[00:03] <kelvinliu> hi veebers, I just saw u mentioned that in standup. I was looking at deploying percona-cluster on caas
[00:06] <kelvinliu> veebers, the problem was `leader-elected` hook failed due to `network-get` could not find relation primary address(it's empty `{MACAddress: InterfaceName: Addresses:[]}`)
[00:06] <veebers> kelvinliu: ah right, thanks. I could have sworn it was a relations thing but wasn't too sure ^_^
[00:11] <kelvinliu> veebers, Is the issue on IAAS ?
[00:16] <veebers> kelvinliu: aye, it's charm related. Not sure of the impact for caas.
[00:16] <kelvinliu> ah, did ur deployment failed at the same hook?
[00:18] <kelvinliu> https://pastebin.ubuntu.com/p/7JXwP44Ry4/
[00:18] <kelvinliu> Is the error u got, veebers ?
[00:20] <veebers> kelvinliu: no, the percona-cluster error was in deb-relation-changed
[00:20] <veebers> it deployed fine, was when I attempted to relate it to a mediawiki app that it fails
[00:21] <kelvinliu> ic
[00:22] <kelvinliu> the problem might be different.
[00:23] <veebers> I believe it is
[08:28] <valenitn> hi guys
[08:28] <valenitn> is it ok to ask for help on deploying with juju here?
[08:29] <rick_h_> valenitn: sure thing
[08:30] <valenitn> great, thanks
[08:30] <valenitn> I'm trying to deploy a xenial charm on azure with storage
[08:31] <rick_h_> valenitn: cool, you've peeked at https://jujucharms.com/docs/2.3/charms-storage ?
[08:31] <valenitn> this worked fine a few weeks ago, but now I get the agent blocked in "allocating" and storage in "pending"
[08:32] <rick_h_> valenitn: hmm, so have to see in the debug log what's up
[08:32] <valenitn> yep, that's where I got the command from
[08:32] <valenitn> there's nothing relevant in the juju logs for the unit or the machine
[08:32] <rick_h_> valenitn: it might be in the controller logs as it sets up the storage to be used
[08:33] <rick_h_> valenitn: the unit itself doesn't make the calls out to the storage provider
[08:33] <valenitn> I think the storage works because I see the disks on the machine with lsblk
[08:33] <valenitn> but they are not mounted
[08:34] <rick_h_> valenitn: hmm, can you `juju switch controller` and check the logs on the controller machine?
[08:35] <rick_h_> if there's nothing in the unit/machine log there it's the only other place I can think to look for any error/issue feedback
[08:38] <valenitn> there is this in the logs, but I remember seeing it even when it was working:
[08:38] <valenitn> machine-0: 08:09:36 WARNING juju.cmd.jujud determining kvm support: INFO: Your CPU does not support KVM extensions KVM acceleration can NOT be used : exit status 1 no kvm containers possible
[08:39] <rick_h_> valenitn: yea, but that's nothing to do with storage. That just means you can't create kvm containers on that platform
[08:39] <valenitn> on the machine syslog I see May  4 08:15:38 machine-1 HV_FCOPY: open /dev/vmbus/hv_fcopy failed; error: 2 No such file or directory
[08:41] <rick_h_> hmm, that looks better more interesting. There's some instances of that out there. https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1614618
[08:41] <mup> Bug #1614618: hv-fcopy-daemon.service system failed start <amd64> <apport-bug> <kernel-da-key> <kernel-hyper-v> <uec-images> <yakkety> <linux (Ubuntu):Confirmed> <linux (Ubuntu Xenial):Confirmed> <https://launchpad.net/bugs/1614618>
[08:42] <rick_h_> valenitn: might have to file a bug.
[08:43] <rick_h_> valenitn: does it work intermittently? I see some links around bits of this stuff being racy
[08:44] <valenitn> it failed consistently in the last days
[08:44] <valenitn> I thought something on azure changed
[08:45] <valenitn> maybe the ami for ubuntu?
[08:45] <rick_h_> valenitn: yea, maybe something with a new kernel or maybe something Juju needs to be doing automatically that Azure changed. I'm not sure.
[08:45] <rick_h_> valenitn: if you can file a bug with the charm/instructions for what you're trying we can chase it down
[08:46] <valenitn> ok, will try to look a little more and file a bug
[08:46] <valenitn> thanks for help
[08:47] <rick_h_> valenitn: yea sorry I don't have a quick fix for you
[08:49] <valenitn> np
[09:32] <TheAbsentOne> rick_h_: do you know a good, recent interface example. I'm not really getting it how I can properly send information from one charm to another through a self-written custom interface :/ it's to_publish right?
[09:34] <TheAbsentOne> https://jujucharms.com/docs/2.3/developer-layers-interfaces <-- does relations.unit in this tutorial represent all connected charms?
[11:15] <srihas> hi guys, I have installed nova-compute on our compute nodes, Now I am thinking to install the percona clulster on another three nodes supposed to act as controllers. What does the "deploy --to lxd:0" mean here? I have skipped the neutron deployment as I want to do it later with cisco ACI.
