14:00:39 #startmeeting oVirt Infra weekly meeting 14:00:39 Meeting started Tue Aug 28 14:00:39 2012 UTC. The chair is mburns. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:39 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:00:44 #chair eedri RobertM 14:00:44 Current chairs: RobertM eedri mburns 14:00:58 * RobertM here 14:01:00 * mburns has no idea what topics/agenda items we have 14:02:01 * eedri here 14:03:25 * ewoud here 14:03:42 think quaid said he was on a plane 14:03:43 Last over items are, New hosting for Gerrit / Jenkins, and Generally Jenkins fun 14:03:58 quaid, is around 14:04:07 But he is on airplane wifi 14:04:15 Last=Left 14:04:25 ok 14:04:31 #topic Hosting 14:04:38 who owns this particular topic? 14:05:08 quaid, I believe but he was looking for input on hosting providers from the group 14:05:14 I think the problem with hosting is now that nobody really owns it 14:05:19 if anyone quaid indeed 14:06:02 * ewoud recalls something about a proposal 14:06:05 14[[07Special:Log/upload14]]4 upload10 02 5* 03Mburns 5* 10uploaded "[[02File:Ovirt-node-2012-08-28.odp10]]": Presentation for oVirt Node at LinuxCon North America oVirt Workshop 2012-08-28 14:06:21 Part of the problem is everyone has there own pet ISP and we have very limited budget for the hosting. 14:06:49 * mburns has no opinion... 14:06:57 i've never really worked with any hosting providers 14:07:33 I only have worked with my employer and another local one 14:07:35 I think as a group we were suppose to find options for getting 2 dedicated boxes and a few VPS for around $150. 14:07:42 and a university 14:07:57 Preferably in the EU or US. 14:08:17 But it would be paid in use Dollars 14:08:44 first I think we should choose if we want to hire boxes or buy boxes and colocate 14:09:42 RobertM: according to http://ovirt.org/meetings/ovirt/2012/ovirt.2012-08-14-14.00.html you would kick off the hosting proposal on the ML 14:09:51 14[[07Special:Log/upload14]]4 upload10 02 5* 03Mburns 5* 10uploaded "[[02File:Ovirt-node-2012-08-28.pdf10]]": Presentation for oVirt Node at LinuxCon North America oVirt Workshop 2012-08-28 14:09:58 ewoud, It has to be rental boxes as the bugget is pretty small. 14:10:06 14[[07Category:Node14]]4 !10 02http://wiki.ovirt.org/w/index.php?diff=4336&oldid=4115&rcid=4436 5* 03Mburns 5* (+69) 10/* Presentations */  14:10:29 ewoud, It was kicked off and went pretty much no where. 14:10:39 14[[07Category:Node14]]4 !10 02http://wiki.ovirt.org/w/index.php?diff=4337&oldid=4336&rcid=4437 5* 03Mburns 5* (+4) 10/* Presentations */  14:10:48 RobertM: true 14:11:30 I don't think we are ready to make a call on this we need more options. 14:11:52 #info not enough information or options to make a decision here yet 14:12:09 #action RobertM to try and find 3 hosts that will fit the budget and needed spec's before next week. 14:12:26 anything else for hosting? 14:12:34 I think not. 14:12:59 anything else to discuss before going to jenkins? 14:13:07 mediawiki upgrade? 14:13:19 #topic medawiki upgrade 14:13:37 I think we don't run special plugins 14:13:42 see http://wiki.ovirt.org/wiki/Special:Version 14:14:16 didn't see any objects, but I don't have access 14:14:22 objections* 14:14:56 * mburns hasn't read the thread yet... 14:14:59 My only concern is downtime and what might break when we upgrade. 14:15:19 RobertM: we can manage that by announcing downtime and a backup 14:15:44 * mburns agrees on announcing downtime/backup 14:15:58 though next week at the earliest for the actual upgrade 14:16:12 ewoud, It is more how long vs the process. 14:16:27 RobertM: not long, my colleague did it for our internal wiki as well 14:16:52 I'll poke on the ML and try to pick a date 14:17:27 #action ewoud to work with quaid and RobertM to pick a date for upgrade. 14:17:55 Ok next topic? Anything else before Jenkins? 