[00:16] <Hey__> is it possible to deploy windows 10 as a juju charm on bare metal?
[00:18] <bdx> Hey__: technically yes
[00:18] <Hey__> hmm.
[00:18] <Hey__> I would expect technically.. yes
[00:18] <Hey__> bdx: would you say this is not a very useful application of juju?
[00:19] <bdx> Hey__: read the "Enterprise support for MAAS" section https://maas.io/
[00:20] <bdx> Hey__: its a great way to deploy Win to bare metal .... you just have to have support, if get the support package then its possible and super useful for those doing windows deploys
[00:20] <Hey__> point taken.
[00:20] <Hey__> understood.
[00:21] <Hey__> bdx: your saying it's only possible with support?
[00:21] <bdx> Hey__: for sure it is
[00:21] <Hey__> is there a feature that is not enabled.. or you saying I'm gonna have to roll my own?
[00:22] <bdx> neither ... you must pay $$
[00:22] <bdx> you can use maas all day long to roll your own deploys of ubuntu and centos
[00:23] <Hey__> bdx: I noticed that.
[00:23] <Hey__> I was reading that I need to customize an init for it to work with Maas
[00:23] <bdx> for what to work with MAAS?
[00:23] <Hey__> Creating the windows image I wish to deploy
[00:27] <bdx> ahhh yeah ... you may encounter many unforeseen issues and processes that are not documented trying to do the things in the $$ tiers
[00:27] <bdx> but hack away!
[00:27] <bdx> don't let me stop you :)
[00:29] <Hey__> lol
[03:46] <bdx> does JAAS have controllers in every region? e.g. if I spin up an instance in eu-west-2, will the Juju agent on that instance talk to a JAAS controller in eu-west-2, or are there only controllers in some regions?
[12:07] <SuneK> kwmonroe, how is your work on vsphere support faring? The 1.9.4 upgrade breaks vsphere cloudprovider support for CDK, so I'm pretty much in limbo here.
[17:03] <rick_h> one hour heads up until Juju show #31 kwmonroe hml bdx cory_fu magicaltrout and anyone else interested
[17:03] <rick_h> cory_fu: going to cover the charm changes if you can make it
[17:12] <cory_fu> rick_h: Sure
[17:13] <cory_fu> rick_h: By charm changes, you mean the changes to layer:basic?
[17:13] <rick_h> cory_fu: rgr, call out to folks on actions, etc
[17:13] <rick_h> cory_fu: does that effect juju run as well?
[17:14] <cory_fu> rick_h: Should be completely transparent to juju run
[17:14] <rick_h> cory_fu: cool
[17:15] <cory_fu> Except in the very edge case of charms installing Python libs that have CLI tools that they use
[17:16] <rick_h> cory_fu: right, like a juju run to execute a django scipt or something is what I was thinking
[17:34] <cory_fu> rick_h: So the work-around would be: juju run unit/0 -- charm-env myscript args
[17:35] <rick_h> cory_fu: yea, or setup make to call the script with the right env and such. Just want to note it out so folks recognize 'wtf'
[17:39] <rick_h> juju show 20min warning
[17:39] <rick_h> get your coffee and tea ready to go
[17:43] <rick_h> the "join the fun" url is https://hangouts.google.com/hangouts/_/atvgoe55qra33nvrw5ogthx6eye
[17:43] <rick_h> and the "watch the fun" url is https://www.youtube.com/watch?v=Zp9DIQ-0NI4
[17:57] <cory_fu> ryebot: Juju show?
[17:57] <cory_fu> Cynerva: Perhaps?
[17:57] <rick_h> party people, starting in 3...
[17:57] <ryebot> can't, stuck on phone internet until next week
[17:57] <rick_h> bdx: what kind of coffee are you making :P
[18:00] <wpk> hm, how to set TRACE logging for some modules for bootstrap?
[18:14] <zeestrat> Hey question for the show. Any hope for some charmhelpers docs love? https://github.com/juju/charm-helpers/issues/27 Where are y'all getting your docs fix currently if not on pythonhosted.org/charmhelpers?
