Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel, 2016-04-19

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:43 xarses joined #fuel
01:03 adam_g mwhahaha, xarses any clue how to quickly resolve the address/port that the rabbit server is listening on?
01:04 mwhahaha It's in astute.yaml
01:05 mwhahaha You can check one of the openstack service tasks
01:05 adam_g ah
01:05 mwhahaha Amqp_hosts I think
01:05 adam_g amqp_hosts in hiera
01:05 adam_g right
01:05 mwhahaha I'm on mobile at the moment
01:06 adam_g mwhahaha, thats for the help on the fly
01:07 adam_g mwhahaha, this might be a silly one, but is it possible to instatiate/realize some puppet resource on a non-compute node that will write out the neutron config as it would be  on the compute node?
01:10 fatdragon joined #fuel
01:14 mwhahaha Check if the task gets run on that node and you could add some astute.yaml value
01:14 mwhahaha We have some override vars that could be set with merge
01:29 mwhahaha alternatively for your plugin you could try to have the existing task run on the node you want
01:30 adam_g yea
01:30 adam_g im trying to get the L2 agent installed like it would be on a compute node
01:32 mwhahaha so in your plugin deployment_tasks.yaml you could steal the existing task and just change out the group/role
01:37 adam_g mwhahaha, yeah, im ahving trouble finding the task that does what i want
01:37 adam_g something tells me its tightly coupled with the nova module and diffulult to reuse
01:50 rongze joined #fuel
01:54 mwhahaha Possibly I'll take a look in a bit
02:07 mwhahaha adam_g: which agent? we've got the l3 agent stuff which is part of the openstack-network tasks
02:08 l4ng1t joined #fuel
02:12 rongze joined #fuel
02:54 jobewan joined #fuel
03:42 charz joined #fuel
04:07 intr1nsic joined #fuel
04:33 intr1nsic joined #fuel
04:56 reddydodda joined #fuel
05:05 xarses adam_g: https://github.com/openstack/fuel-library/blob/stable/8.0/deployment/puppet/osnailyfacter/modular/openstack-network/tasks.yaml#L75-L87
05:11 reddydodda joined #fuel
05:12 eric_lopez joined #fuel
05:44 kszukielojc joined #fuel
05:54 e0ne joined #fuel
06:10 e0ne joined #fuel
06:11 rongze joined #fuel
06:12 l4ng1t joined #fuel
06:18 dcs joined #fuel
06:22 magicboiz joined #fuel
06:23 pal_bth joined #fuel
06:31 e0ne joined #fuel
06:37 javeriak joined #fuel
06:38 magicboiz joined #fuel
07:01 chopmann joined #fuel
07:16 18VAAPPR3 joined #fuel
07:30 magicboiz joined #fuel
07:40 reddydodda joined #fuel
07:50 magicboiz joined #fuel
08:03 javeriak joined #fuel
08:04 fatdragon joined #fuel
08:13 e0ne joined #fuel
08:26 javeriak joined #fuel
08:32 Miouge joined #fuel
08:40 yassou joined #fuel
08:41 yassine joined #fuel
08:42 aleksandr_null joined #fuel
08:54 Alcest joined #fuel
09:03 reddydodda joined #fuel
09:04 magicboiz joined #fuel
09:06 rongze joined #fuel
09:35 rongze joined #fuel
09:45 reddydodda joined #fuel
09:50 magicboiz joined #fuel
10:16 magicboiz joined #fuel
10:36 rongze joined #fuel
10:36 rongze joined #fuel
10:51 sarcasticidiot joined #fuel
11:09 rongze joined #fuel
11:11 reddydodda joined #fuel
11:20 javeriak joined #fuel
11:59 javeriak joined #fuel
12:06 fatdragon joined #fuel
12:16 sarcasticidiot joined #fuel
12:31 yassine joined #fuel
12:36 javeriak joined #fuel
12:42 rongze joined #fuel
12:47 reddydodda joined #fuel
12:57 sarcasticidiot joined #fuel
13:00 rongze joined #fuel
13:05 rongze joined #fuel
13:17 reddydodda joined #fuel
13:20 byrdog55 joined #fuel
13:32 xarses joined #fuel
13:34 jobewan joined #fuel
13:54 rongze_ joined #fuel
14:00 Miouge joined #fuel
14:01 mwhahaha .
14:02 blinky_ghost joined #fuel
14:28 claflico joined #fuel
14:32 rongze joined #fuel
14:39 xarses joined #fuel
14:47 eric_lopez joined #fuel
14:47 Miouge joined #fuel
14:50 blahRus joined #fuel
15:00 blinky_ghost Hi all, I've deployed mos 8.0 and now I cannot centos 6 qcow2 based images, anyone had this problem and knows how to solve it? Thanks
15:00 fatdragon joined #fuel
15:03 mwhahaha blinky_ghost: are you getting any specific error?
