | I'd like to start looking to the proper way to implement cephsupport as an optional config in rpc (only on the openstack side, not the ceph cluster implementation itself); i'd like to discuss teh different things that need to be done, and the right way to do that :) | 12:14 |
| svg | unrelated: overview of the inventory in treeview: http://sprunge.us/gXfA | 12:14 |
| *** subscope has joined #openstack-ansible | 12:15 | |
| *** sdake has joined #openstack-ansible | 12:15 | |
| andymccr | svg: i think what needs to be done to add ceph support would be to adjust the glance/nova/keystone/cinder confs to allow the appropriate settings to be set as part of the anisble run | 12:16 |
| svg | yes, editing the templtes, is the easiest thing I guess | 12:17 |
| svg | I'm still trying to understand how the inventory works etc | 12:17 |
| andymccr | yeh i think adding support wouldnt be too complicated. | 12:17 |
| andymccr | the inventory as in the openstack_inventory.json? | 12:17 |
| svg | afaics, the wouldn't be any extra | 12:17 |
| svg | afaics, the wouldn't be any extra 'ceph" containers | 12:17 |
| andymccr | i dont think you'd need any extra inventory because ansible wouldnt be managing ceph itself | 12:18 |
| andymccr | just how the openstack services interact with ceph | 12:18 |
| svg | andymccr: yes as per the json | 12:18 |
| andymccr | a brief overview is that the inventory is created by taking the openstack_user_config.yml and the openstack_environment file which are "mixed" together to create the inventory (using the dynamic_inventory.py script) | 12:19 |
| svg | well, given that certain config files that need extra parameters for ceph are managed by this project, I'd expect this neeed to be integrated | 12:19 |
| svg | so the files you mentioned, would need extra options for ceph | 12:19 |
| andymccr | svg: thats correct, but the inventory can stay the same, id anticipate the changes would be in "user_variables.yml" and the various defaults/template files for the projects | 12:19 |
| andymccr | so user_variables is generally where conf options are set. the user_config is usually where hosts and host config are defined for the openstack services | 12:20 |
| svg | Then I'd need to write some roles for that, and those need to be mapped on the right containers | 12:20 |
| andymccr | svg: im not too sure you'd need to do that | 12:20 |
| andymccr | since its all just changes to glance/keystone/nova/cinder settings that already exist | 12:20 |
| andymccr | you just need to ensure that those values can be set | 12:21 |
| andymccr | you wouldnt be installing any new services | 12:21 |
| andymccr | unless there is more than just conf options changes | 12:21 |
| andymccr | in which case you might need additional roles | 12:21 |
| andymccr | but the way the roles are setup there is 1 role per service, so you would more likely need new "tasks" within an existing role (for example os-cinder role) | 12:21 |
| svg | ok, as far as those files are concerned you are right, but I also need to deploy /etc/ceph/ceph.conf, deploy authentication files and do some configuration in libvirt | 12:22 |
| andymccr | if ceph isn't managed by ansible itself, would ceph.conf fall within that? | 12:23 |
| svg | also, ceph needs some packages, repferably from the upstream apt repo, which depends on cephs version - needs to be configureable - so I'd need to add repo's too | 12:24 |
| andymccr | hmm] | 12:24 |
| andymccr | ok that is more complicated than i anticipated | 12:24 |
| *** sdake_ has joined #openstack-ansible | 12:25 | |
| svg | I'm not talking about ,maging and deploying the ceph custer, only configuring the openstack components top use the cluster | 12:25 |
| svg | ceph.conf is the exact same file that gets deployed on all ceph nodes | 12:25 |
| svg | So, as ceoph would be optional, just the same way swift is, I'd expect to provide a similar implementation | 12:26 |
| andymccr | ahh i see | 12:26 |
| andymccr | reading through the ceph docs quickly | 12:26 |
| andymccr | so yes it sounds like perhaps a "ceph_client" role would be needed | 12:26 |
| andymccr | and perhaps a separate play to "add ceph support" | 12:27 |
| svg | theoretically, best doc would be http://ceph.com/docs/master/rbd/rbd-openstack/ | 12:27 |
| andymccr | which would then run the ceph_client role | 12:27 |