[18:23] <kwmonroe> doh!  sorry zeestrat, we missed your question.  bad rick_h.
[18:24] <rick_h> oh damn
[18:24] <rick_h> sorry zeestrat! I don't know if cory_fu are we using the pythonhosted docs or I know there's some patches going into the jujucharms.com docs around this lately?
[18:26] <cory_fu> rick_h: I noted on that issue that PythonHosted.org is deprecated, and it's actually reached the point that you can't even update docs on it any more.  We need to move to ReadTheDocs.io, but I don't have access to that repo to enable it.
[18:26] <rick_h> cory_fu: oooooh
[18:27] <cory_fu> rick_h: It would also be really great if jujucharms.com/docs/ could also host Python library docs, but they're in a completely different format have different themes, so I don't see that being feasible any time soon
[18:28] <rick_h> cory_fu: yea, so the jujucharms.com/docs are moving to docs.jujucharms.com and on the standard ubuntu docs setup
[18:28] <rick_h> cory_fu: like maas and other products
[18:28] <cory_fu> Maybe we should open a jujucharms.com doc issue to create a Sphinx theme that matches JujuCharms.com and then have a way to hose the library docs there
[18:28] <rick_h> cory_fu: but yea, don't think it's going to be readthedocs compatible that I'm aware of
[18:28] <rick_h> cory_fu: bug antdillon about it and see if there's something they can do
[18:30] <cory_fu> rick_h: RTD is pretty much the standard for Python libraries now, anyway, so I don't think it's terribly urgent to have them co-hosted on docs.jujucharms.com and it would be quite a bit of work
[18:30] <cory_fu> Still, having a Juju theme for Sphinx in and of itself would be pretty awesome
[18:31] <rick_h> cory_fu: yea just not sure what's easy vs hard with their setup. Figure it's worth a mention.
[18:32] <cory_fu> zeestrat, rick_h: Oh, shoot.  It looks like javacruft might have given me access when I originally commented on that issue and I just never realized it
[18:32] <zeestrat> Oops!
[18:32]  * cory_fu goes to enable RTD for charmhelpers.
[18:34] <zeestrat> Awesome. Maybe check out if there's any way to redirect from pythonhosted or remove it somehow to optimize those Google results
[18:40] <cory_fu> zeestrat: Unfortunately, I've looked into that already.  There's no way to do the redirect.  Best I can do is remove the docs from PythonHosted entirely and add a link to the new docs in the README
[18:53] <bdx> https://unixsurplus.com/collections/4u-servers-1/products/freenas-server-36-bay-supermicro-4u-2x-e5-2680-2-8ghz-192gb-2-port-10gbe-sfp-nic - these are the chassis I'm using for my ceph
[18:54] <bdx> 5 x ^
[18:56] <rick_h> bdx: niiiiice
[18:56] <rick_h> bdx: full disks ?
[18:56] <bdx> yeah 6T
[18:56] <bdx> 36*6T*5 = 1080T
[18:57] <bdx> not sure how it will be replicated and all, so I don't know what my usable will be .... possibly will have to swap out some of the hhd for ssd to use as journal, cache pool etc
[18:58] <bdx> I'll be doing a little write up on it
[18:58] <bdx> rick_h: I'll be sure to keep you posted
[18:58] <zeestrat> bdx: On their way or already got them? We've been running pretty similar chassis/MB for a about a year with ceph.
[18:58] <bdx> on thei rway
[19:00] <bdx> zeestrat: are yours filled with spinning disk only?
[19:05] <zeestrat> bdx: Yeah, we've been doing spinners only for that seutp. Got 12 boxes with 12x8TB regular 7.2k and 12x8TB SMR archive disks in each. 8 more on the way now. The latter disks are a fun experiment in trying to get real cheap ceph workloads on SMR disks.
[19:05] <bdx> I see
[19:07] <zeestrat> We've been looking at how we want to do ssd pools and will probably end up with other nodes with just ssd's though you quickly end up hitting network bottlenecks there.
[19:07] <bdx> zeestrat right ... been there
[19:08] <zeestrat> Used a lot of SuperMicro before?
