15:32:34 <mizmo> #startmeeting
15:32:34 <zodbot> Meeting started Sun Aug 11 15:32:34 2013 UTC.  The chair is mizmo. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:32:34 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:32:34 <flock-ectr-114> (using this one might make the screen a bit more shaky)
15:32:37 <mizmo> go for it :)
15:33:24 <flock-ectr-114> should I also use the meeting commands, or just quote?
15:33:36 <mizmo> flock-ectr-114, just type :)
15:33:39 <mizmo> dont worry about meeting commands
15:33:39 <flock-ectr-114> okay :)
15:34:19 <flock-ectr-114> What I ave been working on lately ios OpenLMI
15:34:22 <flock-ectr-114> Let me start sat this
15:34:24 <flock-ectr-114> Is this you?
15:34:33 <flock-ectr-114> How many people here have been managing a linux system?
15:34:55 <flock-ectr-114> Give me an example of what you had to do
15:35:12 * nb doesn't really know much about the streaming stuff
15:35:15 <nb> spot is the guru
15:35:22 <flock-ectr-114> nb, already fixed I guess
15:35:25 <flock-ectr-114> we're live
15:35:26 <ianweller> i am too, what's up
15:35:48 <ianweller> oh, well, it's working :P
15:36:06 <nb> flock-ectr-114, spot says it is fine on the other end, it only is like that on the local client
15:36:13 <flock-ectr-114> So yeah,when youre managing storage, youre working with speceific tools like parted, pvcreate
15:36:20 <flock-ectr-114> Did anyone have to modify networking on a linux system?
15:36:24 <flock-ectr-114> nb, I hear already
15:36:31 <flock-ectr-114> (more interaction with the audience)
15:37:08 <flock-ectr-114> Trying to get the audience to name tools used for network management
15:37:26 <flock-ectr-114> Basically, this is the entirety of how we manage networking: we hack on a series of config files and restart services
15:37:31 <flock-ectr-114> What do you notice about these examples?
15:37:45 <flock-ectr-114> Both of them require access to shjell on the machine?
15:37:53 <flock-ectr-114> The methods that we use are unrelated
15:37:56 <flock-ectr-114> There is no overlap
15:38:08 <flock-ectr-114> The list goes on, we have systemd, yum, firewalld, performance tuning
15:38:17 <flock-ectr-114> All of these things have completely unique ways to manage
15:38:32 <flock-ectr-114> A lot is historyical, the community grow from multiple projects that all wanted to do one thing
15:38:46 <flock-ectr-114> All projects tried to create a tool that did just what their project did
15:39:12 <flock-ectr-114> The problem is that it provides a very incoherent collection of independant bits that happen to be from the same install dvd
15:39:15 <flock-ectr-114> The problem is you can
15:39:21 <flock-ectr-114> you cant manage a complete system
15:39:42 <flock-ectr-114> You learn how to manage firwalld, you learn storage, but you dont know anything about the other things like service management
15:39:51 <flock-ectr-114> thats why theres so few junior linux admins
15:40:16 <flock-ectr-114> thats why its so hard to get unix sysadmins
15:40:35 <flock-ectr-114> Red Hat did a research to get pain points.
