=== dendrobates is now known as dendro-afk [07:09] hi alll === atul__ is now known as koolhead11 [09:46] morning === daker_ is now known as daker === DigitalFlux_ is now known as DigitalFlux === dendro-a` is now known as dendro-afk === dendro-afk is now known as dendrobates === godber_ is now known as godber === niemeyer is now known as niemeyer_lunhc === niemeyer_lunhc is now known as niemeyer_lunch === dendrobates is now known as dendro-afk === niemeyer_lunch is now known as niemeyer [17:27] hi everyone [17:28] is there any ubuntu 10.04 HVM based ami available for ec2 [17:29] or how i can run ubuntu 10.04 with Compute Cluster Instance [17:33] jahil: hey .. yeah I think there is [17:34] lemme try to find out === dendro-afk is now known as dendrobates [17:37] not sure about 10.04 though [17:38] jahil: seems only natty has one ami-3ec23e57 [17:38] :( [17:39] why can't you use natty :) [17:39] i tried it but there some issue with adobe air [17:40] unable to install air dpkg -i --force-architecture adobeair.deb [17:40] your cluster compute application involves adobe air ?! [17:40] yes [17:41] on lucid its work fine [17:42] on para-virt instance we have issue with pulseaudio [17:43] but on HVM instance we don’t have audio glitches [17:43] check those instructions ? http://kb2.adobe.com/cps/521/cpsid_52132.html [17:44] this might be useful too (same thing basically) http://www.omgubuntu.co.uk/2010/01/how-to-install-adobe-air-on-ubuntu-64bit-2/ [17:46] tried but no luck seems air 2.6 depends on lzma and xz-utils both are installed but not i386 version [17:47] if i extract deb file and place files manually im getting errors ELF mismatch [17:50] jahil: do you have ia32-libs installed [17:53] yes === daker is now known as daker_ [18:17] Is this the wednesday UEC ubuntu cloud? [18:18] is there anything different today compare with other days in the channel? [18:20] jair: hey [18:20] jair: well yeah .. it starts in 40 mins [18:21] great [18:21] I am getting all my questions and diagrams together :) [18:21] kim0: thank you so much [18:21] woot diagrams :) [18:22] o/ kim0 [18:22] obino: hey how are you :) [18:22] good thanks :) [18:22] will you be at the UDS? [18:22] obino: yeah! will you [18:23] it looks like [18:23] perhaps we can have some time then [18:23] oh sounds great then! [18:23] sure thing [18:23] excellent [18:23] jair: so what kinda questions and diagrams you have ready :) === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates [18:35] SpamapS, [18:35] kim0: I am using the "At least 2 physical systems" topology with two virtual machines [18:35] oops. sorry about intruding on meeting. [18:35] smoser: nah it's not even started yet [18:36] I will be pasting a link to my diagram in a second === erichammond1 is now known as erichammond [18:44] kim0: here is my current environment I am using 64bit 10.10 server edition ====> http://i51.tinypic.com/5c0zfn.jpg [18:47] jair: and what's the question :) [18:49] jair: seeing the physical systems for a 2 system cloud, I'd use the beefier system for node, unless you want to add dozens of nodes later on [18:49] jair: given that the beefier system has KVM support [18:50] well this is a virtual environment they are both running in a 16GB memory and 8 cores IBM laptop very powerful [18:50] w510 [18:51] jair: the node controller in a vm will probably not work as kvm support is not working [18:51] but It is not for production at all, will be more for demonstrations on how the system works, and for me to understand it [18:52] TeTeT: hmm I got a few messages about virtual acceleration not being available [18:52] jair: well, just realize if no h/w virtualization is available, the best you can get is qemu and then a boot takes minutes instead of seconds and any workload will be rather slow [18:53] jair: that is a boot of an instance [18:54] "physical" by definition means not a vm :) [19:00] Hi everyone [19:01] Welcome to ensemble cloud community meeting [19:01] let's run through weekly updates [19:01] Who'd like to start [19:02] well I am new here hello everybody [19:02] jair: Hi there [19:02] * hazmat waves [19:02] we're starting off the meeting with an update on ensemble wiki.ubuntu.com/Ensemble [19:02] I am brand new to UEC I have some experience working with kvm and xen, but mostly of my experience is with debian and vmware [19:02] jair: it's a next gen cloud service orchestration framework .. [19:03] cool [19:03] I will check that link [19:03] jair: I propose we go through the ensemble updates .. then I can try to help with your questions [19:03] hazmat: can you get us started please [19:04] kim0, we've had few things land this past week, i've written up a summary of the development for the past onto the list (Ensemble Development Report #3) [19:04] sure, thanks [19:04] jair: awesome :) [19:04] hazmat: That news report was awesome, thanks for that [19:04] hazmat: Yeah, it was incredible indeed ;) [19:04] but the highlights i think are adding formula-upgrade support, and ensemble status showing workflow. [19:05] since our last cloud weekly meeting [19:05] oh and a few minutes, maybe an hr, ensemble will default to using a natty image [19:06] ah awesome [19:06] cut from today's natty daily [19:06] * kim0 goes through the report on mailing list .. absorbing [19:06] An important highlight is also that the principia formulas at launchpad.net/principia are working again with trunk [19:06] Thanks to SpamapS for that [19:06] yeah, that was great [19:07] definitely, that's huge [19:07] But really, in terms of getting up to speed, I'd really advise reading the whole email from hazmat [19:07] * niemeyer looks for a link [19:07] https://lists.ubuntu.com/archives/ensemble/2011-April/000060.html [19:07] hazmat: when you say ensemble status shows workflow status [19:07] does that show hook triggers, or is that different [19:07] hazmat: Aha, thanks [19:07] kim0, its a bit different [19:08] kim0, internally ensemble tracks the state of all the service units of a service, and if something goes wrong in a hook, it tracks that, depending on which hook it is it will cause an error state for the unit or unit relation. [19:08] koolhead17: hey [19:09] now ensemble status shows the workflow state for units and unit relations.. if their working, the unit is in the started state, and if the unit relation is working its in the 'up' state. [19:09] pulling up an ensemble status output to demonstrate... [19:09] talking about status, any thoughts about exporting a 'traditional' sys v init script like status info for services, that could then be fed into openview, tivoli, other management system? [19:09] hi kim0 [19:09] http://pastebin.ubuntu.com/596634/ - example of ensemble status output [19:09] TeTeT: supp [19:09] koolhead17: join the meeting :) [19:10] TeTeT: No, not at this point.. this is a bit orthogonal to what we're striving for at the moment [19:10] kim0: channel plz [19:10] TeTeT, we'd rather facilitate direct integration of tivoli or monitoring tools and the managed services via modeling them as a formula i think. [19:10] Agreed [19:10] koolhead17: it's right here, we're talking wiki.ubuntu.com/Ensemble [19:10] k [19:10] there's some interesting distinctions on it, wrt to gathering machine level data vs. service level data or container for monitoring purposes. [19:11] the implementation ensemble status command demonstrates how this can be done - it's pretty small [19:11] although it could be refactored to simplify this type of reusue [19:11] out of the box, for example munin-node will collect a bunch of machine level data on memory/disk/cpu/io etc, but we also want to facilitate the munin-postgres plugin to talk the postgres service and get detailed stats on a service level [19:12] hazmat: that involves changes in munin-postgres and the formula right ? [19:13] kim0, it involves configuration for munin-postgres in the /etc/munin/plugins directory on where to find the postgres service, much like the relation we have now between mediawiki or wordpress and mysql [19:14] the hooks can retrieve the relation/remote config and write it out to their local service config files [19:14] interesting [19:14] * koolhead17 thinks of landscape [19:15] so taking mysql which we already have a formula.. it would effectively be the same relationship as mysql establishes with other related services/apps, modulo perhaps some additional privileges for monitoring. [19:15] hazmat: wouldn't that still be monitoring per machine not per service ? [19:16] like if I have 5 wordpress service units .. any way to get global info on the word press service [19:16] instead of on 5 machines [19:16] kim0, we'd be deploying a munin service, and watching all the service units of it.. whether or not we aggregate that really depends on the reporting tool and what it supports [19:16] if supports multi-node service rollups [19:16] kim0: Yeah, there's a lot that can be done there [19:17] kim0: We still haven't even scratched the surface of the monitoring support [19:17] Yeah true [19:17] agreed, definitely been thinking about it.. i think there is also room for additional consideration of machine level monitoring, but we're not there yet. [19:18] We've just landed formula upgrade support, which was a big blocker [19:18] kim0, one could readily implement an integration with munin (assuming it's similar to other tools out there) that provides for topology discovery and monitoring of info as known to ensemble [19:18] our primary focus for the next uds, has been making the formula author story as smooth as possible, and covering core use cases. [19:18] this would