[01:28] * SpamapS drums fingers on tray table while waiting for natty t1.micro to boot. :-P [01:30] * SpamapS realizes he's just dumb and didn't allow incoming SSH [01:42] hehe === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk [19:49] I would like to run a VM in amazon or something like that. I'm well aquainted with KVM and similar, but I haven't paid much attention to the cloud. What I want to do now, is to create a local VM, setup all systems and test it for a while and then upload it to some cloud. What do I need to know? [20:46] anyone around? === erichammond1 is now known as erichammond [22:21] jo-erlend: well ec2 is xen and you don't need to make AMIs but you can if you want [22:21] bhook: no [22:24] flaccid, what is an AMI? Can I just make my system as I want it as a KVM instance and transfer it? Or is there some sort of conversion I need to perform? [22:25] jo-erlend: might be a good idea to read the getting started guide so you are familiar with terms such as AMI [22:25] there is no kvm in ec2 [22:25] flaccid, I understand that. And I'm not sure I want to pick Amazon. Is there a uniform way for me to handle these cloud things? [22:25] there are many images (AMIs) public that you can use, no image building required [22:26] yes, but I want to prepare it locally first. [22:26] jo-erlend: well i work at RightScale and we take the MultiCloud approaching of supporting more than just aws [22:26] why do you want to prepare it locally and why do you want to build an image? [22:26] ubuntu has images already available [22:27] I want to do it locally because I see no point in paying for a service when I can provide a better one myself, which I can in the time I spend preparing things. [22:27] jo-erlend: please read about UEC then, it comes with Ubuntu i.e. Eucalyptus [22:28] good luck trying to provide yourself a better service than aws though [22:29] local file access is far faster than the internet is. As long as I don't need any kind of external network, obviously my local server on my gigabit LAN will be better than placing my VM on the other side of the planet. [22:29] sure, that doesn't mean its a better 'service' [22:29] anyway good luck with UEC [22:30] for me? Obviously it is. But is it really necessary to setup an entire cloud for just one VM to prepare websites, databases, etc? [22:31] no [22:31] you can just use something like VirtualBox or VMWare Workstation or whatever [22:31] but then I cannot upload it to any clouds? Because that is what I've been asking how to do. [22:32] you can if you know what you are doing [22:32] jo-erlend: you can make images in KVM and upload them to a cloud service [22:32] ec2 also provides VM Import and also uploading to EBS now. [22:32] jo-erlend: as long as the image isn't using any KVM/Xen specific features [22:32] well the image needs to be xen compatible i.e. the kernel [22:33] flaccid: on my images, the kernel isn't even in the image - the kernel is a separate image [22:34] flaccid: kernel modules, if needed, are in the disk image, but the kernel itself is separate [22:34] well if you want to use your own kernel, pvgrub is required where the kernel is in the AMI. [22:34] bhook: i build images for a living, so i do know all of this [22:35] flaccid: I don't doubt that you do, but you obviously have a very limited way of tackling problems. There are many flexible ways to do this stuff [22:35] which kernel do I need then? I suppose it's in the repos? [22:36] has to be ten years since I used xen :) [22:36] bhook: wtf. limited way of tackling problems? i presented more solutions than you. [22:36] jo-erlend: any kernel you like. read the user_specified_kernels.pdf doc if you want to use your own, or you can choose between the public AKIs [22:37] ... where do I find that pdf? [22:37] online [22:37] aws has bad seo, http://www.google.com/search?client=opera&rls=en&q=user_specified_kernels.pdf&sourceid=opera&ie=utf-8&oe=utf-8&channel=suggest [22:38] if you would like to see how we build our official images at rightscale, https://github.com/rightscale/rightimage [22:39] this seems to have become a lot more difficult than it used to be. Perhaps this is only useful if you want great elasticity? [22:39] flaccid: you've pretty much told him that the best service is always aws - that's all I meant [22:40] bhook: no i did not say that at all. but one would think that one person can't really compete with a multi billion dollar business that has been developed for nearly 5 years, but hey i'd love to see otherwise that would be