| svg | theoretically, I understand openstack can have multiple backends (or at least glance?) | 12:28 |
| *** sdake has quit IRC | 12:28 | |
| andymccr | cinder can also afaik | 12:28 |
| *** markvoelker has joined #openstack-ansible | 12:29 | |
| svg | right now the template for glance checks the default_store var, and implemenst options for file OR swift | 12:29 |
| svg | AFAICS, swift gets triggered by defining storage hosts in openstack_user_config.yml I believe | 12:30 |
| andymccr | installing swift does yes, but i think the var for glance is still set manually | 12:30 |
| svg | hm, I see a var glance_default_store that gets tested if == swift, but nowhere that is being set - so that must be the manual option | 12:31 |
| andymccr | yeh it defaults to "file" | 12:32 |
| andymccr | in the defaults/main.yml for os_glance | 12:32 |
| andymccr | so if you want to use swift as the backend you need to manually specify it in user_variables.yml | 12:32 |
| svg | i found it in user_variables.ym | 12:33 |
| svg | yup | 12:34 |
| andymccr | and then you need to specify a whole bunch of other settings - e.g. it allows you to use a swift deploy that isn't managed by ansible | 12:34 |
| andymccr | (auth url, user/tenant etc) | 12:34 |
| cloudnull | Morning. | 12:36 |
| andymccr | morning cloudnull! | 12:37 |
| andymccr | feel free to chime in if im missing something on the ceph stuff above cloudnull ;D | 12:37 |
| cloudnull | How's it andymccr? | 12:37 |
| * cloudnull reading | 12:37 | |
| andymccr | not bad not bad! how about you? long weekend inc, so all is good | 12:37 |
| svg | hi cloudnull | 12:37 |
| cloudnull | Hi svg | 12:38 |
| cloudnull | Long weekend is where its at ! | 12:38 |
| andymccr | right!?! | 12:38 |
| cloudnull | Andymccr everything you've said so far seems spot on. | 12:39 |
| andymccr | excellent! Yeah its pretty doable, but we may need a "client" role in conjunction with variable setup in ansible | 12:40 |
| cloudnull | A client role is probably the best approach for the ceph specific bits and then add rbd cars to the various services. | 12:40 |
| cloudnull | *vars | 12:40 |
| svg | so i see there is also cinder_backends to set up the cinder specifics | 12:41 |
| andymccr | svg: yeh that'd have to be adjusted in the cinder role to either make the setting up of those backends more generic or in such a way that ceph could be one of those backends | 12:41 |
| svg | I'm a bit cofused seeing some ansible stuff that checks a certain key within cinder_backends, but that key does not exist in all examples | 12:49 |
| *** vmtrooper has joined #openstack-ansible | 12:49 | |
| svg | e.g cinder_backends[item.0]['volume_backend_name'] | 12:50 |
| andymccr | svg: so the cinder_backends are defined in the user_config for the storage hosts | 12:51 |
| andymccr | i think the example conf shows lvm settings for the lvm backend | 12:52 |
| svg | in rpc_deployment/inventory/group_vars/cinder_all.yml it's 'backend_name' (perhaps that is a typo) | 12:53 |
| svg | and also, the user examples have key limit_container_types: cinder_volume; unless I'm missing something, the with_items: cinder_backends|dictsort will provide an item without volume_backend_name | 12:54 |
| *** vmtrooper has quit IRC | 12:54 | |
| svg | I would expect the last task in rpc_deployment/roles/cinder_device_add/tasks/device_add.yml to fail on that? | 12:55 |
| andymccr | i think thats gone in master at least. but yeh that looks like a typo | 12:55 |
| andymccr | looks like that is #'d out though | 12:55 |
| andymccr | so it won't fail since its not actually set | 12:55 |
| *** fawadkhaliq has joined #openstack-ansible | 12:56 | |
| svg | (yeah, I need this poc to work in 10.1.2, so I'm starting hacking on that tree) | 12:56 |
| andymccr | ahh ok cool | 12:56 |
| andymccr | so that var isnt actually set in the cinder_all.yml group_vars its #'d out, but you should set it in etc/rpc_deploy/rpc_user_config.yml storage_hosts section. | 12:57 |
| svg | yes, but I looked at it as a possible example, so I gues iyt's a typo | 12:59 |
| andymccr | i blame cloudnull :) | 12:59 |
| svg | still, don't see how the limit_container_types: value gets handled in rpc_deployment/roles/cinder_device_add/tasks/device_add.yml | 12:59 |
| svg | O, I see | 13:01 |
| svg | the inventory scripts pops it out | 13:01 |
| andymccr | yeh i think it gets handled by the inventory.py when creating the inventory file | 13:01 |
