13:01:14 #startmeeting oVirt Node Weekly Meeting 13:01:14 Meeting started Tue Sep 4 13:01:14 2012 UTC. The chair is mburns. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:01:14 Useful Commands: #action #agreed #help #info #idea #link #topic. 13:01:25 #chair fabiand jboggs pmyers_ READ10 13:01:25 Current chairs: READ10 fabiand jboggs mburns pmyers_ 13:01:27 * fabiand is here 13:01:34 #topic agenda and roll call 13:01:42 * jboggs here 13:01:46 agenda: 13:01:50 1. status updates 13:02:02 2. other topics 13:02:06 * READ10 here 13:02:29 #topic status updates 13:02:30 mburns: might be good to discuss the ovirt node openstack integration work 13:02:41 pmyers_: yes, on my list for other topics... 13:03:00 hey all, welcome back 13:03:15 first time we've had one of these in a few weeks between vacations and linuxcon, etc... 13:03:48 updates from me: 13:04:12 #info ovirt workshop node presentation was well received, good questions, seemed like genuine interest 13:04:36 #info people very excited about plugins and stateless, especially 13:05:35 jboggs: what are you working on at the moment? 13:05:53 some plugin fixes and other rhev-h 6.4 items 13:06:01 * RobertM here 13:06:50 jboggs: excellent 13:06:52 fabiand: ? 13:07:12 #info jboggs working mostly on bug fixes 13:07:18 mburns, I'm actually doing some bug hunting today 13:07:23 #info both in core ovirt-node and plugins 13:07:37 #info fabiand working on bug fixing and openstack integration 13:07:40 mburns, but besides that I'm looking into the openstack+node topic 13:08:17 ok, great 13:08:24 is gerrit dead? 13:08:46 #info additional update: ovirt-node is currently not usable due to a kernel bug 13:08:51 * mburns looking for bz number 13:09:15 lhornyak, Gerrit is workingfor me 13:09:43 RobertM: ah yes, responding, just dead-slow when pushing changes 13:09:45 #link https://bugzilla.redhat.com/show_bug.cgi?id=845660 13:10:00 jboggs: can you chase down where this is? 13:10:05 sure 13:10:09 looks like vdsm might have a workaround for it 13:10:23 but i heard last week that there was something that still needed fixing in vdsm as well 13:10:42 #topic release status 13:10:51 mostly just info 13:11:15 #info ovirt-node 2.5.1 release doesn't build successfully for EL7 builds 13:11:29 #info a couple patches posted to gerrit to fix some of these issues 13:11:46 #info mostly around dracut and systemd 13:12:04 #info need a new upstream tarball to incorporate the changes though 13:12:32 #info so looking to make a 2.5.2 release this week, if we can get those changes merged 13:13:21 i've tested them manually in mock environments and the changes seem to work correctly 13:14:08 #topic ovirt-node/openstack integration 13:14:15 fabiand: any updates here? 13:14:25 Yep 13:14:51 I've been working with apevec from the openstack camp to see whatwe can do on the node side to allow a better integration 13:15:09 we created a simple plugin which pulls in the appropriate openstack components 13:15:17 #info fabiand and apevec working together to make integration simpler 13:15:22 #info simple plugin created 13:15:27 #undo 13:15:27 Removing item from minutes: 13:15:31 #info simple plugin created for openstack components 13:15:47 The plugin is still a bit rough around the edges and the whole functionallity has not yet been tested 13:15:51 fabiand: just fyi, i saw your email earlier, but haven't read through it yet 13:15:53 but the daemons seem to be working 13:15:58 np :) 13:15:58 fabiand: ack 13:16:13 one thing that came up is: there are a lot of things to configure in openstack 13:16:18 * mburns has 7101 unread emails right now 13:16:24 and there rose a demand for some kind of configuration management system 13:16:43 mburns, wey hey ... 13:17:26 #info potential problem identified with openstack plugin 13:17:30 #info lots of things to configure 13:17:41 #info need some sort of configuration management 13:18:03 #info investigating having some sort of puppet plugin 13:18:09 besides that we singled out the important conf params and added cmdline parsing for them to do the initial configuration 13:18:15 so compatible to what we have been doing in the past 13:18:41 excellent 13:19:14 fabiand: what are you doing about vdsm? 13:19:17 disabling it? 13:19:24 removing it? 13:19:29 ah 13:19:30 yes 13:19:41 there are also components conflicting with openstack components, like vdsm 13:19:50 the idea is to carve them out, and put them in separate plugins 13:20:03 this way we are getting a node core, which the other "spins" build upon 13:20:11 would a removal tool in edit-node be useful? 13:20:18 fabiand: ok, what are you doing for testing right now? 13:20:29 jboggs: i'd much rather avoid removing in plugins 13:20:35 ack 13:20:39 jboggs, I'd rather go with bsae node + plugins, rather than fat node - rpms + plugin 13:20:55 mburns, apevec is currently playing with the plugin and will be testing it 13:21:07 one of our basic assumptions is that plugins add net-new (or in some cases updating) 13:21:09 when it's basically working I can setup a job to do the regular testing 13:21:24 btw. the image size was increased by about 10m if i recall it correctly 13:21:30 if we start removing, then we open a whole new can of worms 13:21:53 #agreed want to limit plugins to additions only, no remove 13:22:16 #info image size increased by ~10MB 13:22:25 fabiand: re: image size, that's not too bad 13:22:34 nope 13:22:42 one thing we have to watch is the 256MB limit 13:22:50 okay, we are not reaching that 13:22:51 if we get bigger than that, we break 13:23:09 b/c Root and RootBackup are only 256MB 13:23:13 the worst case was openstack (glance, nova, keystone) + puppet client (jsut into the blue) at abotu 197mb or so 13:23:19 ahh okay 13:23:25 * mburns hit that when he accidently didn't minimize an image 13:23:32 btw. we came to puppet because it is common in the openstack domain 13:23:48 fabiand: seems to be common in general these days 13:24:18 yep, also present in foreman .. 