build on what is done with "ensemble status" now [19:18] niemeyer: what's meant with formula upgrade support? Giving advice when an old formula breaks, like a syntax check? [19:19] TeTeT, it means live upgrading of service formulas [19:19] TeTeT: No, being able to say "ensemble formula-upgrade mydatabase" and have Ensemble taking care of the upgrade on all service units [19:19] it will deploy a new version of the formula to a machine, extract it, run the formula-upgrade hook first, and continue on. [19:19] new upgrade-formula hook [19:19] ok [19:19] am totally new to ubuntu cloud [19:20] newbie|2: Hi o/ [19:20] newbie|2: hello [19:20] koolhead17: hello [19:20] newbie|2: So are we, so to speak, so welcome! ;-) [19:20] kim0, yes.. thanks.. typo on my part.. the new hook is called 'upgrade-formula' [19:20] welcome to the club [19:20] hazmat: how will it be different from existing solution [19:20] niemeyer: thanks [19:20] newbie|2: we're running through ensemble updates ( wiki.ubuntu.com/Ensemble) === MennaEssa is now known as m3ess4 [19:21] That's a lot of great progress [19:21] koolhead17: What existing solution? [19:21] koolhead17, at the core its a service orchestration framework, that is much more dynamic and allows for real reuse, and a higher level of service management, than what existing configuration management tools offer. [19:22] don't know what to do here. am just watching the process .. its funny [19:22] newbie|2: Enjoy, having fun is an excellent indication [19:22] newbie|2, typically if you have a question you ask, else you sit back and enjoy the show ;-) [19:22] hazmat: sure [19:22] m3ess4: Hi there [19:22] hazmat: i am enjoying this show [19:23] cool newbie|2 :) [19:23] For everyone very new around here .. Be sure to check out http://j.mp/ensemble-docs [19:23] kim0: Hey :) [19:23] kim0: thanks i am clicking the links [19:23] hazmat: niemeyer regarding monitoring as the conversation was going we have munin [19:24] Ok, so in terms of what's coming up, jimbaker has started working on the "exposure" functionality, which will enable formulas to flag ports which they depend on [19:24] so are we having something better more flexible than it :P [19:24] and then will follow up with the "expose" command, which will enable an admin to flag that a service should be exposed (IOW, have its ports made available for external use) [19:24] this introduces two new ensemble commands, to expose and unexpose a service, plus supporting functionality [19:25] * kim0 proposes .. questions are fired once the update is complete @everyone [19:25] The other major area being worked on is being pushed by bcsaller, which is configuration options for services [19:25] the current firewall scheme we have in place is not for production - machines are provisioned on EC2 with all ports open [19:25] well I am sure everybody is very busy with different levels of collaboration to improve the UEC thank you all for your hard work and for helping the new people like me as well. [19:25] as I said I have been reading about UEC, requirements, topology, etc... and decide to put together a testing environment running on a linux debian squeeze 64bit with the "at least 2 physical systems" topology [19:25] This will enable users to provide custom options to formulas [19:25] and the formula will get a new hook giving it a chance to react to such changes [19:26] niemeyer: custom options ? like parameters ? [19:26] hazmat started working on the "resolved" functionality, which is related to problem solving [19:26] koolhead11, more like the management of the service deployment and configuration.. say i have 5 mediawiki nodes connected a load balancer, a memcached, and a db.. if i want to add a sixth unit, ensemble makes it just as easy the first, and will configure all those relations for you, so its added to the loadbalancer rotation, gets its db credentials, etc. [19:26] (if a formula gets into a bad state, the admin can poke at it and mark it as resolved so it can move on) [19:26] kim0: Yes, like runtime parameters [19:26] yeah. that's the nutshell [19:27] resolved lets you manually recover from hook or formula errors. [19:27] are those parameteres provided only at launch time I suppose? [19:27] kim0: No, that's the beauty [19:27] they can be changed at runtime [19:27] Yes, that is beautiful indeed [19:27] niemeyer: can you mention a use case [19:27] kim0: The formula may define a new hook which is fired whenever the configuration changes [19:27] kim0: I can mention hundreds.. :-) [19:27] hehe [19:28] Pretty much anything that the formula author want to make configurable rather than static can be made so through that mechanism [19:28] wow, that sounds very flexible indeed [19:28] E.g. a blog title, a db cache size, etc etc [19:28] hooks get triggered, check the