[00:38] smoser: http://dpaste.org/Se9H/ only a couple of things - Couldn't get a file descriptor referring to the console; error: unexpectedly disconnected from boot status daemon; FATAL: Could not load /lib/modules/2.6.32-305-ec2/modules.dep: No such file or directory (need to work out best way to install these on squeeze/debian); [00:38] and Waiting for /dev to be fully populated...udevd-work[258]: kernel-provided name 'evtchn' and NAME= 'xen/evtchn' disagree, please use SYMLINK+= or change the kernel to provide the proper name [00:38] and Setting kernel variables ...error: "net.ipv6.bindv6only" is an unknown key [00:39] most of the above is kernel related. smoser i just need to address/fix the above and all should be good. if you could also advise what the latest actual kernel is that would be great [00:39] "actual kernel" [00:40] all the ubuntu published items conform to https://wiki.ubuntu.com/UEC/Images/NamingConvention [00:40] yes [00:40] (or close to it) [00:40] for released kernels, they'll have ubuntu-images/<...> in their name (per describe-images) [00:40] er.. for kernels it will be [00:40] yeah so what is the latest kernel id ? [00:41] well, for us-east-1, and lucid (released) that would be: [00:41] 099720109477/ubuntu-kernels/ubuntu-lucid-amd64-linux-image-2.6.32-305-ec2-v-2.6.32-305.9-kernel [00:41] aki-0b4aa462 [00:42] is there someone that lists these ? [00:42] i have considered putting kernels in the data at http://uec-images.ubuntu.com/query/ [00:42] but largely, we don't consider kernels aki to be a product as much as a side effect of the image [00:43] that said, http://uec-images.ubuntu.com/query/lucid/server/released.current.txt [00:43] will have a list of the latest ami [00:43] and their assicoated aki [00:43] thanks that helps a lot [00:44] there is 'daily.current.txt' there also [00:44] now back to the errors if you don't mind helping? [00:44] and maverick ... [00:44] so lots to chose from [00:44] :) [00:45] i'vve seen the "Couldn't get a file descriptor" error before, but dont really recall what that was about. [00:45] fwiw, jjohansen is the kernel team member for this ec2 [00:49] i'm looking at some logs [00:49] we see the "Could'nt get a file descriptor" on karmic currently [00:49] cool thanks [00:49] and on lucid prior to alpha2 [00:49] err.. prior to alpha3 [00:50] and would also like your advise on installing/enabling the kernel mods from the ubuntu package in debian and if this is an ok practice etc. [00:50] where did you get the ramdisk ? [00:50] or am i mistaken: [00:50] Begin: Loading essential drivers... ... [00:50] Done. [00:50] Begin: Running /scripts/init-premount ... [00:50] its ubuntu lucid i386 linux image 2 6 32 301 ec2 v 2 6 32 301 4 ramdisk 20100114 1 [00:50] ari-3f2d7c7a [00:50] yeah, so we dont use ramdisk anymore [00:51] right so set to default [00:51] set to default ? [00:51] i'm assuming default=none now [00:52] i'll relaunch with that [00:52] yeah, it probably wont hurt or affect it. [00:52] that has been my experience. the ramdisk doesn't hurt anything. just doesn't help. [00:53] regarding installing ubuntu packages, are you asking from licensing perspective ? [00:53] those modules come from the packages... our ec2 kernels are normal kernle builds and have normal packages. [00:54] https://launchpad.net/ubuntu/+source/linux-ec2 [00:54] i dont know how well those would install onto a debian [00:54] more from an operational point of view. this is to address FATAL: Could not load /lib/modules/2.6.32-305-ec2/modules.dep: No such file or directory [00:54] so yes the question is regarding installing them on debian [00:55] see because the amazon kernels are not working with debian, i am having to use the ubuntu kernels [00:56] yeah. so i dont know whether or not the packages there will "just work" (ie install) on debian or not. [00:56] but almost certainly a dpkg -x linux-image-*.deb would get you the /lib/modules/ tha tyou need [00:56] well i would either have to use the lucid repos briefly and install via apt or manually install the package and its deps via dpkg [00:57] well the linux-image-* pkgs have deps.. [00:59] ok well no ramdisk fixed that [00:59] zcat: invalid magic - another small one to address [00:59] i mean no ramdisk fixed the console descriptor errors [01:10] well it still defaulted to ubuntu lucid i386 linux image 2 6 32 301 ec2 v 2 6 32 301 4 ramdisk 20100108 smoser0 [01:10] that must be tied to the kernel [01:13] wb [01:43] smoser: if i do a reboot and i have installed these modules, this should remove the FATAL: Could not load /lib/modules/2.6.32-305-ec2/modules.dep: No such file or directory message right? [01:44] well maybe [01:44] you may be getting that from the ramdisk [01:44] at least i'm fairly certain that ramdisks do 'modprobe' which will result in a message like that. [01:45] but if the file is there, it should not return No such file or directory right? [01:45] are you using a ramdisk ? [01:45] this kernel appears to default to ubuntu lucid i386 linux image 2 6 32 301 ec2 v 2 6 32 301 4 ramdisk 20100108 smoser0 [01:46] if you are using a ramdisk,and it was built with a differnet kernel then it won't have that /lib/modules/ and it will try to modprobe [01:46] aki-296f3e6c [01:46] kernels dont default to a ramdisk [01:46] this is something that sucks. [01:47] oh but an image defaults to a ramdisk? [01:47] if you registered an image with a ramdisk and kernel, then you are not able (that I know of) to boot without a ramdisk or kernel [01:47] and, actually, it goes back step, since you can give ramdisk and kernel to the bundle command [01:47] ok so ramdisk is bogus, no need to build image with a ramdisk value? [01:47] if you've done that, then there is no way to specify "None' [01:47] right [01:48] then, you can (if you wanted) specify --ramdisk [01:48] but there is no --ramdisk [01:48] sweet then i'll omit ramdisk [01:48] right. thats a superset of the functionality, but then, if you ever wanted a ramdisk, the user has to specify one, which just isn't very friendly [01:49] if ec2-bundle-vol rcvs --kernel --ramdisk --generate-fstab will it simply ignore the no param values [01:49] just do not pass --ramdisk [01:49] or --kernel [01:49] on bundle or on register [01:50] fwiw, there are scripts 'uec-publish-image' that you could use that, in my opinion are more friendly than ec2-bundle, ec2-upload, ec2-register [01:50] well i still want a default kernel [01:51] http://manpages.ubuntu.com/manpages/lucid/man1/uec-publish-image.1.html [01:51] right. [01:51] i *think* (you'd have to test this) [01:51] that you can bundle without anything [01:51] and then add --kernel when you register the image [01:51] which, in this case, would allow you to avoid re-bundling [01:52] thats cool i'll test that [01:54] hmm it doesn't specify params http://docs.amazonwebservices.com/AmazonEC2/dg/2007-01-03/CLTRG-register.html [02:23] see the date in that url [02:24] http://docs.amazonwebservices.com/AWSEC2/latest/CommandLineReference/index.html?ApiReference-cmd-RegisterImage.html [02:24] is what you want [02:26] oh yeah thanks, eek on google! [02:27] thanks muchly :) [02:27] standby, i'll have some results soon [02:27] 1 is building now [02:30] the good thing i've noticed so far is that these AMIs i'm creating work fine with the lucid kernel and boot up really fast [02:35] hmm that doc doesn't mention -K and -C options yet i am using them currently.. [03:01] Required option '-K, --private-key KEY' missing (-h for usage) [03:01] heh its not in their doc [03:51] ah must be in http://docs.amazonwebservices.com/AWSEC2/latest/CommandLineReference/index.html?ApiReference-cmd-RegisterImage.html [03:51] err http://docs.amazonwebservices.com/AWSEC2/latest/CommandLineReference/index.html?CLTRG-common-args-api.html [04:11] Unable to read instance meta-data for ramdisk-id [04:11] Unable to read instance meta-data for product-codes [04:12] ^^ from ec2-bundle-vol. what should i do about this ? [04:12] flaccid: Error: "^" is not a valid command. [04:41] smoser: built and registered with no ramdisk, a launch with default assigns ubuntu lucid i386 linux image 2 6 32 301 ec2 v 2 6 32 301 4 ramdisk 20100108 smoser0 [04:41] the default kernel being assigned is ubuntu-kernels-testing/ubuntu-lucid-i386-linux-image-2.6.32-306-ec2-v-2.6.32-306.11-kernel [04:42] those errors don't surprise me. [04:42] i guess we'll see in console output if it is wanting 306 or 301 modules that will determine if the ramdisk is asking for it or not [04:42] yes but i want to get rid of those errors :) [04:42] no, the "unable to read instance metadat" errors from bundle-vol [04:43] they dont surprise me [04:43] yeah i want to fix those.. [04:43] bundle-vol is a yucky hack. you shouldn't rely on it. [04:43] well thats what aws publishes.. [04:43] bundle-vol basically has to "un-boot" a system. [04:44] which is just simply going to require a bunch of distribution/OS specific HACKs [04:45] its much cleaner to use debootstrap and bundle-img [04:45] well FATAL: Could not load /lib/modules/2.6.32-306-ec2/modules.dep: No such file or directory. so its the kernel that wants it. i'll extra the mods in this build [04:45] where is bundle-img ? [04:45] i am using debootstrap [04:46] ec2-ami-tools [04:47] oh you mean ec2-bundle-img [04:47] yeah [04:47] can you give me output of describe-istances for the image you booted ? [04:47] i think you booted with a ramdisk. [04:47] or, you dont 'have those modules in the root filesystem [04:47] ec2-describe-instances i-abcdefg [04:49] for some reason that returns Client.InvalidInstanceID.NotFound: The instance ID 'i-ba897dfe' does not exist [04:49] yet its running [04:50] i didn't include the modules in that build however [04:50] rightscale dashboard is indicating use of ubuntu lucid i386 linux image 2 6 32 301 ec2 v 2 6 32 301 4 ramdisk 20100108 smoser0 [04:51] it maybe that rightscale is buggy [04:51] if you did not bundle with '--ramdisk' [04:51] and did not