=== mathiaz_ is now known as mathiaz [17:40] I am having trouble registering node controllers [17:41] I cant get them to list on the cloud controller after being discovered by the cluster controller === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates === SuperHark is now known as MichealH [19:14] smoser: you know what would be an awesome use of the landscape cloud stuff / cloud10 ? If I could, in a bug report, grant the user the rights to run a node of any AMI for 1 hour. [19:15] Like "can you repeat that with a fresh install of lucid? Here's 1 lucid node .. go" [19:16] SpamapS, i suggested this in the discussion about awstrial and was suprised that no one else found it a good idea. [19:16] especially for "can you reproduce this on the development release" [19:17] smoser: even a button on each bug that says "I want to try and repeat this bug" that takes you to an appropriate VM. [19:17] as, i think that question is almost absurd for an end user... "Can you re-install your operating system and try this thing that maybe doesn't even boot for you" [19:17] :) [19:18] If we can't do it on cloud10, even just a hook into vmbuilder would be pretty cool.. ;) vmbuilder --bug 559033 [19:18] Launchpad bug 559033 in evolution "Evolution crashes on changing task priority" [Medium,Invalid] https://launchpad.net/bugs/559033 [19:18] SpamapS, yeah, i really think something like that woudlbe valuable. A similar thing is doable with testdrive, but with ec2 you dont have to download anything [19:18] how vmbuilder ? [19:18] well you'd end up with a vm that starts up with the offending package installed. ;) [19:19] (I just discovered vmbuilder 2 days ago btw.. I may not understand its full power ;) [19:23] SpamapS, well as far as ease of use goes, i would recommend testdrive over vm-builder [19:23] test drive - > launch instance -> apt-get intsall ... [19:24] never used testdrive, but maybe its what I should be using instead of fighting with vmbuilder [19:24] vm-builder -> some long command line set -> lots of image building , downloading -> boot instance. [19:24] testdrive just boots the isos, but does so very easily [19:25] if you want to "testdrive" a server install, i suggest the UEC images and booting them in kvm [19:25] testdrive (maybe in natty now) has support for uec images as things to 'test drive' === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates [20:36] does the uec node controller have to have vt extensions on the CPU? [20:48] anyone know why when I update max_cores, I dont see an increase in vm's I can launch on the cloud controller? [20:52] so when I shutdown nodes/controllers, what happens to VM's? [20:53] their process stops execution [20:53] jmgalloway: VT is needed [20:53] yeah, I was just hoping to use some older computers for nodes... [20:54] I still dont understand about the max_cores not updating [20:54] can't help with that [20:54] reboot/waiting solves alot [20:55] I tried rebooting the entire cloud, restarting nc, cc, and eucalyptus....doesnt update [20:55] I'd look it up for you but I broke my right index finger and typing with one hand is a bit time consuming [20:56] I had it working before [20:56] did you try clean start with cc? [20:56] seems like you just update the variable then restart the services and it works [20:56] hmm === dendrobates is now known as dendro-afk [20:56] let me try that [20:58] tried, didnt work [20:58] #MAX_MEM=2048 [20:58] #MAX_CORES="8" [20:58] #MAX_DISK="100" [20:58] you need to set max_cores on all NC's [20:58] that's the eucalyptus.conf of the node controllers [20:59] and remove the # [20:59] lol! wow [20:59] thanks [20:59] :) [20:59] np [20:59] I was searching everywhere! [20:59] lol [20:59] # is for comment [20:59] they've been commented out [20:59] yeah...lol [21:00] y [21:00] what is the max_cores supposed to do? [21:00] rather new to uec [21:01] if you change that variable on the node controllers...it will increase the number of vm's you can execute on the node controllers cpu [21:01] kvm allows for 8*#of physical cores [21:01] so dual core can run 16 VM's? [21:02] thanks, this seems to be something useful [21:02] yes but be careful..make sure you leave cpu power for the underlying os [21:03] would be cooler if you could set that on cc instead of needing to set it on all nc's [21:03] I have dual core cpus..I run 4 vm's per physical core [21:03] I know [21:04] well you could just dd a hard drive and put it in a lot of node controllers...everything setup then [21:05] what kind of setup do you have for your cloud? [21:06] i have a build for adding node controllers for around $400 each.. quad core intel, 4GB ram, 320GB hds.. able to run 16 vm's [21:08] doing a school project [21:08] got around 60 comps for short period of time [21:08] first settings up everythin with around 10 comps [21:08] 60 cool [21:08] 2 cluster controllers [21:09] one clc (that has walrus) [21:09] CC's have storage controllers also [21:09] might be adding more CC's [21:09] I have one clc, one walrus, one cc/sc, and two node controllers [21:10] I am only