15:40:54 <flock-ectr-114> everyone says the problem is that its so dificult because of the differences between all tools
15:41:05 <flock-ectr-114> So we took that feedback and reaslized thats important to know
15:41:27 <flock-ectr-114> If we want Linux to expand into the market and get into the market thats dominated by Windows, we need to make it more coherent
15:41:32 <flock-ectr-114> thats where OpenLMI comes in
15:41:41 <flock-ectr-114> We got a great way with Fedora 19
15:42:12 <flock-ectr-114> When you learn services through LMI, you know how t omanage storage through LMI
15:42:31 <flock-ectr-114> There's been many projects that tried to do this
15:42:49 <flock-ectr-114> They were all trying to invent their own technologies
15:43:04 <flock-ectr-114> With OpenLMI, we build everything upon standard protocols and standards
15:43:16 <flock-ectr-114> To break down
15:43:23 <flock-ectr-114> We built a number of tehcnologies to make this possible
15:43:48 <flock-ectr-114> Forgot to mention: the problem with editing config files and the tools, its very easy to get an inconsistent state by making a typo
15:44:03 <flock-ectr-114> One of the first things we did was storage
15:44:15 <flock-ectr-114> We looked at how its currently implemented: gparted, vg stuff, etc
15:44:33 <flock-ectr-114> anaconda has the largest subset of managing storage, because to install you need to mount and partition drives
15:44:52 <flock-ectr-114> Agreed with anaconda team to split the storage parts into Blivet
15:45:24 <flock-ectr-114> Blivet is a python API that manages the storage from the getting bootloader on drive up to mounting
15:45:32 <flock-ectr-114> Blivet is very consistent in itself
15:46:21 <flock-ectr-114> Internal consistency meaning that operations cannot break stuff without meaning to
15:46:30 <flock-ectr-114> So it should be hard to break without intention to do so
15:46:48 <flock-ectr-114> We've been working with NetworkManager to get it the default way to manage networks
15:46:52 <flock-ectr-114> There's a lot of history
15:47:02 <flock-ectr-114> It came on the scene as Wifi for laptops
15:47:14 <flock-ectr-114> VLAN and Bonding has been added to networkamnager
15:47:26 <flock-ectr-114> These features are made consistent by the NetworkManager API
15:47:40 <flock-ectr-114> Weve been working with NM-teams to get the same API and features in OpenLMIU
15:47:52 <flock-ectr-114> Very controversial as well: SystemD
15:47:57 <flock-ectr-114> it took some time
15:48:06 <flock-ectr-114> It broke a lot of things, including expectations
15:48:21 <flock-ectr-114> Everything was ad-hoc in upstart and sysv
15:48:39 <flock-ectr-114> with systemd we have a comprehensive API that tells whats running and how long and what its been doing
15:49:09 <flock-ectr-114> Weve been working with systemd team to get the apis from systemd in the shape we need for OpenLMI as well
15:49:29 <flock-ectr-114> We chose to use existing protocols and standards
15:49:34 <flock-ectr-114> What we chose was DMTF-CIM
15:49:54 <flock-ectr-114> Its the protocol around most management tools in the world, like server vendors for VMCs
15:50:03 <flock-ectr-114> its used for nearly every SAN as well
15:50:19 <flock-ectr-114> Chances are that if all those vendors chose it, should be good
15:50:42 <flock-ectr-114> CIM is extremely complex
15:50:55 <flock-ectr-114> Normally I get in with the 50 pages that are the table of index for the CIM manual
15:51:01 <flock-ectr-114> It can be intimidating
15:51:39 <flock-ectr-114> CIM models a complete computer system, not just hardware, but also all software components
15:51:57 <flock-ectr-114> It's backed by the Dell's, Cisco's, Microsoft's
15:52:09 <flock-ectr-114> One of our goals is to simplify things
15:52:31 <flock-ectr-114> One of the things Microsoft did well is powershell: write these powerfull shell scripts that interact with CIM models and do useful work
15:52:38 <flock-ectr-114> We wanted to do something similar, but do it better
15:53:05 <flock-ectr-114> We decided to standardize on Python
15:53:24 <flock-ectr-114> We were building specialized specific python modules to implement the most-often used CIM features
15:54:07 <flock-ectr-114> We're building these now, I've got some examples to show during the hackfest after lunch
15:54:19 <flock-ectr-114> We only have query so far
15:55:00 <flock-ectr-114> There are two tings: (slide is old) we have python modules which can be executed themselves, bundled into a command-line app, or can be used in any other python script
15:55:13 <flock-ectr-114> A lot of backend work is already inf Fedora 19
15:55:27 <flock-ectr-114> it's going to be long before we hit 100% coverage
15:55:40 <flock-ectr-114> But we have hit the 80% most-used coverage
15:55:53 <flock-ectr-114> We are not going for 100% coverage by ourselves, but we hope to get contributors to help on that
15:56:04 <flock-ectr-114> The next thing were working on is the LMI scripts
15:56:14 <flock-ectr-114> Theyre very easy to build
15:56:29 <flock-ectr-114> We need to document the creation better
15:56:57 <flock-ectr-114> We have a lot of coverage, but we have issues with storage stuff thats not needed for installation
15:57:13 <flock-ectr-114> so creating volumes is in, but extending not yet
15:57:26 <flock-ectr-114> With the current available tools we cannot setup network storage
15:57:44 <flock-ectr-114> We are working on ways to do this, but the next steps are to add support for initiating the connections
15:57:57 <flock-ectr-114> We can work with anything where we already have device mappings for
15:58:16 <flock-ectr-114> The other thing is getting a proxy service in LMI to the storage vendors
15:58:21 <flock-ectr-114> We hope to have this in Fedora 20 or 21
15:58:48 <flock-ectr-114> He is staring at the packagekit maintainer
15:59:06 <flock-ectr-114> We do not have the metadata to get if a specific bug is fixed on packages installed on a system
15:59:13 <flock-ectr-114> We want to make that available
15:59:23 <flock-ectr-114> So that you can ask LMI "is this bug fixed on this system"?