15:04 blinky_ghost mwhahaha: no, it's just stuck on the boot, with an older image it's gives me a kernel panic. With this one http://cloud.centos.org/centos/6/images/CentOS-6-x86_64-GenericCloud.qcow2 it's stuck "Booting from hard disk"
15:05 mwhahaha that's odd, not sure
15:06 blinky_ghost mwhahaha, with centos 7 and cirros works fine, I guess it could be some kernel issue
15:06 EZPZ joined #fuel
15:07 mwhahaha might need a newer image
15:07 mwhahaha do you know if that centos-6 is like 6.0 or is it 6.7 or whatever is the latest
15:09 rongze joined #fuel
15:10 blinky_ghost mwhahaha I think it's the latest
15:11 mwhahaha yea looks to be based on timestamp
15:21 rongze joined #fuel
15:26 blinky_ghost mwhahaha, I've tested this image on another Openstack Distro (RDO Liberty) and I don't have this problem, could be something related with my libvirt/qemu version
15:27 mwhahaha yea i'd submit a bug if it work on rdo liberty
15:27 mwhahaha perhaps we are configuring something different that is causing it
15:27 blinky_ghost ok thanks
15:28 blinky_ghost I'll try to update the kernel on the image to see if it works
15:54 blinky_ghost same problem, opening a bug
15:59 blinky_ghost https://bugs.launchpad.net/mos/+bug/1572219
16:14 rongze joined #fuel
16:16 gomarivera joined #fuel
16:18 gomarivera can fuel plugins use other puppet-* repos? can they be imported or managed through something like librarian? or is this bad practice?
16:21 xarses gomarivera: yes fuel plugin's can use their own modules, the [deployment_]tasks.yaml format has the option to configure the module path
16:21 xarses gomarivera: what are you looking to do?
16:24 xarses blinky_ghost: there is nearly no information for anyone to troubleshoot with on that bug report. Please include the link to the image you used if possible and a full console log for the instance
16:25 gomarivera I am looking to do a fuel-plugin for zaqar - this is not currently implemented - and there's a puppet-zaqar repo that has manifests that already configure/install zaqar - so I am wondering if there's a way to essentially create a dependency on that library - kind of like berkshelf in chef
16:26 gomarivera but if this is maybe bad practice - then I will implement the plugin stand alone - but I would have liked to reuse working manifests
16:27 xarses gomarivera: you would either include the whole source of the module in the plugin (way less than ideal for long term maintenance) or use the pre_build_hook to fetch it via Puppetfile
16:29 xarses we don't have a plugin example of this, but you can copy the Puppetfile format and update_modules.sh script from fuel-library
16:29 xarses https://github.com/openstack/fuel-library/tree/master/deployment
16:30 gomarivera ok that makes sense - thanks - I'll give that a try - are there any expectations that fuel-plugins should standalone? - not sure i saw in the wiki any kind of guidelines around standards of quality - but maybe i missed it
16:33 xarses there isn't a formal set yet, since we have a diverse set of sources
16:34 xarses you can refer to http://lists.openstack.org/pipermail/openstack-dev/2016-February/085636.html for some 'classes' of quality
16:34 xarses that was proposed recently
16:35 gomarivera cool - thank you
16:35 xarses basically, its up to you
16:42 thumpba joined #fuel
16:59 adam_g xarses, ya, i was pulling that ML2 task in yesterday but it looked like its still missing some networking/bridge/ovs setup that prevented the agents from actually starting
17:04 xarses adam_g: br-int?
17:04 adam_g xarses, yea
17:05 xarses ovs-agent is responsible for creating it, so it's missing some config most likely
17:05 xarses are you providing the ovs agent package?
17:06 adam_g xarses, the neutron-ovs-agent one? no, whatever gets pulled from the ml2 task
17:06 xarses use the start command from the shell and see what it's upset about
17:07 xarses you ran the netconfig task on the node correct?
17:08 adam_g xarses, netconfig was ran as part of the initiail deployment yea.. starting the agent manually shows it spinning in its logs waiting on br-int
17:10 javeriak joined #fuel
17:14 xarses adam_g: ping me in ~10 min
17:14 cr0wrx joined #fuel
17:14 adam_g xarses, sure, thanks
17:15 cr0wrx with fuel is it possible to start with a small non-HA setup and then convert to HA later once needed / more hardware available without rebuilding from scratch? Is it as simple as adding more controller nodes, or is there more to it?