cool [22:41] jo-erlend: what do you mean by 'how it used to be' ? with what sorry? [22:43] it used to be possible to put a VM, get an IP and some resources like 1GB RAM, two cores and x GB disk space. Now I have to read and read ... entire books, probably, just to get something up and running. A few years ago, I could just upload the image and run it. [22:43] jo-erlend: what was this with? [22:43] you're talking about the hypervisor? Both Xen and KVM. [22:44] no, what did you do the above procedure with? i.e. what software on the host, not the hypervisor === EgyParadox1 is now known as EgyParadox [22:45] but yeah, reading docs is expected in computing [22:45] I don't understand that question. I think I scp-ed the image onto the server and issued a command to start and stop it. They had web UIs for it too. It was really simple. [22:46] jo-erlend: right, so what is 'they'? still have not established what you used to do this with.. [22:47] different ones. I don't really remember their names. I just don't understand the benefit of the cloud if it makes things so much more complicated. [22:48] most users don't build images in the cloud, they use already available images. pretty easy stufff. [22:49] hmm. Ok, can I get a _plain_ Ubuntu Server so that I can rsync my local VM to the cloud one? [22:49] sure [22:50] you can just launch an instance and do whatever you like just like any computer [22:50] yes, but I have to use preconfigured images? [22:51] and if I choose Amazon, for instance... Does that mean I'm stuck with it, or can I move it to another provider? [22:51] well unless you learn how to build your own, yes. they are base images though, you can obviously do whatever you want on top via a configuration management system, cloud computing management platform or otherwise [22:52] you are not stuck with it. you would need to know what you are doing to 'move' it to another provider. this is why its a good idea to use base images plus configuration management on top so you are not tied to any one cloud [22:53] jo-erlend: you generally build custom images from within a running instance - it is quite difficult to build an image from scratch [22:53] what does all these complications do for me? [22:53] jo-erlend: you are looking like a troll, dude [22:54] ... [22:55] jo-erlend: in theory a well built image is portable between providers [22:55] I can get a VM up and running on KVM with Libvirt in .. 5 minutes? I don't have to read books. I don't have to understand now "cloud computing management platform configuration tools". I'm just wondering why the cloud is so much more difficult. [22:55] jo-erlend: but there are benefits from within a single provider, like quickly firing up instances in any availability zone [22:56] what is an availability zone? [22:56] jo-erlend: a local KVM instance makes lot more sense for a lot of uses - cloud stuff makes sense when you need the specific benefits of the cloud environment [22:57] jo-erlend: might be a good idea to start reading docs and using the cloud a bit so you get more familiar with how it works. [22:58] you could make your own cloud with libvirt and kvm, but that would be a reinvention of the wheel. UEC provides Eucalyptus atm and is moving to OpenStack iirc in the future. [22:58] both of which support KVM and can be managed via libvirt [22:58] cloud software is essentially an extension or another layer on top [22:59] because for on deman computing, you obviously need something to manage instances per user and bill them etc. if its a commercial service [22:59] *on demand [23:00] oh and the other thing is, if you dont need to use the cloud, you don't have to. [23:02] flaccid, you seem to be misinterpreting my question. My question is why a cloud VM is better than a normal VM. [23:02] jo-erlend: oh, well its not. i didn't see anybody say that it was. [23:03] as far as the instance/vm is concerened it has no idea and the image can be exactly the same. its not about whats better than. [23:03] oh. Seems to me that the entire wen has been trying to tell me that the cloud is the way to go. [23:04] whats wen? [23:04] web. Sorry. [23:04] sure, it has been the biggest buzz word for quite some time now [23:05] you are your own person, you can make your own choices [23:07] thanks for your support. Yes, I try to make my own decisions, but I sometimes prefer to find out what the choices are first. But that seems to be fairly difficult since each provider seems to use their own language of TLAs. [23:07] jo-erlend: cloud resources are generally