| svg | I, womdering how you guys came up with all these alternative ideas to handle 'inventory | 13:02 |
| svg | also, the focus on vars, and only little use of inventory group/host vars was unexpected to me | 13:03 |
| svg | (not saying it's bad) | 13:03 |
| svg | I'm the first to acknowlegde I want be able to handle more logic in the inventory | 13:04 |
| andymccr | i think once we defined that it was going to be containerize we had to come up with a way to define the hosts better | 13:04 |
| andymccr | because having to create a tonne of containers and then set them as hosts would be quite hard | 13:04 |
| cloudnull | Svg one thing I'd eventually like to see is some sorry of an inventory management system outside what we have in the dynamic_inventory.py and the json file. I've not thought a lot about it, but would like to see it more cleanly configured using a more distributed system. | 13:10 |
| svg | cloudnull: can you elaborate a it, not sure what excatly you 're aiming at | 13:11 |
| cloudnull | Ie inventory in a distributed db which a client could be configured to manage the inventory allowing additional host / group var entries. | 13:12 |
| *** galstrom is now known as galstrom_zzz | 13:14 | |
| svg | oh, You're talking about ansible in genaral | 13:14 |
| cloudnull | In that way you'd almost get multi region inventory for free. And the deployment host wouldn't be limited to a single node. | 13:14 |
| cloudnull | Yes the ansible inventory. | 13:15 |
| svg | I have been thinking myself on that subject a lot actually; more specifially on how to manage variables, and avoid pitfalls of group_vars inheritance | 13:15 |
| *** subscope has quit IRC | 13:16 | |
| *** subscope has joined #openstack-ansible | 13:16 | |
| cloudnull | In early prototypes of all this we tried using a lot of group vars which turned out to be fraught with peril, ask due to inheritance. | 13:17 |
| svg | Some friends that have been contributing on ansible have the same feeling, we actually hope to get together to brainstorm on it, and perhaps make that a project | 13:17 |
| cloudnull | That'd be awesome! | 13:17 |
| svg | cloudnull: yes, I see you still user 1.6.x which has some serious bugs in it on that matter | 13:17 |
| svg | 1.7 is far better, if I may say so myself :) | 13:17 |
| cloudnull | Yea in 9/10 we were stuck at 1.6.x because 1.7 had delegation issues. | 13:18 |
| cloudnull | In master we are running 1.8.2 presently. | 13:18 |
| svg | the inventory manages a depth parameter in parent-child relations of group, and that weighs the precedence; before 1.7 that parameter was often plainly wrong | 13:19 |
| cloudnull | But 1.9 is the new hotness ;) | 13:19 |
| svg | yeah, at $cust we still are on 1.7.2, but I have been running 1.7.2 till 1.9.0 to test our scripts, and most if not all works nicely | 13:20 |
| svg | the only reason I sty in the 1.7 tree, is that 1.8 became dead slow for us | 13:20 |
| svg | but thats' because my inventory has someting like 70k variables... | 13:21 |
| cloudnull | That's a lot of vars | 13:22 |
| cloudnull | When we move to 1.9 we can drop our lxc module because we got it merged upstream. | 13:23 |
| cloudnull | Which will be nice not to have to carry any longer. | 13:24 |
| svg | last question for now: do you handle any specific rules to determine whether variables go into vars, or into group_vars? | 13:26 |
| svg | I see some repo urls are in group_vars/all.yml | 13:26 |
| svg | whilst others are in e.g. vars/repo_packages/... | 13:27 |
| svg | I can imagine certain people (like we) to now want to depend on external resources, nd needing to mirror everything locally | 13:27 |
| svg | I might e.g be te,pted to put the rabbitmq deb package into a local repository, instead of just wgetting it, but that 'd need some changes on how it gets installed y ansible | 13:28 |
| cloudnull | In the roles everything has a default (in master). The groupvars are what we feel could be used for production. But if you wanted to override all of that you could and I'd recommend doing that in the user_variables.yml file. | 13:29 |
| svg | also, some *_repo_url's have things like "stable main" hardcod, and that could change on a local mirror (we use aptly) | 13:29 |
| cloudnull | As that for can be encrypted using vault and will take the most precedence when referenced on the cli. | 13:29 |
