[05:41] <gaurangt-> hi, I've a requirement where I need to get the specific network IP address in the charm for further processing. I know we can get unit IP address using private-address or public-address but what happens if the unit has more IP addresses dedicated for specific purposes (for example, separate network for storage, management, external access etc)
[06:03] <thumper> wallyworld: ^^ you have dealt with this haven't you?
[06:36] <wallyworld> gaurangt-: the hook tool you want to use is network-get. that should give you a list of all the addresses on the machine hosting the unit. it also takes account of and space related endpoint bindings
[06:37] <wallyworld> unit-get is deprecated
[06:37] <wallyworld> it will just return the one "preferred" address. i'm not sure off hand how that address is determined
[06:48] <wpk> wallyworld: with unit-get the only thing guaranteed is that it will be the same address every time (unless network setup changes)
[06:50] <gaurangt-> wallyworld, thanks for the info. I think network-get is the way forward.
[06:51] <gaurangt-> wpk, but I guess we cannot control which ip unit-get picks up. It can be any one of the many that are available, right?
[06:52] <wpk> gaurangt-: exactly, that's why network-get is the way to go.
[06:53] <gaurangt-> wpk, yep, sounds good. Let me try that out.
[12:49] <gaurangt-> wallyworld, looks like there is no python interface to call network-get
[12:49] <gaurangt-> is it something to be run from cmdline only?
[12:51] <wallyworld> gaurangt-: network-get is the name of the hook tool. it calls the NetworkInfo() api on the uniter facade
[13:35] <ryebot> Does deploying to a subnet work on AWS?
[13:36] <ryebot> Docs seem to indicate it might be MAAS-only, but I'm not 100% sure.
[13:55] <rick_h> ryebot: should for bootstrap constraints and such.
[13:56] <ryebot> rick_h: excellent, thanks
[15:42] <stormmore> o/ juju world
[18:07] <ybaumy> anyone at work is into nutanix?
[18:08] <ybaumy> would be interessted in hearing what this is about and how good it is
[18:09] <ybaumy> ive seen some yt videos and they look very promising
[18:36] <rick_h> ybaumy: not checked it out
[18:44] <ybaumy> rick_h: have you used hyperconverged systems yet? if not take a look at nutanix. i downloaded the community edition today and will take a good look at the hypervisor and so on
[18:44] <rick_h> ybaumy: cool
[18:44] <ybaumy> rick_h: my boss got a presentation of it and is really hyped now. so i have to evaluate
[18:46] <ybaumy> the cool thing is that you have local storage on your nodes and not like ceph only network nodes and the compute is running somewhere else
[18:47] <ybaumy> anyways i think we get easily and appliance for testing. at least i hope so
[18:48] <rick_h> ybaumy: cool, let me know how it works out.
[18:49] <ybaumy> rick_h: sure thing i will report back. i was testing opesntack for several month now. and openstack clearly has its trade offs. so im looking forward to a new cloud solution
[18:51] <ybaumy> that https://www.simplivity.com/ would be another solution which is interessing.
[18:51] <ybaumy> but personally i dont like HP
[18:53] <ybaumy> but HP bought simplivity so its not their product. which is good if you take 3PAR storage for example which was also aquired
[18:53] <ybaumy> and 3PAR storage is really nice
[19:07] <ybaumy> with ssd approaching 50TB and ethernet 100Gbit who need fibre channel and traditional storage anymore
[19:09] <ybaumy> and storage is the problem in our company not compute power so the future looks bright i guess
[19:10] <ybaumy> we are operating so 2010ish currently
[19:15] <ybaumy> http://www.prnewswire.com/news-releases/supermicro-previews-1u-petabyte-nvme-storage-system-supporting-new-ruler-form-factor-for-intel-ssds-at-flash-memory-summit-300501628.html
[19:16] <ybaumy> https://www.youtube.com/watch?v=OcaID9XLHBU
[19:16] <ybaumy> that will kick ass for bigdata
[19:29] <thnkpad> rick_h ping
[19:29] <rick_h> thnkpad: pong
[19:30] <thnkpad> Hiya Rick - did you message marco 'bout that discourse webpage update to Xenial, or nay ?
[19:33] <thnkpad> I have had puri.sm talking about it - that's really why Iam asking.
[19:35] <rick_h> thnkpad: not gotten a hold of yet. Working on it. Apologies, was out for a week last week so behind the 8-ball. My fault
[19:36] <thnkpad> rick_h, wait a second - wasn't there talk of setting up a Launchpad group to cover it while Marco's away ?