17:17 xarses cr0wrx: since 6.0 all deployments have all of the ha components (there is no choice between ha or simple modes) you're choice is how many controller nodes to meet your availability needs
17:17 xarses because all of the ha parts are configured regardless, adding more controllers later is simple, you just need to add them to the cluster and deploy the changes
17:17 cr0wrx sure. so if I start with 1 controller node initially, there should be no issue moving to 3 nodes and having everything HA just take over?
17:18 cr0wrx sounds good
17:18 cr0wrx now I just gotta figure out hardware to purchase and then we will have a stack! hooray
17:18 xarses you can even start with 3, roll up to 5, scale back to 1
17:19 cr0wrx neat
17:19 xarses your only cavieat is that you need an odd number of nodes for quorum
17:19 cr0wrx sure
17:20 xarses otherwise you are at a real risk for split brain and manual intervention
17:20 cr0wrx sure
17:21 cr0wrx from a hardware side I'm assuming it can all be heterogenous as well (so multiple controllers can have diff hardware specs, it doesn't care, outside of performance)
17:23 xarses thats correct, but you can get into heavily nuanced issues with regards to load causing service availability issues if the controllers don't perform the same. This is especially visible if you use ceph and/or combine other roles onto the controller node
17:25 rongze joined #fuel
17:27 xarses adam_g: you will want to browse the ovs logs to see why it doesn't want to bring br-int up. According to the neutron guys, ovs-agent will tell ovs to start it. either ovs-agent is missing configuration (/etc/neutron/plugins/ml2/ml2_conf.ini || /etc/neutron/plugin.ini) or something is wrong in ovs
17:27 chopmann joined #fuel
17:27 adam_g xarses, yeah ill check it in a few. my VMs seem to kernel panic if left alone overnight, so i need to reboot the whole env soon
17:28 adam_g thanks
17:31 AJSchroeder joined #fuel
17:31 AJSchroeder hello
17:33 AJSchroeder would anyone know how to add VPNaaS to an already deployed cloud?
17:33 AJSchroeder the cloud was deployed from fuel 8.0 so it is openstack liberty running on ubuntu
17:37 mwhahaha you'd have to do it all manually
17:38 AJSchroeder Ok, I'm assuming that I can follow the manual install procedure in the docs?
17:41 javeriak_ joined #fuel
17:43 javeria__ joined #fuel
17:45 mwhahaha most likely
17:45 mwhahaha we're working on being able to change stuff after deployment in 10+ but for 8 we don't have a way to go back and do something like that
17:46 AJSchroeder I am alright with installing it manually, but I didn't want to screw anything up with fuel
17:47 AJSchroeder being able to change things after deployment would be amazing, but I understand that is a lot easier said than done
17:49 xarses you can technically bring in a plugin and you can manually execute it's tasks, the engine in 8 simply cant orchestrate that for you. This would be beneficial in that if you had to deploy the controllers again, they should get configured with it correctly
17:50 xarses In this case, you will likely have to ensure that all services are restarted correctly.
17:51 xarses on the plausible || hacking side, you can again bring in a plugin and hack the nodes with tasks to run into error state. then when you click deploy changes the tasks graph should be recalculated correctly
17:52 xarses but that should not be done in a environment with workloads unless you know its going to work out well for you (it could cause outages and other oddness)
17:53 xarses if you do configure it by hand, any deployment of a controller role will trigger all controllers to re-deploy, in this case any by hand changes would be reverted so you'd have to have an script handy to re-apply your changes
17:54 AJSchroeder I have a test env setup in virtualbox that I could try to manually execute the tasks on that way I don't screw up my environment with the workloads
17:55 AJSchroeder do you have any docs on how to bring in a plugin manually?
17:57 xarses nope, you will want to start with changing the plugin metadata to is_hotpluggable: ture https://github.com/openstack/fuel-plugins/blob/master/examples/fuel_plugin_example_v4/metadata.yaml#L23
17:58 xarses if it's hot pluggable is already true you might be able to just deploy it and it should work but I don't think anyone uses this feature yet and its use in 8 is limited
17:58 xarses then you can rebuild the plugin
17:58 xarses install it
17:59 xarses examine it's deployment_tasks.yaml
17:59 xarses the task names are the `- id:` line https://github.com/openstack/fuel-plugins/blob/master/examples/fuel_plugin_example_v4/deployment_tasks.yaml#L13
17:59 AJSchroeder and I was wondering if I could do that too (i.e. make it compatible with 8) because the packages required should change
17:59 xarses the roles it runs on are https://github.com/openstack/fuel-plugins/blob/master/examples/fuel_plugin_example_v4/deployment_tasks.yaml#L13
18:00 AJSchroeder *shouldn't change
18:00 xarses then you can use the `fuel node` sub commands for tasks `--tasks` and select the nodes `--node` to run the tasks on
18:01 xarses the problem with `is_hotpluggable` in 8 is that its only guaranteed to work when the task doesn't rely on tasks in the regular deployment graph
18:02 xarses so a plugin like LMA (which has `is_hotpluggable: true`) which goes and configures a bunch of stuff not related to openstack works well
18:03 xarses when you get to VPNaaS it becomes more complicated
18:03 AJSchroeder because it relies on neutron which is part of the regular deployment?
