[00:22] <skay> I'd like to use juju with a rackspace account. based on a quick search, people use manual provisioning?
[00:22] <skay> if there is another approach, or if someone has tips, please let me know!
[00:28] <marcoceppi> skay: not at the moment. The Rackspace openstack setup is to heavily modified from upstream openstack so our openstack provider code won't work against it
[00:28] <marcoceppi> skay: it's possible, with some code, to wrap the rackspace API and provide a "manual provider" like was done with digital ocean, but someone would have to do that
[00:30] <skay> marcoceppi: I may not bother for the rackspace stuff. thanks for telling me about it
[00:30]  * skay has donated rackspace services to work with
[14:29] <arbrandes> Hey guys!
[14:29] <arbrandes> Is there a reason why many Juju/OpenStack/HA guides recommend Percona over, say, MariaDB or MySQL?
[18:29] <captine> Hi all.  Am a bit of a tech hobbiest and am looking at using a mac mini server for either a hypervisor like proxmox to play with containers and vms, or installing MAAS and learning a bit of juju.  anyone done anything like this and have advice?
[18:37] <jcastro> captine, we've used juju and maas with intel nucs before
[18:37] <jcastro> minis should work fine if it's similarly powerful hardware
[18:38] <captine> jcastro, mini has bios based raid0 2x1TB HDD, 16Gig ram and i7 processor.  only have 1, so hence was thinking of using proxmox to then run 3 or so VMs to mimic MAAS.
[18:39] <jcastro> yeah that sounds much more powerful than the i5 nucs we have
[18:39] <captine> can MAAS and juju give me similar virtualization ability to a hypervisor?
[18:40] <jcastro> I've not done that personally
[18:40] <mbruzek> captine: MAAS can *manage* those VMs that you create on the mini, and by that I mean install Ubuntu on them when the boot pxe
[18:41] <mbruzek> captine: Once juju is orchestrating the MAAS you can deploy charms and or bundles to them.
[18:42] <captine> mbruzek, thanks.  I think i will go with a hypervisor + some "MAAS" vm's.  that way, if I understand you correctly, I can run a windows vm or other linux distro's/ubuntu flavours which doesnt sound possible with MAAS only.
[18:42] <captine> appreciate the responses all.
[18:44] <mbruzek> captine: Your original idea looked fine to me, I have 2 intel NUCs (i5) and created 8 VMs on one of them.  Then set up MAAS to manage those VMs.  Then I have juju controlling the MAAS
[18:45] <mbruzek> captine: I believe the latest version of MAAS can do other linux distros, but you will have to check the documentation on that, I haven't played with that yet.
[18:47] <mbruzek> captine: I am not a MAAS expert, but I have set it up on my NUC machines, several versions ago.  For more information on that you might want to ask in #maas
[18:48] <captine> mbruzek, thanks a mil.  what were you using for hosting the vms?  e.g. hypervisor wise?
[18:49] <mbruzek> captine: I used KVM.  But as long as the VM can boot pxe (from network) and MAAS controls that network interface it will set up the VMs for you.
[18:52] <mbruzek> captine: So my setup is:  NUC1 Ubuntu 14.04 with Juju and MAAS installed.  NUC2 Ubuntu server 14.04 with KVM, and 8 VMs created with 2GB of RAM each.  The two NUCs are connected via a switch and MAAS controls NUC1's eth0 adapter.
[18:53] <mbruzek> captine: NUC1 has a separate wireless adapter that I use to connect to the Internet, and all traffic is routed through there.  In the MAAS documentation you will see it recommended to have more than 1 network adapter.
[18:54] <mbruzek> captine: It is worth noting that MAAS can be installed on a VM itself, but again needs to control a network interface so it can send boot images over pxe.  I have not done this, but version 2 might be set  up with MAAS in a VM.
[19:03] <captine> awesome info.  thanks
[22:34] <marcoceppi> mbruzek: captine you'll need more than pxe boot because Maas needs to be able to turn the vms on. So you'll need anything that works with libvirt/qemu
[22:50] <designated> I have a service in a dying state that will not remove.  What can I try besides destroying the machine?
[22:51] <designated> I've tried "juju destroy-service ceph" and "juju destroy-unit ceph/0", neither produce a result or any logs.
[22:53] <marcoceppi> designated: is it in an error state?
[22:55] <designated> marcoceppi: agent state is started
[22:55] <designated> marcoceppi: life shows dying
[22:56] <marcoceppi> designated: okay, was anything related to it?
[22:56] <designated> nope
[22:57] <marcoceppi> designated: can you pastebin your juju status output?
[22:58] <designated> marcoceppi: http://pastebin.com/Mm3Tg4ar
[22:59] <designated> marcoceppi: I recently deployed ceph-osd but no relationship was built between ceph and ceph-osd
[23:00] <designated> marcoceppi: this all started out of the ceph charm not forming a ring due to it not finding the keyring.  So I tried to destroy the ceph service and reinstall.
[23:00] <designated> well tried to destroy, have to get past that first.
[23:00] <marcoceppi> designated: there are probably hooks still running in the machines
[23:01] <designated> marcoceppi: it's been at least an hour
[23:02] <marcoceppi> You can be patient (lame) or terminate-machine --force 2 3 4
[23:02] <marcoceppi> designated: a him might be stuck and blocking
[23:03] <marcoceppi> That will destroy the service since there are no units at that point and the service is dying
[23:03] <designated> marcoceppi: i want to avoid destroying the machine if possible as this will not be a suitable solution in production.
[23:08] <designated> marcoceppi: so i have to destroy the whole machine because one service messed up?
[23:28] <designated> For anyone that might be interested, SSHing into the machine and restarting the problem juju agent allowed me to resolve the unit and eventually destroy the service as a whole.
[23:28] <designated> does james page hang out in here?