| svg | O, that's the 'global_overrides:' thing? | 13:30 |
| cloudnull | Yes. User_variables.yml will override everything. | 13:31 |
| cloudnull | As it pertains to variables. | 13:31 |
| svg | sorry, I referred to global_overrides: in rpc_user_config.yml | 13:31 |
| cloudnull | The openstack-ansible wrapper will source yml files in the /etc/openstack_deploy directory. | 13:32 |
| andymccr | cloudnull: just an fyi svg is using the 10.x branch so its rpc_deploy sections. but i think what you said above still applies | 13:33 |
| cloudnull | Global_overrides is a method to add variables to the inventory. | 13:33 |
| cloudnull | Which will take precedence over most things but items in user_variables will always win | 13:34 |
| svg | Can I conclude that this way of setting up handling of vars is coming from ansible's inventory model that has bugs and limitation? | 13:35 |
| cloudnull | Yes. Its very much following ansibles inventory model. | 13:35 |
| cloudnull | Which is why we have limited group config in inventory, set specific items in the meta values for hosts and rely mostly on variables set outside inventory. | 13:36 |
| svg | following it? I'd say, you use it less. (as I see/name things, inventory is about group and host vars, plain 'vars' (files, extra vars, ..) are imho more on the playbook side | 13:37 |
| svg | But that's a way of looking at it of coursee :) | 13:37 |
| cloudnull | For sure. | 13:38 |
| svg | Ok, thanks a lot for all these insights, helps a lot to understand rpc's structure. No let's try to hack on that ceph support... | 13:38 |
| svg | thx & ttyl | 13:38 |
| *** sdake has joined #openstack-ansible | 13:39 | |
| cloudnull | Anytime svg talk to you soon. | 13:39 |
| *** sdake_ has quit IRC | 13:43 | |
| *** fawadk has joined #openstack-ansible | 13:47 | |
| *** fawadkhaliq has quit IRC | 13:48 | |
| *** sdake_ has joined #openstack-ansible | 13:51 | |
| *** fawadkhaliq has joined #openstack-ansible | 13:52 | |
| *** fawadk has quit IRC | 13:53 | |
| *** sdake has quit IRC | 13:55 | |
| *** subscope has quit IRC | 13:57 | |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:01 | |
| *** subscope has joined #openstack-ansible | 14:15 | |
| *** Mudpuppy has joined #openstack-ansible | 14:16 | |
| *** Mudpuppy has quit IRC | 14:18 | |
| *** Mudpuppy has joined #openstack-ansible | 14:19 | |
| palendae | mattt: Thanks for your reply to my email - I did not realize you were checking the console log | 14:23 |
| palendae | Though I have certainly been able to SSH into instances once I set up the security group stuff | 14:23 |
| *** stevemar has joined #openstack-ansible | 14:26 | |
| *** fawadk has joined #openstack-ansible | 14:37 | |
| *** hj-hp has joined #openstack-ansible | 14:37 | |
| *** vmtrooper has joined #openstack-ansible | 14:38 | |
| *** fawadkhaliq has quit IRC | 14:41 | |
| *** ishant has quit IRC | 14:43 | |
| *** vmtrooper has quit IRC | 14:44 | |
| *** fawadk has quit IRC | 15:03 | |
| *** hj-hp has quit IRC | 15:06 | |
| *** fawadkhaliq has joined #openstack-ansible | 15:30 | |
| *** galstrom_zzz is now known as galstrom | 15:31 | |
| *** jwagner_away is now known as jwagner | 15:31 | |
| *** hj-hp has joined #openstack-ansible | 15:35 | |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Fix typo in neutron.conf.j2 https://review.openstack.org/170177 | 15:37 |
| *** sdake has joined #openstack-ansible | 15:38 | |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Fix typo in neutron.conf.j2 https://review.openstack.org/170177 | 15:38 |
| stevelle | mattt: good eyes | 15:39 |
| andymccr | eagle eyed! | 15:39 |
| *** sdake_ has quit IRC | 15:42 | |
| Sam-I-Am | mattt: how the heck did that get in there | 15:46 |
| Sam-I-Am | of course, this all needs to change again | 15:46 |
| *** sacharya has joined #openstack-ansible | 15:50 | |
| mattt | i was so happy thinking i may have solved the networking issue, but alas that didn't do it ;P | 15:51 |
| mattt | it is def. a bug tho | 15:51 |
| mattt | and i no longer see the 404 as outlined in https://gist.githubusercontent.com/mattt416/81dbf106107c21a0edab/raw/37002b1379b3275f0fa30aec64adabbedbdd2c2e/gistfile1.txt | 15:52 |
| mattt | so unfortunately w/ that patch my instances still do not get IPs :( | 15:52 |
| openstackgerrit | Hugh Saunders proposed stackforge/os-ansible-deployment: Test instance connectivity in compute scenarios https://review.openstack.org/170187 | 15:52 |