cheaper and more geographically redundant than what people have at their finger tips [23:07] whats a TLA> [23:07] three-letter abbreviation. [23:07] jo-erlend: but, if you need one instance and you are using it for you (not to deliver a public facing server), then the cloud is overkill [23:08] acronym, even. [23:09] well rs supports 4 clouds at the moment. of course to do that we invested money. the images we build will always basically be xen or kvm and can be pretty stock. of course with commercial products there is going to be 'product's and different TLAs. [23:10] bhook, well. I was thinking about moving my blog. I want to manage my own services and I want some additional things such as a coudhdb and a mailserver. Pretty simple stuff. [23:10] ^ you can do that in the cloud or anywhere with one instance/vm [23:10] flaccid: Error: "you" is not a valid command. [23:10] flaccid, yes, but I don't even understand what they represent. It seems to me that these aren't normal VMs? [23:11] uvirtbot: tripped you out did i? :p [23:11] flaccid: Error: "tripped" is not a valid command. [23:11] jo-erlend: no such thing as a normal VM. well you are new to the cloud, so much to learn and read at this stage [23:14] yes, but it seems I can't just ask for xGB RAM, x CPU cores, x GB hdd? _Why_ is there so much to learn? [23:14] well you can do that easily in eucalyptus and ec2 [23:15] why is there so much to learn? because there is? [23:15] http://aws.amazon.com/ec2/instance-types/ [23:17] yes. That seems nice. But then there is storage, which as I understand it is a completely different service? [23:17] jo-erlend: yes, read up on s3 and ebs [23:19] ok. So what I need is an EC2, an S3, an EBS and I connect these things using a cloud computing management platform configuration tool? [23:19] sure [23:19] jo-erlend: a simple vps (virtual private server) from something like vpslink.com may be more like what you have used in the past. far less flexible than aws, but much simpler [23:19] though you can do all of that already in the aws console or in UEC [23:20] i assume that you have not setup UEC yet, so you may like to get started with AWS/EC@ [23:20] ec2 [23:21] ok, if I want to be able to move my VM back onto my own servers someday, then Amazon is the way to go? Because I really don't want to make a life long commitment. [23:22] thats not the right approach, 'moving vms'. you can always get data off an instance. you can directly download data from s3, you can't with ebs, but you can just ssh or rsync it out or whatever [23:23] jo-erlend: linux is generally fairly forgiving if you rsync the contents of /. There are a few special files you have to remove and/or tweak, but all of your data and such should be fine if you chose to move later. [23:23] store you data in s3 or ebs or vcs etc., not on the instance itself [23:24] oh, ok. Because it sounded like these clouds needed special kinds of setups. [23:24] no, no idea where you got that from. ubuntu is ubuntu etc. [23:24] jo-erlend: I had a local cloud controller crap out on me - I rsync the contents of the volumes onto flash drives and booted them on a bunch of spare laptops - like I said, linux is very forgiving [23:25] sure, I know. I'm not new to Linux. I'm new to these clouds. [23:27] the simplest scenario is backup/restore with s3 in a tar ball for example. pretty simple stuff at the end of the day. the idea is you design things as appliances so if an instance dies/fails, you can just launch another one and restore the same state. [23:27] if the images needed to be generated in their systems and not in my own, then I just assumed that was because there were special considerations when running it in the cloud. [23:28] flaccid, yes, pretty much what I do at home. [23:28] jo-erlend: they don't need to be generated in their cloud. you can build images anywhere, but it is pretty handy to just spin up an instance for 1 hour to do a build [23:29] the main consideration with ec2 is that it runs an old version of xen so that can be a problem at times. [23:30] oh, ok. So I can get a finished image that really is plain ubuntu server without any modifications? [23:30] jo-erlend: yes thats what ubuntu provides [23:31] http://cloud.ubuntu.com/ami/ [23:31] in terms of the UEC images, they are available somewhere, should be mentioned in the docs, but you don't have UEC setup yet it seems anyway [23:33] https://help.ubuntu.com/community/UEC/Images#UEC Images on Local Hypervisor [23:34] https://help.ubuntu.com/community/UEC/Images#UEC%20Images%20on%20Local%20Hypervisor [23:35] jo-erlend: and as you can see it shows you how to run that without the cloud env [23:35] oh, ok. What is the difference between an ebs-image and an instance-store-image? [23:36] the root device is ebs or s3 basically [23:36] oh, there is ebs-hvm-cluster too. Does this depend on how I store my data? [23:37] funnily the doc bhook referred answered all your questions. which is kind of why my first response was to read the docs. [23:37] no, this is where the root device comes from [23:37] instance-store, the image is stored in s3 [23:38] jo-erlend: the one thing to be extra careful with when playing with the cloud is to remember that ephemeral storage is NOT persistent. If you shut down the instance without storing the data somewhere, it goes bye-bye. [23:38] yep, this is the key to cloud computing. design your stuff so if it dies, launching a new instance restores state on boot. [23:39] oh, ok. My VM isn't automatically HA? [23:39] HA has nothing to do with anything here [23:39] HA doesn't have anything to do with a computer dying? [23:39] not in this context [23:40] you could have a HA setup, redundant via horizontal scaling and all instances die at the same time. [23:40] but I have to design my VM so that if a harddisk dies, then there is a copy somewhere? [23:40] its the same situation, HA is irrelevent [23:40] jo-erlend: yes [23:40] by die he means terminates, whether intentionally, accidentally, or due to hardware failure.... even a HA server can be shut down manually [23:40] this is what storage and backups is for. [23:41] if the instance "dies" the ephemeral storage is gone [23:41] yes, but I thought those systems were supposed to do this automatically. [23:41] and so is the instance itself which yeah is usually regarded as ephemeral [23:41] jo-erlend: probably best to disregard all your misconceptions of the cloud [23:42] systems are available, but of course you have to use them [23:42] that seems to be a good idea. I've been under the impression that it was a replacement for VPSes, but they seem to hardly have anything to do with one another. [23:46] VPS vs AWS is about like Linux vs OSX... the ability to run ifconfig on both doesn't make them anything alike [23:48] how much time do you think I'll have to spend getting ubuntu server up and running on aws? [23:48] jo-erlend: takes a few clicks to launch the AMI and a few minutes to boot, thats it. [23:48] jo-erlend: wrong question - the question is how much time will it take you to learn the tools and process [23:49] jo-erlend: what have you done so far and what docs have you read? [23:49] right, but that's when you've already learned how stuff works. I'm not used to all this. I've just used two physical boxes with KVM and Libvirt for virtualization and drbd for storage availability. This seems much more complicated. [23:50] i wouldn't call logging into a website and clicking a few things complicated. [23:51] well, I have to make sure that my data is persistent and I have to learn how to use these mechanisms you talked about to make sure that if a harddisk dies, my data is still instact, etc. I can't just click buttons. I need to know which ones to click. [23:52] like bhook and i both say, you need to read docs and learn [23:53] sure. That's why I'm asking how much time I should expect to spend on that. If we're talking weeks then it'll be far cheaper for me to pay someone to do that for me. [23:53] jo-erlend: if you put your mind to it you can have the basics down in a couple of days if you have previous experience with non-cloud virtualization [23:53] its not a question we can answer. it depends on your intelligience. [23:53] about 140. [23:54] 140 what? [23:54] jo-erlend: well, at 140 IQ you are probably screwed [23:54] jo-erlend: hire someone [23:54] hehe [23:54] depends on which side of the pond you're at. :) [23:55] but ok. The general idea is that it's similar to a blade server? That you have a computer with some memory and then you have a storage area some other place where you can get persistent storage? [23:56] i guess so [23:56] and if you want to make sure your data is safe, you need to get two storages and synchronize them somehow? [23:57] nope [23:57] oh? Perhaps I misunderstood your reply earlier then. [23:57] you might like to read the docs and try it out. [23:58] you're right. Perhaps I should just take a break and start over. [23:58] jo-erlend: ephemeral storage isn't "real" storage - it's more like a RAM disk [23:58] well doesn't seem like you started at all