register with '--ramdisk' [04:52] then you should not have a kernel [04:52] err... you should not have a ramdisk [04:52] it is possible that rightscale is being nice, and trying to use that one, in which case, that really sucks. [04:52] but describe-instances certinaly should be able to give you output [04:52] what about describe-images of the ami [04:54] that returns does not exist as well. no idea why [04:54] hmm are these region specific? [04:55] i think its only returning us-east when i need us-west [05:00] yes. all of it is region specific [05:00] each region is its own "cloud". the only data they share is your account id [05:00] nps [05:02] this is the image [05:02] IMAGE ami-b5b8e9f0 041819229125/RightImage-Debian-squeeze-20100618033019-i386-v5.4.6 041819229125 available private i386 machine aki-3197c674 instance-store [05:03] this is the instance [05:03] INSTANCE i-ba897dfe ami-e7b6e7a2 ec2-204-236-179-7.us-west-1.compute.amazonaws.com ip-10-162-143-68.us-west-1.compute.internal running CJF: AWS US-West (testing) 0 m1.small 2010-06-18T03:39:43+0000 us-west-1b aki-c18edf84 ari-8f2978ca monitoring-disabled 204.236.179.7 10.162.143.68 instance-store [05:04] so i guess either rightscale is allocating ari-8f2978ca or ec2 (likely rightscale) [05:06] can you help confirm that there is no such thing as a 'default' ramdisk when registering or launching? [05:07] we seem to be using 'default' with a reserved ari then.. [05:07] yep [05:08] its rightscale is being "nice" [05:08] well, when you register or bundle and pass '--ramdisk' that gives you a "default ramdisk" [05:08] and if you do that, you can't ever run an instance without a ramdisk. [05:09] but, the describe-images info above shows that the image has no ramdisk associated with it. [05:09] which means either [05:09] a.) the user must provide one in run-instances [05:09] b.) there will not be one used [05:09] that would of came up in the describe-images but right? i took this out of the build process [05:09] yep we are on same level [05:09] right. it would be shown in describe-images output [05:09] i will raise a bug for the rightscale dashboard issue [05:09] look at any other one, you'll see an ari [05:10] fwiw, that 'smoser0' is a debug ramdisk :) [05:10] (with some interesting characteristics, it can process user data) [05:10] hehe, looks like we chose your ramdisk image for the default... :) [05:15] our choice to not have ramdisks has thrown errors much furrther than i thought it would have [05:16] there is issues with no ramdisk? [05:19] is there any way to remove a debootstrap target dir? i.e. you can't remove /proc and /dev [05:36] you have to unmount proc and dev first [05:36] then remove [05:36] there are no issues with no ramdisk other than tools that expected a ramdisk [05:37] UEC had fits (Eucalyptus) [05:37] righscale apparently had fits [05:37] the expectation of an ari was just hard coded many places [05:37] oh yeah, other places. [05:37] landscape [05:37] and the canonical "image store" (part of UEC) [05:38] i see [05:39] ok now launching with a new image [06:14] smoser: extracting from the .deb for those modules doesn't give the desired files [06:14] no? [06:15] you're right.sorry. [06:15] i'll cp them across from the worker which is lucid [06:16] just the wrong package [06:16] the linux-ec2 is a metapackage [06:16] that depends on the right linux-image-2.6.32-305-ec2 [06:17] i'm extracting linux-image-2.6.32-305-ec2_2.6.32-305.9_i386.deb [06:17] i tell ya what but it boots so fast [06:17] http://packages.ubuntu.com/hu/lucid/linux-image-2.6.32-305-ec2 [06:17] yeah, thats what you want [06:22] hopefully cp -R /lib/modules/2.6.32-305-ec2 "$DEBOOTSTRAP_TARGET_DIR"/lib/modules/ is good enough.. [06:28] i'd think so [06:29] i'm getting 1.2min startups [06:30] 14.5mins to create and build an AMI [07:00] smoser: only a few things left to address now, http://dpaste.org/5fAp/ [07:01] start up is all over the map [07:01] what do you think is left ? [07:01] it looks pretty good [07:03] what do you mjean by "1.2min startups" ? [07:03] total time to login/ssh from what [07:04] flaccid, [07:04] i think thats the boot time, not sure how its configured [07:05] need to address the ' WARNING: Currently emulating unsupported memory accesses' [07:05] there is no /lib/tls on that instance [07:05] also Waiting for /dev to be fully populated...udevd-work[86]: kernel-provided name 'evtchn' and NAME= 'xen/evtchn' disagree, please use SYMLINK+= or change the kernel to provide the proper name [07:06] and the few under cannot unset powersave mode [07:06] and Setting kernel variables ...error: "net.ipv6.bindv6only" is an unknown key [07:06] and maybe /build/buildd/linux-ec2-2.6.32/drivers/rtc/hctosys.c: unable to open rtc device (rtc0) [07:06] also Setting kernel variables ...error: "net.ipv6.bindv6only" is an unknown key [07:13] i'll build a lenny now [07:37] lenny works but not with rightlink [07:39] i'm going to bed. good luck. [07:39] thanks; cya later