[00:42] <smoser> mathiaz, i dont think so
[00:42] <smoser> anything in console output ?
[00:44] <mathiaz> smoser: well - I'm trying to log with the ubuntu account
[00:44] <mathiaz> smoser: once I've configured the system with puppet
[00:44] <smoser> and aparently you cannot. :)
[00:45] <mathiaz> smoser: the ubuntu account is then locked
[00:45] <mathiaz> smoser: not sure what does this though
[00:45] <smoser> hmm..
[00:45] <smoser> i really dont know.
[00:45] <mathiaz> smoser: me neither
[00:45] <mathiaz> smoser: I've worked around it for now
[00:46] <mathiaz> smoser: might investigate it later
[00:46] <mathiaz> smoser: keep up the good work! the ec2 images are great base to build upon!
[00:47] <smoser> well the account *is* locked initially
[00:47] <smoser> as in no password set
[00:47] <smoser> (usermod -L)
[00:47] <smoser> but obviously you can get in
[00:48] <mathiaz> smoser: yeah - using my public key
[00:48] <mathiaz> smoser: however there is a password set for ubuntu
[00:48] <mathiaz> smoser: in /etc/shadow
[00:48] <mathiaz> smoser: the string starts with !
[00:48] <smoser> there should not be
[00:48] <smoser> oh
[00:48] <smoser> yeah
[00:48] <mathiaz> smoser: but there is a long string there
[00:53] <elro> Is there a way to specify that my rc.local should be run after mounting all filesystems? Putting a mount in rc.local shows that my ebs volume /dev/sdf has not yet been mounted.
[00:54] <elro> I guess it is because of the nobootwait
[00:59] <smoser> mathiaz, just for kicks i tried to see if just doing a 'usermod -U' would end up locking you out, but i can still get in
[01:00] <mathiaz> smoser: I think it may be related to the sshd_config file I put in place
[01:00] <mathiaz> smoser: because I can log in before I install the instance via puppet
[01:01] <mathiaz> smoser: using the ubuntu account
[01:01] <mathiaz> smoser: once puppet has run, the ubuntu account cannot be used anymore
[01:02] <smoser> mathiaz, i'm sure you're as capable as me to debug.
[01:02] <smoser> but i'd start by copying off /etc/ before and comparing to after (assuming you can still get in)
[01:12] <smoser> RoAkSoAx, i have another ec2 scripts task for you at some point :)
[01:13] <RoAkSoAx> smoser: ok cool :)
[01:13] <smoser> right now the publish is all serial. first publishes us-east-1 , then us-west-1 , then eu-west-1, then ap-southest-1 ...
[01:13] <smoser> it'd be nice if we just did those in parrallel
[01:13] <smoser> a full publish takes like 6+ hours at the moment
[01:13] <smoser> probably could cut that in half
[01:14] <smoser> and there are probably other things we could do to speed it up.
[01:14] <smoser> well, i've got to run. later.
[01:14] <RoAkSoAx> alrighty
[01:14] <RoAkSoAx> have a good one
[08:27] <superxgl> hi all, is there anyone who have ever written a webservice like this ?   http://open.eucalyptus.com/participate/wiki/autoscaling-behalf-monitoring-virtual-machines
[08:32] <superxgl> hmm... i am a newbie , i haven't written a webservice before, can anyone here point me somewhere how to write a webservice ?
[08:40] <flaccid> google.com
[08:44] <superxgl> k, i am seaching now..
[08:54] <superxgl> hmm...using axis or xfire..
[09:41] <kim0> superxgl: I suppose if you're not a developer .. hire someone who can implement this cleanly. I also don't think it has to be implemented that way
[09:42] <kim0> superxgl: potentially just a shell script that ssh'es .. checks load .. and decides to launch more instances might do huh
[09:42] <flaccid> superxgl: what are you trying to achieve.? i'll let you know what to do
[10:09] <superxgl> kim0,flaccid: thanks very much :)  i want to achieve the load of instances , then i can decide if it is need to start a new instance (i.e a hadoop slave node)
[10:10] <flaccid> load of instances?
[10:10] <superxgl> yes
[10:10] <flaccid> what is that?
[10:11] <superxgl> the system metrics
[10:11] <superxgl> like the cpu load
[10:12] <TREllis> system load I'd imagine
[10:13] <superxgl> yes, i want to get the load average of the cluster
[10:17] <kim0> superxgl: for s in servers ; do ssh $s uptime ; done
[10:17] <kim0> there you have it :)
[10:17] <kim0> decide to launch more instances ..
[10:20] <superxgl> kim0: hmm. yeah, seems like just a shell script is ok
[10:22] <kim0> yeah .. I'm saying it can be .. ofc if it gets more complex .. perhaps some proper code needs to be written
[10:28] <superxgl> hmm, maybe i also need cron
[10:34] <superxgl> i found a good example :   http://bash.cyberciti.biz/monitoring/get-system-information-in-html-format/
[10:38] <flaccid> superxgl:  i use RightScale which uses collectd and RRD
[10:40] <superxgl> flaccid: what is collectd and RRD ?
[10:40] <flaccid> google.com
[10:41] <superxgl> finding now :)
[11:04] <superxgl> http://support.rightscale.com/12-Guides/RightScale_Methodologies/Monitoring_System/Writing_custom_collectd_plugins/Custom_Collectd_Plug-ins_for_Linux
[11:05] <superxgl> collectd is very cool ..
[11:06] <flaccid> indeed it is
[11:08] <superxgl> yeah, i want to have a try :)
[11:10] <flaccid> go for it. if you want to use rightscale, its free
[11:10] <flaccid> you get free monitoring for the first week of each instance/server
[11:18] <superxgl> great . i want to apply collectd and rrd into my private cloud
[11:18] <flaccid> can do that too. its just a package..