able to run 16 vm's right now [21:10] kinda just have had only trouble with the cloud [21:10] the cloud? [21:10] cloud controller? [21:10] mostly getting VM's running [21:11] but it's all fine now [21:11] have you tried hybridfox? [21:11] yes we use it [21:11] It seems sketchy...sometimes it works, sometimes it doesnt [21:11] had trouble getting bundled images to run, and storage images are crap [21:11] hybridfox works nice after you restart it couple times [21:12] like if you get an error, close the tab and open new obe [21:12] one [21:12] the comps we got are C2D E6300 and 4Gigs RAM [21:12] yeah, I have problems with the keys and permissions using hybridfox. It will launch vm's but wont let me connect to them even though I use the key it made for me [21:13] 60 of those? [21:13] yes [21:13] around 60 [21:13] +/-5 [21:13] I have 6 total machines [21:14] I am writing a front end for my cloud...since everything I'm building is hosted on a private network(s) I have to give users that are not attached to my private network the ability to launch vm's [21:16] php is running on my cloud controller..communicates in the browser with the WAN users and spawns vm's with a gui interface. it has to keep up with ports and update the iptables since I only have one WAN address [21:18] sounds nice [21:19] with that amount of machines, why don't you integrate CLC/CC/Walrus? [21:19] then you could have second storage controller and 2 more nodes [21:19] I want it to be fully scalable [21:20] which it would be..but making them work with different physical machines is more interesting than combining them together. [21:20] does the sc and cc still have to be on the same box? [21:20] trnope [21:20] nope* [21:20] atleast I think [21:20] :D [21:20] I thought it did in euca 1.6...I'm using ubuntu 10.10 and euca 2.0 [21:21] well it can be [21:21] yeah, mine are combined [21:21] as I say, rather new to this [21:21] sc/cc [21:21] open eucalyptus doesnt natively support windows images either [21:21] but it can be done [21:22] all of my vm's are linux, single core, 256mb ram [21:22] you planning to run the cloud for a long time? [21:22] until I graduate [21:22] 2 years possible [21:24] I think you should be able to put storage controller and cc to 2 diff machines [21:24] but just reading in between lines [21:24] of ubuntu instructions [21:24] CC needs a storage controller [21:24] but it can be on other pc afaik [21:25] the one thing I'm not sure about is how to make persistent storage that is mounted automatically for users when they launch vm's [21:25] create storage volumes and script to launch vm? [21:25] one that launches vm and mounts the volume [21:26] or perhaps launches vm, creates volume and mounts it [21:26] I need to look into that [21:26] also, I need to look into quotas [21:26] so a single user cant spawn enough vm's to use all resources [21:27] ye [21:27] I'm in rather tight schedule so I try avoid completely custom solutions [21:27] In my setup, users will not be able to spawn vm's by command line [21:27] yeah [21:27] I have 2 years haha [21:27] or 1.5 years to go [21:28] I think we have like 4 weeks left [21:28] oh cool. is your cloud interface public? [21:28] nope [21:28] it's a school project [21:28] mine is...although I just threw together a front end web site [21:28] we need to figure out if it's useful to teach cloud to students [21:29] got like less than 2 months in total to get everything setup, demo'ed, written documents etc [21:29] sure it is, its compatible with aws [21:29] really tight schedule [21:29] are you working in a group? [21:29] group of 4 [21:29] ah cool. I'm doing this all alone [21:30] not to sound cocky, but I seem to be the only one knowing wtf were doing [21:30] there's a lot to wrap your head around to administrate a cloud setup [21:30] meaning I do almost everything [21:30] and teach others lol [21:31] naw the other guy gets around too [21:31] but the other 2 [21:31] gah [21:31] my advisor is teaching a cloud class in the spring, and she's using this cloud in the class [21:31] I have to lock it down [21:31] I would need to walk them thru by hand [21:31] hehe [21:31] remember to make lot's of security groups [21:32] well i really just need two security groups..one for me, and one for everyone else [21:32] can you change the security groups between machines after launching them? [21:32] atleast hybrid fox didn't seem able [21:33] I dont know. I am going for a custom interface that gives the user a gui..so they click to launch vm's, not caring about how it launches or security groups..they will just have access to the vm [21:33] heh ok [21:33] I'm off to bed, night [21:33] php will generate the script to launch the vm, vm size, security groups, etc [21:33] later [21:33] we got a blog @ silvercloud.wordpress.com if you want to see how our project is doing [21:34] sorrt [21:34] sorry [21:34] silvercloudproject.wordpress.com [21:35] (some blatant advertising :) === dendro-afk is now known as dendrobates [21:44] ok cool === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk