[14:57] Hello [14:58] how can i capture all packets which the SYN flag using tcpdump? [14:59] kraut: It got you stumped too? [14:59] Gargoyle: sorry, i need to fix a breakdown [14:59] no time atm [15:00] np [15:02] nealmcb: Around? [15:02] howdy [15:03] nealmcb: Hey. [15:03] back home? [15:03] Yeah. [15:03] nice [15:03] I also finally got round to looking at your ubuntu-jeos changes. [15:03] Apart from a few minor thing, it looks great! [15:03] :-) [15:04] I stumbled a while on the need to mark the vm type..... [15:04] I also am wondering how this might hook in with some other similar tools... [15:04] Such as? [15:05] libvirt and virt-manager [15:05] qemu-make-debian-root [15:06] I'm working on libvirt and virt-manager right now. [15:06] https://bugs.edge.launchpad.net/ubuntu/+source/qemu/+bug/63429 [15:06] Launchpad bug 63429 in qemu "[Edgy] qemu-make-debian-root hasn't worked since Debian Sarge, and not even then" [Undecided,Confirmed] [15:07] I still haven't gotten back to getting networking working right - need for local bridge-aware interfaces or something in qemu [15:07] have you also read that forums thread? [15:07] Which forum thread? [15:09] xen-create-image: http://www.debian-administration.org/articles/533 [15:09] http://ubuntuforums.org/showthread.php?t=549222&page=4 [15:11] how do you intend to package it? hardy and backport for gutsy? [15:12] the forum thread has lots of questions about this release vs the demo last summer, which seemed to be a much smaller or somewhat different thing - can you shed light on that? [15:14] No :) I'm still trying to get some sort of overview over the entire situation. [15:14] still no jeos iso released. is the md5 341ca65a187c71643079a2f9ee5523b5 ubuntu-7.10-jeos-i386.iso (seen in the wild) [15:17] hello [15:17] huh - what is http://cdimage.ubuntu.com/jeos/releases/7.10/release/ [15:20] i cant install landscape-client, i get : landscape-client: Depends: smartpm-core (>= 0.52-gutsy1-landscape1) but... [15:20] nealmcb: that is virtualization based ubuntu [15:21] m1r: right - I'm chatting with soren about it - and wondering why that one is there but not at http://releases.ubuntu.com/7.10/ [15:21] ah [15:21] i was looking for it other day and cojuldnt find it [15:23] m1r: I think it only just got moved there a few days ago. [15:24] nealmcb: I'm not sure why it's not on releases. [15:24] ok , tnx soren [15:24] i thought i was searching for somthing unavailable :) [15:25] whats the problem with landscape-client? [15:25] avatar_: i get error when adding 2nd host [15:25] nealmcb: I've just asked our release manager. He put it there, so he should know :) [15:26] m1r: Have you added the repository to the machine or are you just copying the landscape-client deb around? [15:26] deb http://landscape.canonical.com/packages/gutsy ./ [15:26] thats were you can get it [15:26] soren, added rep and key [15:27] % apt-cache policy landscape-client [15:27] landscape-client: Installed: 0.15.0-gutsy1-landscape1 [15:27] nealmcb: the general answer to that is that releases.ubuntu.com is for the highest-traffic things, and has limited space [15:27] and when i try install i get missing dependencies [15:27] landscape-client: Depends: smartpm-core (>= 0.52-gutsy1-landscape1) but... [15:28] Please put the output of "apt-cache showpkg smartpm-core" onto pastebin [15:30] soren ; http://pastebin.ca/773448 [15:30] m1r: What does "apt-get install smartpm-core=0.52-gutsy1-landscape1" do? [15:33] let me check [15:33] i would like landscape-server to run localy [15:33] not so happy with canonical having al my data [15:33] paranoia avatar_ ? :) [15:34] but sales@canonical.com is slow with answering my questions [15:34] soren : smartpm-core: Depends: libc6 (>= 2.6-1) but 2.5-0ubuntu14 is to be installed [15:34] m1r: imho it conflicts with privacy terms [15:35] m1r: what version of landscape are you installing? [15:35] hi ivoks [15:35] the one from CP [15:35] canonical can view all information from our and our customers servers. Installed packages, users etc [15:35] m1r: You're running feisty on that box... [15:35] soren, i think is gutsy [15:36] m1r: No :) [15:36] m1r: it isn't [15:36] m1r: 2.5-0ubuntu14 is from feisty. [15:36] omg :P [15:36] m1r: If it doesn't