passed value, and act on it [19:29] newbie|2: cool. seems like i have to dig my nose lttle more inside the docu,entation. :P [19:29] sounds great [19:29] *litt;e [19:29] People will be able to say something as simple as "$ ensemble set mydbcache=1024" [19:29] Sorry [19:29] People will be able to say something as simple as "$ ensemble set mydbservice mydbcache=1024" [19:29] koolhead17: ya you are right [19:29] niemeyer: I get it now [19:29] koolhead17: its complicated for me [19:29] the formula author can stipulate what are the options supported, and what they mean [19:30] newbie|2: for me too currently :P [19:30] ensemble feels like layering a management API on top of legacy services .. would you guys agree [19:31] In a nutshell, that's what has been happening in Ensemble development world I guess [19:31] well on top of services in general i'd agree [19:31] koolhead17: just keep on watching the show... :-) [19:31] kim0: s/legacy// [19:31] Yeah .. [19:31] Thanks everyone for the updates [19:31] - open floor - All questions welcome [19:31] kim0: You can pick the latest cool aid and safely make an Ensemble formula for it :) [19:32] niemeyer: which language code is being written :P [19:32] Is there a current way to trigger a hook manually ? [19:32] koolhead11, any language, to ensemble the hooks written by a formula authors are just executables [19:32] koolhead17: The project is being written mostly in Python internally, at the moment [19:32] ensemble coordinates and sequences when to call those hooks [19:33] koolhead17: But, as hazmat says, the *formulas* can be written in any language at all [19:33] seems like have to pickup python ASAP. planning to start playing with django tonight [19:33] koolhead17: Just like packages in Ubuntu don't have a specific language [19:33] awesome [19:33] koolhead17: you can do lots of magic with bash formulas :) [19:33] kim0: +1 [19:33] koolhead17: If you want to develop Ensemble with us (which would be awesome), yes. If you want to develop formulas for Ensemble, no, that's not necessary. [19:34] niemeyer: am in [19:34] all of our examples and about half of the principia formulas are bash [19:34] * kim0 hugs niemeyer :) [19:34] * kim0 hugs koolhead17 :) [19:34] kim0: thanks to you for taking me to this world :) [19:34] actually .. but it never hurts to hug two people hehe [19:34] ;-) [19:34] hehe [19:34] koolhead17: Jump into #ubuntu-ensemble after the meeting then [19:35] done [19:35] jair m3ess4 .. Hey folks .. Did you have questions or comments [19:35] thank you kim0 sorry for being impatient :( [19:35] jair: no problemo [19:36] jair: I've read what you've written [19:36] just didn't find a question yet [19:36] Yeah, this would be a good time if anyone has anything at all to say about Ensemble [19:37] here is my scenario: http://i51.tinypic.com/5c0zfn.jpg [19:37] OK [19:37] niemeyer: jimbaker hazmat bcsaller thanks team! [19:37] 1. does the demo environemt I have works to demo the basics for the environment? [19:37] :) [19:37] kim0, thanks! [19:37] kim0: Thank you! [19:38] i am curious about Ensemble [19:38] I also thank the ubuntu-cloud team and wish one day be smart enough to help [19:38] jair: TeTeT had mentioned you need real/physical/metal machines .. otherwise you will only use "qemu" which is too slow (but good for proof of concept) [19:38] rsthdn: welcome! [19:38] rsthdn: http://j.mp/ensemble-docs [19:38] kim0: perfect that is the requirement I have for now [19:39] rsthdn: feel free to shoot any ensemble questions at me [19:39] kim0: thanks [19:39] rsthdn: the ensemble devs hang-out in #ubuntu-ensemble [19:39] rsthdn: are you currently using the cloud somehow [19:40] kim0: i guess i am totally new on cloud [19:40] rsthdn: no problemo .. aren't we all [19:40] I am new as well but I have some understanding on debian and ubuntu [19:40] 2. can I run windows guests on the uec? [19:41] jair: I think with the current version .. it's possible, but not supported [19:41] kim0: i use ubuntu one for music... is it also a cloud? [19:41] with the next version, it will be supported [19:41] kim0: perfect [19:42] 3. What is the Eucalyptus client? [19:42] Is this the node server? [19:42] has anyone succeeded with natty and lxc and openstack yet? [19:42] or node controller? [19:42] jair: 2 - no windows support in eucalyptus officially, though a google for it might reveal some hacks [19:43] jair: not sure what eucalyptus client might mean myself [19:43] it is on the documentation and I was wondering what does that concept refers to [19:44] is this referring to the node controller system? [19:44] I want to know about zookeeper [19:44] jair: can you point me to that part of the documentation? Out of context I do not understand what it means [19:44] jair: there's the "euca-xxx" cli tools .. also