18:04 xarses it working in 8 also contingent on all the tasks being executed from a new role which causes the deployment engine to calculate the graph correctly
18:04 xarses it depends on the plugin but I'm guessing it would
18:05 Mike_G__ joined #fuel
18:08 chopmann left #fuel
18:22 fuel-slackbot joined #fuel
18:59 adam_g xarses, its safe to assume the pwd from which a plugin will apply its puppet manifest is /etc/fuel/plugins/$plugin/ ?
19:05 xarses the pwd is relative to the mcollective process, I don't think it chdir's anywhere
19:06 xarses what do you want the relative path for?
19:08 neophy joined #fuel
19:09 adam_g xarses, im trying to exec a script from the plugin's ./scripts from a puppet exec {}
19:10 xarses if you are following module spec the exec can expand the relative path of the scripts / templates folder
19:14 AJSchroeder xarses, thank you for the help with the plugins, I'm going to give it a shot later tonight
19:20 xarses adam_g: hmm, I see $module_name, but not path...
19:24 thumpba joined #fuel
19:27 thumpba joined #fuel
20:00 elo joined #fuel
20:03 rongze_ joined #fuel
20:06 adam_g xarses, fyi the issue i was having trying to get l2 setup correclty was that simply pulling in the ml2 task doesnt set required values in neutron.conf, specifically root_helper + rabbit info
20:06 rongze__ joined #fuel
20:14 xarses adam_g: oh, you probably need to do neutron-common
20:15 adam_g xarses, is there a task for that or do you mean the package?
20:15 adam_g package should get pulled in as a Depends
20:15 xarses there is a task
20:16 adam_g ah
20:16 xarses https://github.com/openstack/fuel-library/blob/stable/8.0/deployment/puppet/osnailyfacter/modular/openstack-network/tasks.yaml#L39
20:16 adam_g im just setting it manually ATM
20:17 adam_g thanks
20:26 adam_g xarses, also, i forgot what you told me on friday WRT ordering of post_deploy tasks
20:27 adam_g will post deploy tasks on a primary node execute before those on a non-primary?
20:29 xarses before fuel 9 yes
20:29 xarses but typically you only differentiate because primary will only run once
20:29 xarses where there may be multiples of the others
20:34 adam_g xarses, is it safe to have a functional dependency between a post_deploy on a non-primary and one on a primary?
20:34 adam_g ie, the primary's post deploy publishes an image, the non-primary post deploy polls glance until the image shows up
20:41 xarses ah, ya make sure you ask for the internal endpoint URI back from keystone
20:42 xarses or for the interum source /root/openrc
20:42 xarses keeping in mind that openrc may be out of date
20:42 xarses on an already deployed cluster
20:58 adam_g xarses, ack
21:02 dims_ joined #fuel
21:20 adam_g xarses, oh also, can a plugin for a configured (but not deployed) environment be updated between minor versions? it looked like doing so wouldn't update the environments deplyment tasks in the DB
21:21 xarses `fuel plugin --sync` should fix that
21:21 adam_g ah!
21:21 xarses didn't I reply your mail on that?
21:22 xarses hrm, oh ya chrome was all blown up yesterday, maybe not
21:22 adam_g xarses, doesnt look like it, unless its lost in gmails horrible non-threading
21:51 eric_lopez joined #fuel
21:57 adam_g any reason why the VMs setup by fuel-devops dont use virtio drivers for networking?
22:02 mwhahaha yea i think we had problems with them
22:04 adam_g just manually updating the VMsi found the admin node comes up with interfaces named traditionally (eth0) instead of the enp0s3
22:12 rongze joined #fuel
22:14 mwhahaha that's odd i thought we switched to enp0s3, which version?
22:17 adam_g mwhahaha, 8. yeah, biosdevname is enabled and all ndoes have the enp0s* schema. im saying, when i enable virtio manually on the VMs it screws with the NIC namign
22:17 mwhahaha ah
22:17 rongze joined #fuel
22:26 intr1nsic joined #fuel
22:38 xarses joined #fuel
23:39 elo joined #fuel

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary