=== dendrobates is now known as dendro-afk [00:31] are there any videos around of a controller assigning images to nodes and making them work? [00:31] Also - any more elaborated documentation on the matter other than the community pages? [03:46] I have a questions regarding https://help.ubuntu.com/community/UEC/Topologies it shows breaking systems out with some systems label just NC. The question is does the system running other services not showing NC run VM's or is it just for managemnt stuff? [05:29] hello, i'm having a cpu load problem under medium network traffic on 10.04LTS, large instances, can anybody help me? [05:29] what is using the cpu? [05:29] apache [05:30] then that is your problem [05:30] it would make sense, except that traffic to the app does not significantly change [05:30] how does that conclude anything [05:30] you could have 0 traffice and apache do 100% [05:31] true; however, the application responds pretty much the same way to everybody that uses it. it's a social game, and tens of 1000's of players play it all day [05:31] however, at certain times of the afternoon load will surge in apache withotu the app doing anything different per se [05:32] sounds normal to me [05:32] its your problem to put it simply [05:32] not serving more users, different game play, etc. the same at 10am as at 4pm. but between 3 and 6, the normal load for its traffic (3) spikes to 50 [05:32] you can't expect the open source community to debug your app and shit [05:32] if it were a ton more players, i would think apache would become more allocated (looking at /server-status) [05:33] yeah bad assumptions once again [05:33] i don't expect that, certainly [05:33] and why is that bad assumptions? [05:33] because you don' need a load of users to have a mem and/or cpu leak or other indirect issue [05:34] it does not look like the problem with your app is anything to do with aws or ubuntu [05:35] i am merely hoping to hear that either a) this release of ubuntu has issues with high network load and to use something different or b) load on ec2 instances can vary even without your instance doing any work at all, since i've done a ton of debugging on the app and do not see an issue except when the app is in the cloud [05:35] keen to hear any evidence to suggest otherwise [05:35] for example, the app has been heavily loaded outside the cloud by browsermob, without issues [05:35] yeah sorry, not a good hope. [05:36] flaccid: hmm, ok, i'll take your feedback as a valid single data point [05:36] i do hope that other users here come out with other responses.. [05:36] a hope i share as well :-) [05:38] flaccid: as regards your suggestions, certainly there could be a ram leak, but i don't see evidence of it using the usual tools (top, iostat, free, etc) [05:38] flaccid: swap is never allocated, e.g. [05:39] flaccid: as far as CPU leak, also possible, i don't expect the community here to debug anything like that. however, again, if code doesn't change hour over hour yet system behavior does, it leads one to suspect OS or metal [05:42] can you show stats at least? [05:43] sure [05:44] flaccid: which stats would you want to see [05:44] flaccid: i have some output from iostat, for example [05:46] well [05:46] whats the actual issue first [05:46] cpu usage? [05:46] yes... [05:47] well, load overall, causes tasks that take say 300ms to return to take 1500ms, etc [05:47] is top basically correlating to the load reported in w [05:47] yes, it does exactly [05:48] well what can i say, apache is using that cpu [05:48] it doesn't look like a load accounting bug [05:48] how about you try maverick? [05:48] it definitely is the app but what i don't understand is, why the app doing the same thing it does all day, should suddenly demand so much more cpu [05:48] nobody here can answer that question [05:48] is maverick production ready? [05:49] sure [05:49] no ubuntu is production ready [05:49] thats the nature of it [05:49] but use it if you like [05:49] ok, i wil give it a tryout then, thank yo [05:53] i'd be curious to see if the same thing occurs on debian 6 [05:56] i might just switch to centos after reading this posting: http://alestic.com/2010/10/ec2-ubuntu-maverick#comment-484 [05:57] willkessler: we offer alternate builds @ RightScale if you want to try them out [05:57] that post seems a little speculativ [05:58] well it is. you don't report bugs on blog comments for a start. [05:59] as was pointed out [05:59] where do those rightscale builds live? [06:00] willkessler: http://support.rightscale.com/18-Release_Notes/OS_and_Software_Package_Support [06:00] im an engineer at rightscale, so please feel free to join us