[12:06] <roadmr> srihas: "deploy --to lxd:0" means it will deploy into a new lxd container which will be created on machine 0
[12:06] <roadmr> srihas: https://jujucharms.com/docs/2.3/charms-deploying-advanced#deploying-to-specific-machines
[12:26] <srihas> aha
[12:27] <srihas> roadmr: what if the machine is not yet in the list of machines on juju ? then I go the way I do --constraints tags=db --to lxd ?
[12:29] <srihas> but then I cannot specify :0 or 1 or something, right?
[12:30] <srihas> "To deploy to specific, pre-existing machines the --to option is used. When this is done, unless the machine was created via add-machine, a charm has already been deployed to the machine."
[12:46] <roadmr> srihas: AFAIk there's always a machine 0, but you may be able to specify any existing machine
[12:46] <roadmr> srihas: it sounds like you're following a document, that should tell you the machines that are available
[12:59] <srihas> roadmr: I want to install percona on a new fresh machine thats not yet in the list of machines in juju
[12:59] <srihas> juju machines doesn't show a machine with id 0 for me :o
[14:31] <zeestrat> srihas: Are you following along https://docs.openstack.org/charm-deployment-guide/latest/install-openstack.html#deploy-openstack?
[14:34] <srihas> zeestrat: yeah
[14:34] <srihas> I am on it
[14:34] <srihas> but its sometimes confusing me
[14:35] <srihas> like now the swift-server.config has the bind_ip = 0.0.0.0 instead of specific IP
[14:35] <srihas> :7
[14:37] <zeestrat> srihas: Gotcha. I'd recommend maybe looking at using bundles which is mentioned in the next step https://docs.openstack.org/charm-deployment-guide/latest/install-openstack-bundle. It will make it easier to define things and map applications to machines in a predictable way
[14:38] <srihas> zeestrat: thats a good point. I will study it though I can't deploy directly
[14:39] <zeestrat> Also if you find some doc bugs you can file them here https://launchpad.net/charm-deployment-guide
[14:39] <srihas> zeestrat: is there a way to add all the "ready" nodes to juju
[14:39] <srihas> like I get them when I do the juju machines
[14:40] <srihas> without any charms (atleast OpenStack related) deployed
[14:42] <zeestrat> Just so I understand, is there a reason you can't deploy bundles directly?
[14:43] <srihas> zeestrat: yeah, we will be using cisco ACI for neutron and ScaleIO for storage. so no ceph and neutron
[14:45] <zeestrat> srihas: Right. But you're using openstack charms with juju for the other components such as Nova compute? You can just remove them from the bundle
[14:47] <srihas> zeestrat: aha
[14:47] <srihas> zeestrat: I will look around it, thank you for the suggestions
[14:47] <srihas> will get back to you
[14:48] <srihas> zeestrat: how can we add custom charms, for example if we use ScaleIO ?
[14:50] <srihas> 16:39 < srihas> zeestrat: is there a way to add all the "ready" nodes to juju
[14:50] <srihas> 16:39 < srihas> like I get them when I do the juju machines
[14:50] <srihas> I think this is needed, I am missing some thing crucial here
[14:55] <zeestrat> srihas: I'm actually not sure. I only use bundles which define which machines I want from MAAS. I'm thinking out loud now, but you can check if you can do a `juju add-machine` which should deploy a machine, then if you do something like `juju deploy nova-compute --to <machine-id>` where <machine-id> is the one you get after the add-machine.
[14:56] <zeestrat> srihas: But I really think you should look at using bundles where you can define constraints which allows you to select which machines you want to deploy on from MAAS.
[14:56] <srihas> aha
[14:57] <zeestrat> srihas: Regarding ScaleIO, are you planing to use the charms from https://jujucharms.com/u/cloudscaling/ or write your own ScaleIO charms? I haven't used them so don't know the state of them.
[14:57] <srihas> zeestrat: the ones from juju, they are old?
[14:58] <srihas> https://github.com/thecodeteam/juju-scaleio -> I looked at them before
[14:59] <zeestrat> Looks like last activity was in Jan 2017 so I'm not sure of the state.
[15:00] <srihas> zeestrat: ok
[15:01] <srihas> zeestrat: is there a guide to write "hello-world" juju charm ? (please bare with me)
[15:02] <zeestrat> srihas: https://jujucharms.com/docs/stable/developer-getting-started
[15:02] <srihas> zeestrat: thank you
[15:05] <zeestrat> srihas: No problem. The bundle docs page talks a bit more about constraints and how to place things on different machines. https://jujucharms.com/docs/2.3/charms-bundles If you tag machines in MAAS you can target those using the `tags` constraint which is one of the constraints available: https://jujucharms.com/docs/2.3/reference-constraints
[15:06] <zeestrat> srihas: I can whip up a short example, but you'll have to remind me next week
[15:06] <srihas> zeestrat: sure, I will get back to you :)
[20:28] <otama> HELP: Hi Team, is it possible declare a Cross Model Relations into a bundle files ?