| palendae | mattt: So needy | 15:53 |
| Sam-I-Am | mattt: so... vif confirmation wasnt working? | 15:56 |
| Sam-I-Am | is vif plugging = false ? | 15:56 |
| cloudnull | meeting in a few: cloudnull, mattt, andymccr, d34dh0r53, hughsaunders, b3rnard0, palendae, Sam-I-Am, odyssey4me, serverascode, rromans, mancdaz, dolphm, _shaps_, BjoernT, claco, echiu, dstanek | 15:56 |
| mattt | Sam-I-Am: vif_plugging ? | 15:58 |
| mattt | cloudnull: have to jet in a few mins, will not be able to make meeting due to clocks going forward :( | 15:58 |
| cloudnull | np. | 15:59 |
| Sam-I-Am | mattt: yeah, that thing | 16:00 |
| mattt | Sam-I-Am: i don't see that under nova or neutron | 16:00 |
| * mattt is afk | 16:00 | |
| Sam-I-Am | mattt: vif_plugging_is_fatal = False | 16:02 |
| Sam-I-Am | thats probably why your networks worked anyway | 16:02 |
| Sam-I-Am | seems to be a default in the nova.conf.j2 | 16:02 |
| Sam-I-Am | cloudnull: i thought this was configurable somewhere | 16:03 |
| *** galstrom is now known as galstrom_zzz | 16:09 | |
| *** galstrom_zzz is now known as galstrom | 16:11 | |
| *** Bjoern__ has joined #openstack-ansible | 16:20 | |
| *** javeriak has joined #openstack-ansible | 16:21 | |
| *** vmtrooper has joined #openstack-ansible | 16:27 | |
| *** sdake has quit IRC | 16:29 | |
| *** sdake has joined #openstack-ansible | 16:30 | |
| *** Bjoern__ is now known as BjoernT | 16:31 | |
| *** vmtrooper has quit IRC | 16:32 | |
| *** fawadkhaliq has quit IRC | 16:35 | |
| *** jwagner is now known as jwagner_away | 16:44 | |
| *** sdake_ has joined #openstack-ansible | 16:46 | |
| *** daneyon has quit IRC | 16:49 | |
| *** sdake has quit IRC | 16:50 | |
| *** fawadkhaliq has joined #openstack-ansible | 16:53 | |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: WIP Update Glance service to Kilo https://review.openstack.org/170211 | 17:02 |
| *** mahipal has joined #openstack-ansible | 17:06 | |
| *** cbaesema has quit IRC | 17:15 | |
| *** cbaesema has joined #openstack-ansible | 17:19 | |
| *** galstrom is now known as galstrom_zzz | 17:24 | |
| *** sdake has joined #openstack-ansible | 17:43 | |
| *** jwagner_away is now known as jwagner | 17:45 | |
| *** sdake_ has quit IRC | 17:46 | |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: WIP Update Glance service to Kilo https://review.openstack.org/170211 | 17:47 |
| *** galstrom_zzz is now known as galstrom | 18:02 | |
| *** vmtrooper has joined #openstack-ansible | 18:16 | |
| *** sdake_ has joined #openstack-ansible | 18:19 | |
| *** javeriak has quit IRC | 18:19 | |
| *** javeriak has joined #openstack-ansible | 18:19 | |
| *** vmtrooper has quit IRC | 18:21 | |
| *** sdake has quit IRC | 18:23 | |
| *** server730 has joined #openstack-ansible | 18:24 | |
| server730 | hi, i have vagrant vms that im trying to use the playbooks on, they randomly loose connectivity midway through the first host-setup playbook, and sometimes just go into an aborted state | 18:27 |
| server730 | they are fine before starting and i can still ping them from deploy even after the playbooks give ssh errors | 18:30 |
| server730 | unless they go into abort | 18:30 |
| *** daneyon has joined #openstack-ansible | 18:51 | |
| *** galstrom is now known as galstrom_zzz | 19:05 | |
| *** fawadkhaliq has quit IRC | 19:19 | |
| *** sdake has joined #openstack-ansible | 19:21 | |
| *** sdake_ has quit IRC | 19:24 | |
| cloudnull | server730: its likely an issue with containers being unreachable due to resource issues. Asnible uses SSH for connections and when running an AIO with < 8GB ram we've seen the connections drop. We did create an SSH retry connection module that should resolve a lot of that but if you need, in ansible.cfg you can set that higher. | 19:32 |
| cloudnull | server730: which branch are you running ? | 19:33 |
| server730 | master | 19:33 |
| cloudnull | the ssh retry is in there and is the default connection module. | 19:34 |
| server730 | I also configured it to create 3 infra's, but I am finding that one or the other goes into an aborted state in the first playbook | 19:35 |
| cloudnull | aborted due to connections issues or something else? | 19:35 |
| cloudnull | are you able to provide a stacktrace on when / where it fails ? | 19:36 |
| cloudnull | also have you, in your 3 infra nodes, setup all of the network interfaces to all communicate to one another ? | 19:36 |