in #rightscale if you want [06:00] do i need to be a customer? [06:00] no [06:01] certainly, rightscale has gotten high reco's from my colleagues, and if we could afford it , i'd probably talk to you via the front door... but not quite yet [06:02] rightscale is free unless you would like a subscription [06:04] where can i read more about rightscale/free? that same site? [06:05] i'll just look at the links, nvm [06:05] thanks and good night [06:06] willkessler: http://support.rightscale.com/03-Tutorials/01-RightScale/1._Signing_Up_for_RightScale/Sign-up_for_a_Free_RightScale_Account [06:06] good night then. [06:07] Question I selected the wrong options during install is there a way to re-run the install script? [06:07] what install? [06:08] I selected CLC, Walrus, on the wrong box. [12:35] hi all, i wanna ask a question, now in every nodes i use Wget to get the ip of nodes, and i want this IPs be sent to cc, how should i do ? [12:35] in nodes, i use cron to execute the a shell script to run Wget every minutes [12:36] and i want it to report to the cc [12:36] how would I specify --addressing private for uec-run-instances? When I run 'uec-run-instances --attach-volume=$vol -- --addressing private $emi' I get an error. I use the TREllis modified cloud-utils packages for support of attaching volumes [12:37] if the cc can not receive the ip of nodes execeed 3 mins, i will think the node dies and start a new one... [12:38] TeTeT: hmm.. i also use --addressing private [12:38] superxgl: do you use euca-run-instances or uec-run-instances? euca-run-instances works fine [12:39] TeTeT: i use euca-run-instances, hmm.. don't know then .. [12:40] TeTeT: --run-args=RUNARGS pass option through to run-instances [12:42] TREllis: doesn't seem to work either, http://pastebin.ubuntu.com/568724/ [12:45] TeTeT: hmm I see the same on trunk [12:46] TREllis: ok, this is the syntax: uec-run-instances -v --attach-volume=$vol $emi --run-args="--addressing private" [12:46] note the quotes around the run-args, tried it without the -- and the quotes first [12:47] I normally put the emi at the end, but testing here that doesn't work either [12:47] superxgl: hmm, you try to detect if the nodes are still up and running? [12:49] TREllis: request for improvement: check if the volume is in an attachable state ;) if it's already been attached to a running instance, print an error or warning [12:51] TeTeT: hmmm I tried with a different runargs and it worked -k mykey [12:51] TREllis: but otherwise, works very nicely! Thanks for providing that option, will introduce it in my next class :) [12:51] TeTeT : yes..i call it heartbeat, if the cc can not receive the heartbeat execeed 3 mins, i would think this node dies... [12:52] superxgl: why don't you use an existing hearbeat system then? [12:52] TeTeT: Is there have one ? [12:53] TeTeT: did you get the run-args working then? [12:54] TeTeT: thanks for the suggestion, thats one on the list :) [12:54] TREllis: yes, got it working with the quotes and the -- [12:54] superxgl: please have a look at 'apt-cache show pacemaker' [12:55] superxgl: there's also https://wiki.ubuntu.com/UbuntuHighAvailabilityTeam/PacemakerHeartbeat for more doc on how to get going [12:56] infact, RoAkSoAx has been working on a resource agent for the CLC too [12:59] TeTeT: tnx:) go and checking now ..hmm.. i am really a newbie to this.. [12:59] superxgl: no worries, we've all been newbies once [13:01] TeTeT : :) [13:51] TeTeT：hmm...PacemakerHeartbeat is very cool :) but since it is my first time to do something like this, now i want to do it myself. my though is the nodes do a Wget like http://cc.com/?thisismyip=192.168.1.2 , that is the Ip be embedded into the GET parameter, so that the cc can get it's ip address.. [13:51] anyone here have any idea? [13:58] TeTeT: that documentation is outdated [13:59] superxgl: you might be interested in this: http://www.roaksoax.com/2010/10/high-availability-uec-clc-howto [14:01] TREllis: and as a matter of fact, the RA for the CLC is ready. I'll publish it later today [14:02] RoAkSoAx : tnx very much :) go and checking now:) [14:09] RoAkSoAx: cool :D nice work [14:10] TREllis: the hardest thing was to test it lol! [14:11] RoAkSoAx: heh [14:29] RoAkSoAx: great, thanks for the update! [14:31] TeTeT: wiki.ubuntu.com/ClusterStack/LucidTesting is the most updated, which still stands for natty [14:36] RoAkSoAx: ok, will bookmark it, superxgl might be interested in it as well [14:38] TeTeT: you might also be interested in http://www.roaksoax.com/2010/10/high-availability-uec-clc-howto (I'll update it later today with a better way to monitor the CLC though) [14:44] i am really interested