[11:24] <superxgl> but i am still a little confused...
[11:43] <superxgl> that is : how graphs  be drawn  ?
[11:51] <flaccid> checkout rrdtool
[11:54] <superxgl> ok,tkx
[13:29] <superxgl> i have a question, how about ganglia ? which one is better ?
[13:33] <superxgl> why RightScale not use ganglia instead of rrdtool?
[14:19] <RobertLaptop> Does eucalyptus support more then 1 Public interface Nic?  In other words I have 2 diff ISP with there own IP range.  So I would want Public IP range A = br0, Public IP range B = br1, and Private IP range C = br2
[16:01] <SpamapS> smoser: hey.. so I've been thinking more about cloud-init and the cloud type setting.. I am curious.. could we check the install media type to auto detect it?
[16:02] <smoser> i'm not sure i follow the question
[16:02] <smoser> but its a runtime decision
[16:02] <smoser> it has to be
[16:02] <smoser> the idea is that we're making one image that runs on multiple clouds (or, non-clouds)
[16:06] <SpamapS> smoser: right, but if the install media is CD, the cloud type should be local file
[16:06] <SpamapS> smoser: as happy as I am that the boot only slows by 30 seconds now instead of 20 minutes .. I'd really prefer it to not slow things at all :)
[16:08] <smoser> it should not stop you for 30 seconds
[16:09] <SpamapS> it definitely does
[16:09] <smoser> the reason looking  at install media is not valid is that you should be able to do an install in kvm (or bare metal for that matter) and then publish that on EC2 (ideally)
[16:09] <smoser> hm.. do you know why you'r ebeing slowed down for 30 seconds ?
[16:10] <smoser> you really shouldt be. i suspect the ovf datasource is what is giving you fits
[16:10] <SpamapS> well 1 thing is I'm gettng a DataSourceNotFoundException ...
[16:10] <superxgl> oh..ganglia is based on rrdtool...
[16:10] <smoser> and there may be a fix for htat in trunk already
[16:11] <SpamapS> cloud-init start-local running: Thu, 17 Feb 2011 16:00:42 +0000. up 16.88 seconds
[16:11] <SpamapS> I get that right after root is mounted..
[16:11] <SpamapS> then after a bi
[16:11] <SpamapS> t
[16:11] <SpamapS> no instance data foudn in start-local
[16:11] <SpamapS> init: cloud-init-local main process (269) terminated with status 1
[16:12] <SpamapS> cloud-init start running: Thu, 17 Feb 2011 16:01:03 +0000. up 36.45 seconds
[16:14] <smoser> yeah.. i'm pretty sure its the ovf
[16:15] <smoser> it tries to mount cds. i had expected mount to timeout quicker
[16:15] <smoser> in trunk now it does a read of the device before trying to mount, the read (in my tests in kvm) returns immediately with failure, where mount was taking quite a long time.
[16:16] <smoser> if its -local thats taking a long time, its definitely not EC2.
[16:16] <SpamapS> ahhh
[16:16] <SpamapS> Makes sense
[16:18] <SpamapS> smoser: well then thank you for indulging my crazy idea section for the day. :)
[18:36] <hazmobile> niemeyer, heading back from doc.. eta 30m
[18:36] <niemeyer> hazmobile: How are you?
[18:36] <hazmobile> stitched and anesthesized
[18:37] <niemeyer> hazmobile: Awww..
[18:38] <hazmobile> on the bright side found aideable Android irc client
[18:38] <hazmobile> useable
[18:40] <niemeyer> hazmobile: Clearly ;-)
[18:40] <niemeyer> hazmobile: Which finger was it?
[18:49] <Edulix> hi people
[18:50] <Edulix> I want to do a cpu intesive task that can be divided into small chunks that can be sent to different computers, using some kind of clustering
[18:51] <Edulix> is ubuntu cloud recommended for cluster apps? I want to be able to plug in a new computer to my farm, then boot a thin client on it that runs a node of the cluster..
[19:08] <RobertLaptop> Edulix, can it yes.
[19:09] <Edulix> RobertLaptop: I think I might use MPI for this (I need to patch my app to make parallelizable)
[19:09] <RobertLaptop> Although it isn't designed to do cluster work so interconnections between the VM will likely be slower then in something designed to be a cluster.
[19:10] <Edulix> RobertLaptop: I'm new ubuntu cluster. ubuntu cloud always run images in an VM?
[19:11] <RobertLaptop> Edulix, Yes.
[19:11] <Edulix> ok then it's definitely not what I'm looking for
[19:11] <Edulix> thanks RobertLaptop
[19:14] <Edulix> RobertLaptop: isn't there anything like ubuntu cloud but that runs the images non virtualized? :P
[19:15] <RobertLaptop> Edulix, not that I know of not saying it doesn't exist just not something I know about.
[19:16] <Edulix> it'd be nice if there was a distro specialized in cluster support, so that you just install & configure your mpi-enabled software in an image, then start the server and connect new nodes
[19:17] <Edulix> and then the new nodes boot via ethernet etc
[19:26] <Edulix> RobertLaptop: how does ubuntu server deal with hardware?
[19:26] <Edulix> RobertLaptop: how do I configure a new node, for example. does it boot over network, or something?
[19:26] <RobertLaptop> Edulix, just to be clear I don't work for Canorical just a user.
[19:26] <Edulix> RobertLaptop: me too :P
[19:27] <RobertLaptop> Edulix, It deal with hardware like most Linux Server.
[19:28] <Edulix> RobertLaptop: so for each node, you need to install ubuntu cloud server etc
[19:28] <RobertLaptop> The cloud stuff uses either Xen or KVM to handle visualization and if you you setup your images just right the VM will get pretty close to bare metal ussage.