13:24:31 fabiand: jboggs: we have some work to do in general with minimizing and plugins, but that's the next topic 13:24:40 fabiand: anything else here? 13:25:06 The current plugin code is hosted here http://fedorapeople.org/cgit/fabiand/public_git/ovirt-node-openstack-plugin.git/ 13:25:21 #link http://fedorapeople.org/cgit/fabiand/public_git/ovirt-node-openstack-plugin.git/ 13:25:22 I'm done for now 13:25:30 ok, thanks for the update 13:25:50 #topic plugins 13:25:56 jboggs: any updates here? 13:26:30 booting on latest node should be working once dracut patch is acked 13:26:51 jboggs, which one? 13:26:54 jboggs: ok, /me will look today 13:26:55 and also the patch for isomd5sum but otherwise stable 13:27:05 ok, great 13:27:20 some things came out of the workshop, w.r.t. plugins 13:27:26 1. we need example plugins 13:27:40 2. we need guidelines for writing/developing plugins 13:28:14 yeah 13:28:21 3. we need to make sure we have support for things like firewall and service generation 13:28:59 wouldnt that be handled with the rpm itself? 13:29:12 nothign against adding it just seems like the %post would handle it 13:29:16 service stuff probably yes 13:29:30 firewall config probably needs work on our part 13:29:46 pretty sure changing the firewall in %post is frowned upon 13:30:05 apevec, and myself also came to this points - the question is if we should provide a common api to wrap the different services providing the functionality 13:31:03 my thought is that the services should be installed 13:31:28 * fabiand thinks that we still need a way to en-/disbaled services at runtime ... there was a bz# 13:31:38 and started based on whether or not they're configured 13:31:43 similar to how we do cim 13:32:18 and we can open/close firewall ports the same way 13:32:55 fabiand: i agree, but it's just not easy to do in livecd environment 13:33:22 and enable/disable would probably require either commandline access or new UI 13:33:44 once the services are figured out as a base it shouldnt be hard layering on top of that firewall etc 13:33:54 ok 13:34:03 firewalld should mak this easier too 13:34:14 I have the services bz I belive 13:34:52 jboggs: i was thinking we could turn cim and snmp into example plugins 13:35:10 and drop them from the base image 13:35:23 maybe we can also use the openstack plugin too 13:35:27 yeah that should be fine 13:35:32 I've tridd to make it quite clear, it has also cmdline parsing 13:36:09 fabiand: yes, openstack is a good candidate as well 13:36:19 but it's much more complex, i think than cim and snmp 13:36:31 okay 13:36:44 though i haven't really looked at it yet 13:37:10 ok, anything else on plugins? 13:37:17 re. big cmdline - I'd look at removing direct access to /proc/cmdline, instead reading e.g. /etc/cmdline 13:37:40 ovirt-early would append /proc/cmdline to that file 13:37:58 while plugin could inject base values into it 13:38:11 just an idea, I'll try to express it as a patch :) 13:38:18 apevec, I'd favor /run/cmdline - because the cmdline can change between boots, /etc is more static in my eyes 13:38:30 ahyes, /run is the new location 13:38:38 apevec: yes, something like that could work 13:38:44 I think dracut is doing smthg like that 13:38:57 apevec: i was thinking more like having /etc/default/ovirt file 13:39:09 /etc/default/ovirt should be gone :) 13:39:10 which we already write out to 13:39:19 right, but whatever replaces it 13:39:32 it's not FSTD compliant, I have posted one mega RFC some time ago 13:39:41 * mburns makes note to review apevec 's patch to remove /etc/default/ovirt 13:39:50 apevec: it was posted, but just bad timing 13:40:00 Node is not devel frozen now? 13:40:05 no time to review it 13:40:12 apevec: no, not frozen 13:40:25 was just too busy with other issues to review and incorporate 13:40:26 so time is right for bigger changes! 13:40:27 I updated apevec 's patch a couple of times 13:40:37 (rebased) 13:40:51 apevec: yes, time is right to review and get it in for 2.6.0 13:42:15 ok, anything else to cover today? 13:42:43 fabiand, could you please take over that fstd patch? 13:42:52 I even lost link to it :) 13:42:53 apevec, sure, I'd be happy to do so 13:43:09 apevec, this one http://gerrit.ovirt.org/#/c/2686/ ? 13:43:21 yep, that's the one 13:43:46 mburns, I'm good 13:44:29 #action fabiand to take over http://gerrit.ovirt.org/#/c/2686/ and update it 13:44:54 fabiand, did you push your rebases to gerrit? There's only original patch set 1 13:44:54 ok, if nothing else... 13:45:18 apevec, no, I never pushed it, just had it locally, and rebased it with my other pending branches 13:45:38 apevec, I'll have to rebase it on the lastest cdoe style changes, I'll push it afterwards 13:45:39 Hi 13:46:00 Anyone know how I can find a revision ID from which a branch was made? 13:46:16 gitk ;) 13:46:30 fabiand, Smart-ass ;) 13:46:45 * mburns could have pointed to gitweb... 13:46:55 hehe 13:46:55 dneary, revision ID = hash? 13:46:56 but that would be even uglier... 13:46:58 In svn there was a "stop on branch" argument to svn log 13:47:15 ok, /me ending node meeting 13:47:17 #endmeeting