in part: Load Balancing with Pacemaker/ldirectord :) [14:45] superxgl: ah!! I have one how-to for that I just haven't make it public yet [14:55] RoAkSoAx: i'll wait for it then :) [15:48] SpamapS, awake ? === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates [17:59] smoser: only have a couple minutes.. wassup? [18:00] i had some upstart questions. [18:00] hold on. let me get a pastebin [18:01] SpamapS, http://paste.ubuntu.com/568842/ [18:02] smoser: ok.. whats the ? [18:03] so i'm trying to force the order of those 3 things [18:03] cloud-init-local -> cloud-init-nonet -> cloud-init [18:04] ah ok [18:05] since they're all tasks.. [18:05] stopped and started are equivilent [18:05] start on stopped cloud-init-local ... start on stopped cloud-init-nonet ... [18:07] also since they're tasks.. [18:07] you can do this (I think this is evil but it may be necessary for fine grained control... have a task that just does [18:07] start cloud-init-local ; start cloud-init-nonet ; start cloud-init [18:07] note that && would maybe be better ;) [18:08] so do you not think what i have there would work ? [18:12] it should actually [18:12] the mounted is tricky tho [18:13] I'd only do it on the first one [18:13] unless you MUST block mounting / on all 3? [18:13] or rather, mounting anything except mount [18:13] mounted blocks mountall [18:39] well, we want to block. [18:40] smoser: just make sure those tasks *always* finish or the system won't boot [18:41] mostly because filesystem won't fire .. once / is mounted a lot is already done :) [18:41] not really. [18:41] not much is done on MOUNTED / [18:41] most stuff requires filesystems [18:54] true.. like rc-sysinit [18:55] and ssh, just about everything. and really you should'nt start before that [18:55] you're not guaranteed that tmp files in /tmp wont get deleted while you write them [18:58] can someone tell me where the uec website is located on the cloud controller? [19:11] SpamapS, shoot. my plan doesnt work [19:11] i think i know why [19:12] smoser: ? [19:12] http://paste.ubuntu.com/568842/ doesnt work. i get [19:12] cloud-init-nonet waiting for a network device. [19:12] cloud-init-nonet gave up waiting for a network device. [19:12] even though there was a netowrk device [19:13] i think the problem was that eth0 came up before cloud-init-nonet was running [19:13] (ie, while cloud-init-local was running) [19:15] SpamapS, make sense ? [19:21] jmgalloway: /etc/eucalyptus/cloud.d/www defines admin.xml, which refers to /var/lib/eucalyptus/webapps/root.war I think [19:23] oh thanks, I'll take a look there. I'm running apache and another website on my cloud controller..I just wanted to know if I could customize the web gui uec already has [19:27] smoser: right so you also need to create another task that emits an event on net-device-up or a timeout [19:27] getting messy :-( [19:27] network detection is messy [19:27] wait is nonet supposed to figure out if there is or isn't a network? [19:27] yeah. [19:28] so make a task that sends it a SIGUSR1 every time net-device-up fires [19:28] and handle that event by checking again [19:28] s/event/signal/ [19:28] thats how mountall does it [19:29] the problem is that network comes up before it runs. [19:29] i think [19:30] so its failing to detect that fact.. sounds like it should do/while rather than while .. ;) [19:30] ? [19:30] i was trying to force order... cloud-init-local -> cloud-init-nonet -> cloud-init [19:31] but (i think) the network comes up while cloud-init-local is running [19:32] so cloud-init-nonet does not get the 'net-device-up' to stop it [19:32] and so it waits until timeout [19:32] Right so you need to check if its up *first* [19:33] while (!netup) sleep(); [19:33] while (!netup()) sleep(); [19:36] who do i check if netup? [19:37] SpamapS, [19:37] suggestions? [19:37] smoser: ifconfig.. /sys .. /proc .. not sure actually ;) [19:37] yeah. [19:38] there may be a single command that does it [19:54] SpamapS, looks like ifquery might hold some information ... and also that /var/run/network/ifstate might just be suffiicent [19:55] smoser: yeah that sounds like you're on the right path. [19:55] gotta go afk.. bbl === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk [21:26] Question how do you define more then 1 DNS server? VNET_DNS="192.168.100.213" to you add 2 lines or 1 line with a comma or a space? [21:41] hggdh, ping [21:57] smoser: pong [21:58] i think i don tneed you now, hggdh [21:59] smoser: k [21:59] and i'm done with rig [22:02] smoser: thank you, I will reset it for Daviey === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates === dendrobates is now known as dendro-afk