know a libc > 2.6, it's not running gutsy. [15:37] ye i see [15:37] There's a feisty version of landscape, isn't there? [15:39] ok solved, many tnx guys [15:40] m1r: np [15:40] soren: so, how are you? :) [15:40] ivoks: I'm rocking! [15:42] soren - cool - thanks for the trunk commit - yeah the kvm vs qemu needed cleaning up also [15:42] nice changes [15:42] though I don't know that I rate as an "author" :-) [15:42] nealmcb: Sure you do. [15:48] ivoks: howdy! thanks for your blog writeup - it was indeed a pleasure to hang out with you! [15:50] nealmcb: hi! no problem... :) [15:50] hopefully, we'll meet again on next UDS, or sooner ;) [15:50] anyone use drbd? [15:51] Gargoyle: me [15:52] after you fail over, and back again. if drbdadm state all returns Primary/Unknone and Secondary/Unknown on the two machines, is there a command I should be running to resync them? [15:52] they sync automaticaly [15:52] i have primary/primary [15:53] and when one dies, on reconnect, becomes secondary, but outdated [15:53] then they sync, and when that's over i have a mechanisam that makes it primary again [15:53] hey mathiaz [15:53] hi zul [15:54] ivoks: I am using heartbeat, and it put things back over onto node 1, which is reporting as primary. [15:54] but node 2 is unknown.... Is this because heartbeat might have switched back too soon [15:55] oh... are you mounting fs? [15:55] ypu [15:55] yup [15:55] you can't do that if it's secondary [15:56] ivoks: Nahh that's handled by heartbeat. only gets mounted on promary [15:56] ok [15:57] so, primary has Primary/Unknown [15:57] and the 'rebooted' one has? [15:57] It not been rebooted yet, but it has Secondary/Unknown [15:57] then you have to connect it [15:58] It's like each node is in the correct state, just not aware what the other one is upto! [15:58] if you want to drop it's data: [15:58] drbdadm -- --discard-my-data connect all [15:58] and on the primary: [15:58] drbdadm connect all [15:59] do you have to do that each time it fails over, or was it just in the middle of a sync or something? [15:59] you have to do that only on split brain situtation [16:00] so, in real world, you will never use it :) [16:01] right, so if I unplugged my nic cables. and watched it fail over. [16:01] then hb should take care of making secondary - primary, and mounting filesystem [16:01] then plugged them back in, and heartbeat switched back to node 1 (even though I have told it not to) could that have caused split brain [16:01] no [16:01] that should be normal situation [16:02] old primary should resync first [16:02] to become secondary [16:02] after that it can become primary and get services back [16:02] imho, two primaries would be better solution [16:03] and some redhat tools instead of heartbeat [16:03] It'f for a high available mysql [16:03] i see [16:04] i did mysql ha like this: [16:04] mysql master-master replication with VIP over heartbeat [16:04] but i will replace heartbeat with redhat-cluster-suite [16:06] with master-master replication, you can also do load balancing [16:06] is rh cluster suit a lot better? [16:06] well, it includes a lot more than heartbeat [16:07] sine i'm using GFS on drbd, having redhat-cluster-suite is just natural way to go... [16:08] ivoks: I read quite a bit on GFS, got confused with some of it. Can GFS give you no single points of failure with 2 servers? [16:09] GFS is just a filesystem which can be mounted on more than computers [16:09] drbd provides high availability [16:09] with drbd you can create 'network mirror' [16:10] and then on top of drbd is GFS, and if you have primary/primary drbd, then you can mount /dev/drbd0 on two machines [16:10] and write on it at the same time [16:12] I just checked node 1 with the nic's stull unplugged. drbd still thinks it's primary. [16:13] I pulled the cables out 10 mins ago. [16:13] well yes [16:13] it is primary [16:13] it doesn't know if node1 or node2 is out [16:14] grr [16:14] both nodes think they are the one that survived, and both are primary, right? [16:14] but heartbeat does. since it released the vip to node 2. [16:14] Gargoyle: did you fixed your hb issue? [16:14] kraut: Not yet [16:15] Gargoyle: did you add drbd relevant stuff to ha.cf? [16:15] soren: have you