elastic-fox ..etc [19:44] also do you have documentation for UEC in other languages today other than English? [19:45] let me check where I read about the eucalyptus client === dendrobates is now known as dendro-afk [19:46] https://help.ubuntu.com/community/UEC/Prerequisites [19:46] I think kim0 is right [19:46] jair: I don't think much documentation in non-english exists today [19:46] jair: but hey .. you can help write it in your own language [19:46] that's how it works for ubuntu [19:46] everyone contributes something :) [19:46] kim0: once I understand how the UEC works I can help with some spanish translation [19:46] Awesome! [19:47] jair: the main installation steps are https://help.ubuntu.com/community/UEC/CDInstall [19:47] for installing from CD [19:47] in the future with help with some ... [19:47] or https://help.ubuntu.com/community/UEC/PackageInstall [19:47] I guess the chat does not recognize other characters [19:47] for installing from packages [19:47] I am installing from cd [19:48] great .. that's easiest [19:48] so what does the documentation means with eucalyptus client? [19:48] also: 4. why do I need to install euca2ools? [19:49] what does that package offers? [19:49] euca2ools .. is the "client" [19:49] it offers euca-xxx different commands [19:49] hmm is this a CLI tool to control the images in the cloud? [19:50] jair: those are tools .. to launch and control image in UEC yes [19:50] jair: the cd install document I pointed you to, mentions them and how to use them .. [19:50] read it and if you have questions later ask in this channel and wait for an answer [19:51] jair: so the eucalytpus client here is a machine (even a vm) from which to control your UEC [19:51] just watching this show [19:51] jair: I personally refer to it as 'cloud control host' [19:52] which could just be your laptop [19:52] jair: the cloud control host needs the credentials for accessing the cloud. I like to place this on ~/.euca-/ [19:53] jair: with username being usually admin for the first default user [19:53] jair: it also needs the eucalyptus tools to access the cloud, this is the need for euca2ools [19:53] got it [19:54] so looks like we will need to work on some name consistency I like eucalyptus client (my laptop, my desktop, my android mobile) [19:54] with the euca2ools [19:55] I will check the https://help.ubuntu.com/community/UEC/CDInstall and https://help.ubuntu.com/community/UEC/PackageInstall I like to have understanding on the different scenarios [19:56] so people running mac or windows clients will not be able to install the credentials to manage the UEC client at this moment will be supported only on gnu/linux clients [19:56] jair: generally the CD based install is easier for a first timer [19:57] jair: if you want to use windows/mac, you could either use the ec2tools from amazon or a graphical frontend like elasticfox [19:57] jair: albeit I have next to no experience with these tools on Windows/Mac [20:06] great information I will probably give it a try [20:07] What is the difference between public and private cloud? [20:07] shit [20:10] rsthdn: i would suggest you to do a simple google for that :P [20:11] rsthdn: yeah something like http://en.wikipedia.org/wiki/Cloud_computing [20:11] Public: can be accessed from anywhere in the world without any vpn, special port, or firewall rule restriction. And most of the time is managed by a big company not by a small business or individual. [20:15] Private: you manage the environment in your network no one know about the access this cloud but internally in your company or home network [20:17] is landscape a tool we can use in the UEC for free? [20:17] if I want to use a cloud management platform is this eucalyptus or landscape or this are two different things? [20:17] jair: as far as i know 5 instances can be managed with free account [20:24] koolhead17: it is provided by canonical only correct? [20:24] I guess I will need to ask canonical [20:24] what I am doing right now is only demonstration environment [20:24] I am using both virtual machines [20:25] jair: true [20:25] jair: it will work for you even in that case i suppose [20:45] nice [20:46] can I run other images in UEC other than the mediawiki and ubuntu ones? for example debian, redhat, sles? [20:47] jair: i think you can only if those images are bundled. [20:48] jair: or you can use EC2 images and import them into the UEC [20:48] or build them yourself if you have the expertise [20:51] jair: http://j.mp/h2CEY5 see they have explained bundling as well i suppose. [21:22] koolhead17: thank you so very much [21:22] obino: thank you so much [21:22] I will defenitely lookinto that === dendro-afk is now known as dendrobates [21:23] jair: mention not. its barter of knowledge here :D === dendrobates is now known as dendro-afk