15:59:47 <flock-ectr-114> A demo of setting up a RAID array
16:00:07 <flock-ectr-114> I'm calling lmishell which is awrapper arround python with some imports
16:00:12 <flock-ectr-114> It checks the current config
16:00:33 <flock-ectr-114> We create a partition table create the partitions, create the raid system and then verify its done and working
16:00:38 <flock-ectr-114> All of that was performed over network
16:00:42 <flock-ectr-114> This is real working code
16:01:01 <flock-ectr-114> The other demo is on interacting with oenlmi to get service information from systemd
16:01:13 <flock-ectr-114> We connect to a service
16:01:24 <flock-ectr-114> I want to get services that start at boot and their current status
16:01:33 <flock-ectr-114> For purpose of dempo, the output is slowed down
16:01:41 <flock-ectr-114> then we restart a service remotely
16:01:54 <flock-ectr-114> All was over the network
16:02:06 <flock-ectr-114> No direct shell
16:02:51 <flock-ectr-114> The demos are in tinyurl.com/openlmi-demo
16:03:17 <flock-ectr-114> We need info and contributions from community
16:03:26 <flock-ectr-114> - help us plan the API, what do we need to have RIGHT NOW?
16:03:38 <flock-ectr-114> What are the most important things you need?
16:03:48 <flock-ectr-114> - Provide a public API
16:04:01 <flock-ectr-114> it doesnt need to be LMI, but it should be a public python API
16:04:19 <flock-ectr-114> We are not trying to do anything of the actual business logic in OpenLMI, its only a compatibility layer
16:04:43 <flock-ectr-114> One of the pushes is to encourage to build reusable APIs
16:05:09 <flock-ectr-114> - Contribute these LMI modules (called scriptons previously, but not allowed by legal anymore)
16:05:22 <flock-ectr-114> so now called LMI (python) modules
16:05:32 <flock-ectr-114> It is in Fedora 19
16:05:39 <flock-ectr-114> Try it!
16:05:50 <flock-ectr-114> Tell us what you like or hate, and give feedback
16:05:59 <flock-ectr-114> With that, I will take questions
16:06:05 <flock-ectr-114> (any IRC questions?)
16:07:59 <flock-ectr-114> #info Webpage: www.openlmi.org
16:08:10 <flock-ectr-114> #info Mailing list: openlmi-devel@lists.fedorahosted.org
16:08:20 <flock-ectr-114> #info bug tracker: https://fedorahosted.org/openlmi/
16:08:34 <flock-ectr-114> #info HOWTO: https://fedorahosted.org/openlmi/wiki/Hosto
16:08:45 <flock-ectr-114> #info Demo http://tinyurl.com/openlmi-demo
16:11:28 <flock-ectr-114> #endmeeting
16:11:34 <flock-ectr-114> mizmo, please end it
16:11:49 <flock-ectr-114> or anyone else, but I'm no chair
16:15:40 <mizmo> #endmeeting