tried installing the jeos iso in qemu? I just get initramfs on reboot [16:15] respawn hacluster /usr/lib/heartbeat/dopd [16:15] apiauth dopd gid=haclient uid=hacluster [16:16] ivoks: Not dopd [16:16] god, damn it [16:16] Gargoyle: this way hb desides which one should be primary [16:16] decides [16:16] the guide I have says respawn hacluster /usr/lib/heartbead/ipfail [16:16] that's for IP [16:17] you need it also for drbd [16:17] ahhh! [16:20] soren: since http://releases.ubuntu.com/7.10/ doesn't have all the releases, it should say so and link to the others..... [16:22] ivoks: does it matter which order they are in the file? [16:22] no [16:23] nealmcb: I honestly don't remember. [16:24] Gargoyle: one more thing [16:24] Gargoyle: in /etc/drbd.conf [16:24] yup... [16:24] Gargoyle: you need handlers { outdate-peer "/usr/lib/heartbeat/drbd-peer-outdater"; } [16:25] Gargoyle: under common {} [16:25] my haresources is just - mysql-01 drbddisk Filesystem::/dev/drbd0::/drbd::ext3 mysql 192.168.0.215 [16:25] not haresources [16:25] ha.cf and drbd.conf [16:26] yeah, but I was just checking my haresources looked ok to you - since this guide seems to have an error or two. [16:27] ivoks: I have "outdate-peer /usr/sbin/drbd-peer-outdater" in the resource section. [16:28] well, do you have that file? i don't :) [16:29] there is /usr/lib/drbd/outdate-peer.sh, tough... [16:29] damn! [16:29] :) [16:29] Gargoyle: don't worry [16:29] so do I change it on my resource or common? (I only have 1 resource so I am assumng it doesnt matter!) [16:30] HA and LB isn't trivial to set up... [16:30] i've put it in common [16:30] but should be ok in resource too, yes... [16:33] dendrobates: hi there [16:33] ivoks: hi === dantalizing is now known as dantalizing_away [17:08] someone say they were working on libvirt and virt-install? [17:16] Me. [17:16] zul: ^^ [17:18] soren: ah ok...good.. [17:20] let me know if you need to bounce stuff off of me [17:25] zul: I'll keep that in mind. === dantalizing_away is now known as dantalizing === gamble6x is now known as gamble|fude === oly_ is now known as oly- [18:59] soren, zul - so do we need a plan for bridging/combining libvirt, virt-manager, virt-install, ubuntu-jeos-builder for xen and qemu? [18:59] has anyone written up the virtualization blueprint based on the uds discussions? [19:00] Possibly. I need to look into how much of ubuntu-jeos-builder's job virt-install can do. [19:00] * nealmcb nods [19:18] afternoon [19:21] zul afternoon here in Boulder too :-) where are you? [19:23] ottawa,canada === dantalizing is now known as dantalizing_away === dantalizing_away is now known as dantalizing === dantalizing is now known as dantalizing_away === dantalizing_away is now known as dantalizing === dantaliz1ng is now known as dantalizing [23:03] what determines the contents of "Suggested Packages" when you install something? [23:04] IE apt-get install blah [23:04] Suggested Packages [23:04] debian/control file [23:04] some-other-package [23:04] who would change one? moatas? [23:04] mralphabet: Each package can Suggest: other packages. [23:04] er, motus? [23:05] Depends on the package. [23:05] k, thanks [23:05] If it's in universe, yes. If it's in main, core-dev's can do it. [23:05] mralphabet: motus for universe, core for main [23:05] i feel like in real server [23:05] everything is redundant :) [23:06] *g* [23:07] I also feel like a real server... [23:07] ...but that's because I seem to never stop working. [23:09] hehe [23:09] then you are ubuntu-server :) [23:10] we should rethink heartbeat :/ [23:11] i know i was the one asking for it, but every day it takes one year of my life... [23:20] ivoks: Heh. Have you looked into redhat-cluster-suiter? [23:20] *suite [23:22] yes [23:23] i'm planing to replace my heartbeat installations with redhat cluster suite [23:50] zul, soren: fwiw, I registered the libvirt project for the hardy libvirt package and configured the cvs import [23:52] nealmcb: Ah, handy. [23:52] I'm happy to hand it over to Daniel Veillard or the "registry" group or whoever [23:55] as long as you stick around it shouldn't be necessary [23:56] or put ubuntu-server as the team to own it [23:56] so that team members can update things where necessary