| server730 | the playbook output gives SSH errors as usual, but when I go check vagrant status, I see the problematic node shows an aborted state | 19:37 |
| cloudnull | IE https://github.com/stackforge/os-ansible-deployment/tree/master/etc/network | 19:37 |
| server730 | yes I followed the example of the first infra and gave the rest similar network configurations | 19:38 |
| server730 | I find that it goes into that state usually after the Container Restart task | 19:38 |
| cloudnull | are your infra nodes running out of resources ? | 19:39 |
| cloudnull | are all of the containers running ? | 19:39 |
| cloudnull | if you do an lxc-ls -f | 19:40 |
| cloudnull | you can see if the conatiners are running with the various networks as they've been defined witihn your openstack_user_variables.yml file. | 19:40 |
| server730 | resources might very well be an issue as I'm doing everything on one physical machine | 19:41 |
| cloudnull | if you're looking to just kick the tires you can do it all in one node using the gate commit check script. | 19:42 |
| cloudnull | which will do everything on one node but creates clusters for things like galera , rabbitmq and a few others. | 19:43 |
| cloudnull | check https://github.com/stackforge/os-ansible-deployment/blob/master/scripts/gate-check-commit.sh <- this is how we do our gate jobs. | 19:44 |
| cloudnull | a fresh install of Ubuntu 14.04, clone master, run script. it should do the needful for testing everything. | 19:44 |
| cloudnull | and will give you a functional cluster, all be it on one node. | 19:45 |
| cloudnull | server730: also worth a read https://github.com/stackforge/os-ansible-deployment/blob/master/development-stack.rst | 19:46 |
| server730 | cloudnull: thanks I'll take a look at these resources | 19:47 |
| cloudnull | np hit us up if you have any questions. | 19:48 |
| server730 | will do. | 19:48 |
| *** vmtrooper has joined #openstack-ansible | 20:05 | |
| *** vmtrooper has quit IRC | 20:11 | |
| *** sdake_ has joined #openstack-ansible | 20:20 | |
| *** sdake has quit IRC | 20:24 | |
| openstackgerrit | Jacob Wagner proposed stackforge/os-ansible-deployment: Fixes cinder backend support for backends without a volume_group key https://review.openstack.org/170274 | 20:43 |
| openstackgerrit | Sudarshan Acharya proposed stackforge/os-ansible-deployment: WIP: Managing policy file with default file and user variables. https://review.openstack.org/168104 | 20:45 |
| openstackgerrit | Miguel Alejandro Cantu proposed stackforge/os-ansible-deployment-specs: Add spec for BP implement-ceilometer https://review.openstack.org/169417 | 20:50 |
| *** BjoernT has left #openstack-ansible | 20:50 | |
| openstackgerrit | Miguel Alejandro Cantu proposed stackforge/os-ansible-deployment-specs: Add spec for BP implement-ceilometer https://review.openstack.org/169417 | 20:52 |
| *** galstrom_zzz is now known as galstrom | 20:52 | |
| server730 | ok so I've been trying ti run the gate-check-commit.sh and keep getting a [ Status: Failure ] at the end, no further info on whats wrong | 20:53 |
| cloudnull | in the scroll back do you see anything that would indicate what the issue is? did it fail on some tempest part? also in the /openstack/log/ansible_logging directory theres a verbose ansible log which should help to debug the failures. | 20:57 |
| openstackgerrit | Sudarshan Acharya proposed stackforge/os-ansible-deployment: WIP: Managing policy file with default file and user variables. https://review.openstack.org/168104 | 21:01 |
| *** sdake_ is now known as sdake | 21:02 | |
| svg | cloudnull: still around? | 21:07 |
| cloudnull | yesir | 21:07 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updated log file entries for neutron https://review.openstack.org/170284 | 21:09 |
| svg | we talked about inbventory earlier today, just wanted to let you know, if you have questions on (python lib wise) ansible.inventory, or thoughts or whatever, it's the part I am quite familiar with, and have particular interest with. | 21:11 |
| svg | So far, little new things happened there, and I hope after v2, there will be some serious work onde on it. | 21:11 |
| svg | done | 21:11 |
| svg | Your insight and experience might be valuable and interesting, and how that can evolve to serve this project better | 21:12 |
| svg | just fyi... | 21:12 |