[19:38] <rick_h> thnkpad: sec, otp
[23:07] <bdx> are x-model relations supported on 2.2.2?
[23:07] <rick_h> bdx: no, it's a 2.3 thing
[23:07] <bdx> ok, does x-controller come with that too?
[23:08] <rick_h> bdx: unsure, probably not.
[23:08] <bdx> ok
[23:08] <bdx> rick_h: thx
[23:08] <rick_h> bdx: it'll probably be feature flagged in 2.3 as kinds get worked out across providers
[23:08] <bdx> totally
[23:08] <thumper> well kinda
[23:08] <thumper> I think the intent is to not have a feature flag for 2.3
[23:08] <thumper> for CMR
[23:09] <bdx> thumper: CMR will be included in the 2.3?
[23:09] <bdx> not behind any flags
[23:09] <thumper> that is my current understanding of intent
[23:09] <thumper> it won't be perfect
[23:09] <thumper> but should be usable
[23:10] <thumper> wallyworld: is that your understanding too?
[23:10]  * wallyworld reads
[23:10]  * rick_h just wants to play safe in promises :) 
[23:11] <thumper> rick_h: sure
[23:11] <wallyworld> CMR is intended to be in 2.3 without a feature flag
[23:11] <wallyworld> working to get data model and APIs stable
[23:12] <wallyworld> includes cross controller. there may be some tooling missing and polish etc that will be done in a point release
[23:12] <bdx> oh man
[23:12] <bdx> :) :) :)
[23:12] <wallyworld> cross controller is usable now if you run the edge snap
[23:13] <bdx> Ooooooo
[23:13] <wallyworld> there's networking aspects that will require some charm changes
[23:14] <bdx> ok, thats great news none the less
[23:14] <bdx> going to take it for a spin
[23:14] <wallyworld> so not all charms are guaranteed to work out of the box, but most should if they have been written crrectly
[23:14] <bdx> I have the perfect use case for x-controller
[23:14] <wallyworld> try it out and let us know of any issues
[23:14] <bdx> I have fiber to us-west-2 from my datacenter
[23:15] <wallyworld> what charms were you looking to use?
[23:15] <bdx> and no xfer charge to and fro
[23:15] <bdx> wallyworld: my own
[23:15] <bdx> alongside postgresql, redis, elasticsearch
[23:16] <bdx> I wanted to also look at getting a beats/logstash stack going
[23:16] <wallyworld> ok. make sure that you use either network-get or pick "ingress-address" from relation-get data to figure out the address to connect to on the otjer side of the x-model relation
[23:16] <wallyworld> if you use unit-get it won't work
[23:16] <wallyworld> and many older charms still use unit-get
[23:17] <bdx> ok, that good to know
[23:17] <bdx> lol
[23:17] <wallyworld> but just ask if you try it and get stuck
[23:17] <bdx> I'll start looking around to see where I'm using that
[23:17] <bdx> will do
[23:18] <bdx> wallyworld, rick_h, thumper: thx
[23:18] <wallyworld> np, good luck with it
[23:20] <bdx> wallyworld: are there any prelim docs on creating and consuming offers?
[23:22] <wallyworld> bdx: there's been some blog posts, but doc is a little sparse right now as it's all wip. let me dig something up
[23:22] <bdx> ok, thx
[23:28] <wallyworld> bdx: http://paste.ubuntu.com/25428383/
[23:29] <wallyworld> there may be some typos, i just splatted it out
[23:30] <wallyworld> bdx: also, http://mitechie.com/blog/2017/7/7/call-for-testing-shared-services-with-juju
[23:31] <bdx> that is great
[23:31] <bdx> rick_h: nice blog post!
[23:32] <rick_h> bdx: :)
[23:32] <bdx> super exciting stuff
[23:33] <bdx> wallyworld: many thanks
[23:34] <wallyworld> my pleasure
[23:34] <wallyworld> it's a wip
[23:34] <wallyworld> so nit fully cooked
[23:34] <rick_h> the man's put in some major work there so <3
[23:35] <wallyworld> rick_h: the latest is I have a PR where you can suspend/resume individual x-model relations
[23:35] <wallyworld> suspend = run departed/broken hooks etc. resume = run joined/changed hooks
[23:36] <wallyworld> useful if someone needs to be locked out temporarily