[19:08] <bdx> yeah
[19:09] <bdx> I've had to beef up the networking for ssd ceph clusters too .... totally hit that first hand
[19:10] <bdx> zeestrat: I ordered them full of spinning disk so I can get baseline metrics with full spinning disk with journal on the osd drive
[19:10] <bdx> zeestrat: then I'm just going to iterate from there
[19:11] <bdx> making one change at a time and benching inbetween each change
[19:17] <zeestrat> Sounds like a plan. Will be fun to hear about your experiences. The ceph charms have been getting a bit more love lately.
[19:18] <bdx> yeah, so legit
[19:19] <bdx> zeestrat: I have 5 racks all in a row, my plan is to just build ceph, and the rest of the openstack evenly across the 5 racks
[19:20] <bdx> the ceph nodes across the bottom, 10G networking across the top, and compute in the middle
[19:20] <bdx> I've already go the HA MAAS going on (deployed via my maas charms)
[19:21] <bdx> I know there will be a workload upgrading that maas and doing the whole bionic dance
[19:21] <zeestrat> Yeah that's pretty much what we did, though started with 3xHA over 3 racks
[19:22] <bdx> cool
[19:23] <zeestrat> HA in MAAS for region controllers was more pain than I wanted though when we started with Maas 2.1 so planning to just drop that to 1 region and just do multiple rack controllers
[19:23] <bdx> zeestrat same
[19:23] <bdx> its much nicer now in 2.3
[19:23] <bdx> but there is still the proxy issue
[19:24] <bdx> in the sense that you have to mitigate things manually that aren't immediately obvious when setting up ha maas
[19:25] <bdx> zeestrat: I tried to address a good amount of those paper cuts and bumps in the road in my rendition of the maas charm
[19:25] <bdx> zeestrat: https://jujucharms.com/u/omnivector/maas/1
[19:26] <zeestrat> Cool. I'll check it out. I guess I was hoping for something a bit more opinionated or ready packaged but I guess I'm just spoiled.
[19:26] <zeestrat> *regarding the maas ha docs that is
[19:27] <bdx> zeestrat: right, they want paying customers
[19:27] <zeestrat> :)
[19:27] <zeestrat> Not us freeloaders right?
[19:28] <bdx> aha right
[19:38] <zeestrat> Speaking of SuperMicro. The cluster I mentioned above was my first rodeo with them (both using and procuring) so I presumed those BMC's would have a nice and easy way to update BIOS firmware from the CLI or web just like iDRAC which I was used to. Was a fun day at the office (and not DC) and I wanted to do just that and I found out you need buy this separate OOB license to do that. So now every time I hear someone
[19:38] <zeestrat> (like you mentioned above) buying SuperMicro, I instinctively want to tell them "remember to buy those damn licenses".
[19:55] <bdx> zeestrat: thanks for the heads up, I'lll look into this
[19:58] <zeestrat> bdx: sft-oob-lic is the part#. Everything else works fine in the web interface, but BIOS (and perhaps BMC itself, I can double check if you want) upgrades via web are locked down. There are workarounds/hacks, but those licenses paid for themselves pretty quickly.
[19:59] <bdx> zeestrat: how do I know if I need it? Possibly it is already the latest?
[20:12] <zeestrat> bdx: We have a bit newer MB and I'm having a hard time finding anything on the SuperMicro specs page about what's included with OOB mgmt. It wasn't included with ours (X10DRi-T) though that's might be just our VAR forgetting.
[20:14] <bdx> zeestrat: this sounds like a darn rat trap
[20:15] <zeestrat> bdx: Welcome to supermicro.com. If it's important on day 1, I'd ask your seller/VAR or just support@supermicro.com directly.
[20:16] <bdx> zeestrat: so .. yeah purchasing these from unixsurplus .. outside of our vendor
[20:17] <bdx> zeestrat: I have 1 of them already ... it appears to be functioning just fine, how can I tell if I need the update?
[20:20] <cory_fu> zeestrat: https://github.com/juju/charm-helpers/pull/131
[20:29] <zeestrat> cory_fu: Great stuff. Thank you very much!