| cloudnull | thats awesome! | 21:13 |
| cloudnull | i'd love to participate in something that makes inventory more scalable/manageable. | 21:14 |
| server730 | sorry to interrupt your conversaton guys, cloudnull: the ansible_log is empty and the only error i have found in the scroll back is "fatal error: yaml.h: No such file or directory compilation terminated." | 21:15 |
| cloudnull | so ill definitely reach out to you once we embark on the "make inventory moar better" adventure. | 21:15 |
| svg | super! | 21:16 |
| cloudnull | server730: that would seem that your box is missing yaml. | 21:17 |
| cloudnull | you running ubuntu 14.04? | 21:17 |
| server730 | cloudnull: yep | 21:18 |
| *** sdake_ has joined #openstack-ansible | 21:18 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment-specs: Add modularize-neutron-plays specification https://review.openstack.org/169189 | 21:18 |
| server730 | cloudnull: the development-stack.rst doesnt talk about installing any dependency packages first..? | 21:18 |
| cloudnull | the "./scripts/gate-check-commit.sh" should do that for you. | 21:19 |
| sigmavirus24 | Is there another core who wants to workflow https://review.openstack.org/#/c/166986/ ? | 21:19 |
| cloudnull | server730: from the box run the "./scripts/teardown.sh" script which will cleanup/reset the machine. Then rerun "./scripts/gate-check-commit.sh" and let me know if it still fails. | 21:21 |
| *** sdake has quit IRC | 21:22 | |
| server730 | cloudnull: yes i tried that as well, no luck [ Error Info - 275 0 ] | 21:23 |
| *** sdake has joined #openstack-ansible | 21:23 | |
| cloudnull | "Error Info - 275 0" is that all you have as a return ? | 21:24 |
| server730 | cloudnull: yes that's all it says before failure | 21:26 |
| cloudnull | so lets step through a bit of this. | 21:26 |
| *** sdake_ has quit IRC | 21:27 | |
| cloudnull | can you run the ./scripts/bootstrap-ansible.sh | 21:27 |
| server730 | cloudnull: done | 21:34 |
| *** jlvillal has quit IRC | 21:34 | |
| cloudnull | no failures ? | 21:34 |
| *** jmeridth has quit IRC | 21:34 | |
| server730 | cloudnull: mope | 21:35 |
| cloudnull | ok. now. | 21:35 |
| cloudnull | rm logs && ./scripts/bootstrap-aio.sh | 21:35 |
| *** docaedo has quit IRC | 21:35 | |
| cloudnull | ^ removes log link in your working directory and then rebootstraps the all in one hosts. | 21:35 |
| server730 | ok this took slightly longer but is also done, no errors | 21:36 |
| *** docaedo has joined #openstack-ansible | 21:37 | |
| *** persia_ has joined #openstack-ansible | 21:37 | |
| *** persia_ has quit IRC | 21:37 | |
| *** persia_ has joined #openstack-ansible | 21:37 | |
| server730 | cloudnull: should i re-run the gate-check-commit now? | 21:38 |
| cloudnull | no. | 21:38 |
| cloudnull | cd playbooks | 21:38 |
| *** erikmwilson has quit IRC | 21:38 | |
| *** bgmccollum has quit IRC | 21:38 | |
| *** aslaen has quit IRC | 21:38 | |
| *** hughsaunders has quit IRC | 21:38 | |
| *** lbragstad has quit IRC | 21:38 | |
| *** persia has quit IRC | 21:38 | |
| cloudnull | openstack-ansible setup-everything.yml | 21:38 |
| cloudnull | check tht | 21:39 |
| *** bgmccollum has joined #openstack-ansible | 21:39 | |
| *** erikmwilson has joined #openstack-ansible | 21:39 | |
| *** jlvillal has joined #openstack-ansible | 21:39 | |
| *** aslaen has joined #openstack-ansible | 21:39 | |
| cloudnull | openstack-ansible haproxy-install.yml setup-everything.yml | 21:39 |
| *** lbragstad has joined #openstack-ansible | 21:39 | |
| server730 | cloudnull: Error -> got an unexpected keyword argument 'vault_password' | 21:39 |
| cloudnull | have you encrypted your config files ? | 21:40 |
| server730 | cloudnull: not that I'm aware of | 21:40 |
| *** jmeridth has joined #openstack-ansible | 21:40 | |
| server730 | I just cloned it awhile ago | 21:40 |
| *** hughsaunders has joined #openstack-ansible | 21:40 | |
| cloudnull | that error came from "openstack-ansible haproxy-install.yml setup-everything.yml" command ? | 21:41 |
| server730 | cloudnull: no just -> openstack-ansible setup-everything.yml | 21:41 |
| server730 | and yes for the later now too | 21:42 |
| cloudnull | we're not using vault anywhere. | 21:42 |
| cloudnull | that would come from something in your config that is being parsed that has a vault setup. | 21:42 |
| *** Mudpuppy_ has joined #openstack-ansible | 21:42 | |
| server730 | cloudnull: this is the full output: | 21:43 |
| cloudnull | ansible-playbooks -e @/etc/openstack_deploy/user_variables.yml haproxy-install.yml setup-everything.yml | 21:43 |
| cloudnull | ^ try that if you dont mind | 21:43 |
| server730 | File "/usr/local/bin/ansible-playbook", line 160, in main; inventory = ansible.inventory.Inventory(options.inventory, vault_password=vault_pass); TypeError: __init__() got an unexpected keyword argument 'vault_password' | 21:44 |
| *** sacharya has quit IRC | 21:45 | |
| server730 | cloudnull: same error with the full playbook command as well | 21:45 |
| cloudnull | also can you run "ansible --version" ? | 21:45 |
| cloudnull | i remember that error in ansible <= 1.5.x | 21:45 |
| cloudnull | did you install ansible via the apt packages ? | 21:46 |
| cloudnull | let me see if i can find the bug im thinking of. | 21:46 |
| *** Mudpuppy has quit IRC | 21:46 | |
| server730 | cloudnull: ok this is odd, i have checked version before, but now it says AttributeError: 'module' object has no attribute 'DEFAULT_LOAD_CALLBACK_PLUGINS' | 21:47 |
| *** jwagner is now known as jwagner_away | 21:48 | |
| server730 | cloudnull: and the version should be 1.9, yes i installed through apt | 21:49 |
| cloudnull | ah. ansible is a bad state . | 21:50 |
| cloudnull | we install ansible from source. | 21:50 |
| cloudnull | and our source install defaults to 1.8.x | 21:50 |
| cloudnull | apt-get remove --purge ansible | 21:50 |
| cloudnull | and then rerun the bootstrap-ansible.sh script | 21:50 |
| cloudnull | that should resolve ansible being angry. | 21:51 |
| server730 | ansible is now 1.8.2, and bootsrap failed [ Error Info - 275 0 ] [ Status: Failure ] | 21:52 |
| svg | sounds like a borken/mixed up pip/other install | 21:53 |
| svg | cloudnull: https://twitter.com/svg/status/583748319308095489 | 21:53 |
| *** vmtrooper has joined #openstack-ansible | 21:54 | |
| cloudnull | server730 seems like there is some pip vs apt cruft which is conflicting on your host. | 21:55 |
| * cloudnull svg reading | 21:55 | |
| server730 | svg: yes I guess easiest way is to now install a fresh ubuntu, I didnt actuallt start with a fresh ubuntu, i was using it for the vagrant-ansible workflow so I just removed the VMs and tried running the aio scripts | 21:56 |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: Enable offline compression for resource files https://review.openstack.org/170298 | 21:57 |
| server730 | cloudnull: thanks for your help, I will try with a new setup | 21:57 |
| cloudnull | sorry i couldn't help you make it go . but if you try again please let us know what happens and if you have further questions. | 21:58 |
| *** vmtrooper has quit IRC | 21:59 | |
| svg | server730: that's a typical ansible install problem, fresh oinstall should solve it | 22:07 |
| *** Mudpuppy_ has quit IRC | 22:12 | |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: Move SYMLINK_DIR declaration to bootstrap-aio.sh https://review.openstack.org/170303 | 22:14 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set Glance large_object vars as ansible variables https://review.openstack.org/169693 | 22:16 |
| *** sigmavirus24 is now known as sigmavirus24_awa | 22:27 | |
| *** hj-hp has quit IRC | 22:38 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set Glance large_object vars as ansible variables https://review.openstack.org/169717 | 22:38 |
| *** hj-hp has joined #openstack-ansible | 22:41 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Fixes cinder backend support for backends without a volume_group key https://review.openstack.org/170274 | 22:46 |
| *** hj-hp has quit IRC | 22:56 | |
| *** galstrom is now known as galstrom_zzz | 23:00 | |
| *** server730 has quit IRC | 23:21 | |
| *** hj-hp has joined #openstack-ansible | 23:26 | |
| *** sdake_ has joined #openstack-ansible | 23:27 | |
| *** sdake has quit IRC | 23:30 | |
| *** sdake has joined #openstack-ansible | 23:33 | |
| *** sdake_ has quit IRC | 23:37 | |
| *** hj-hp has quit IRC | 23:39 | |
| *** vmtrooper has joined #openstack-ansible | 23:43 | |
| *** vmtrooper has quit IRC | 23:48 | |
| openstackgerrit | Miguel Alejandro Cantu proposed stackforge/os-ansible-deployment-specs: Add spec for BP implement-ceilometer https://review.openstack.org/169417 | 23:54 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!