14:18:16 #topic Jenkins 14:18:37 * mburns puts hard stop at 7:40 14:18:47 currently it's 7:18 14:19:37 RobertM: eedri these are your points usually 14:19:52 Jenkins is still dual brain. Ovirt.org does the master branch and jenkins.ovirt.info does patches. There have been a few issues with the patch process in part because of a copy and paste mistake on my part when I brought 3 new slaves online. 14:20:20 ewoud, were there open issues on jenkins from last meetings? i can't remmeber 14:20:52 ewoud, i think next thing we can push forward is running automation test on an hypervisor 14:20:53 eedri: I wasn't really attending the last meeting, but I'll check the minutes 14:21:09 ewoud, as long as we have a physical server to run it on 14:21:34 ewoud, meaning a test that install install rhevm + add host + run rest api/cli/sdk tests 14:21:36 eedri: but that depends on the hosting 14:21:36 jenkins.ovirt.info has 2 physical and 3 virtual servers. 14:21:54 ewoud, we can start using ovirt.info hosts 14:21:58 true 14:22:24 Even if we merge the 2 once on hosting those 2 boxes could become slaves. 14:23:23 RobertM, we'll need a phsyical host that won't be slave in jenkins 14:23:40 RobertM, since the hypervisor might reboot during install,etc.. 14:23:53 eedri: any chance of using KVM nested virtualisation? 14:24:04 but the per patch processing is kicking jenkins.ovirt.info butt. Even with the 3 new VM's The jobs run an hour or more behind in the Morning hours in EST timezone 14:24:33 * eedri thinks of using my server jenkins-upstream-f16-host01 14:24:54 its a test physical server i've added 14:25:18 RobertM, we can connect it to ovirt.info instead if it will be faster 14:25:51 ewoud, never tested nested virtualization 14:26:24 I am not sure what is needed I am getting some more Hardware at the end of the year what do we need? 14:27:02 RobertM, does ovirt.org specify min req for a hypervisor? 14:27:06 Ebay hardware can be found cheap for less then $200 you can get a pretty decent piece of hardware. 14:27:18 nested virt is questionable... 14:27:23 it's pretty slow in general 14:28:29 i think we can start with the current host, i'm not using it at the moment 14:29:09 eedri, Do we know how we are going to manage these hosts? PXE? Etherboot? 14:30:02 We need something that will allow us a lot of control over the OS so we can install what is need and clean up failures. 14:31:03 or is it something that we just need ssh access to a base f17 build? And we can just install unistall software on a static host. 14:32:12 RobertM, well 14:32:34 RobertM, once we'll have foreman+puppet we'll be able to provision and reinstall hosts 14:32:57 RobertM, but, if we have a static host, f17 for example. we don't need to pxe install it all the time 14:33:24 RobertM, the test will run vdsm-bootstrap and install vdsm/libvirt/etc... but not the OS 14:33:45 you're talking about running vm's on a base host and how to manage the vms, right? 14:34:18 base os should not ever be reprovisioned 14:34:21 mburns, i'm talking about running automatic tests via restapi/cli/sdk 14:34:27 mburns, That is what I am asking? It sounds like they want raw hardware because of the virtual needs. 14:34:30 mburns, exactly 14:34:36 eedri: right 14:34:50 mburns, it will need to reprovision only if we want to use the host for other os, like rhel 14:35:03 mburns, if we want to test ovirt for rhel hypervisors 14:35:10 two options for the vms: run local pxe server on the machine, or have a set of base disk images that you always reset to 14:35:24 we're using cobbler 14:35:41 for rhevm, i guess we can install a cobber server for ovirt 14:35:52 start your vms against copies the base images, then after the test, reset back to base images 14:35:59 does cobber support WAN? 14:35:59 then no need for cobbler 14:36:06 we don't want cobbler 14:36:22 we want puppet/foreman eventually 14:36:47 but short term using either base disk images that are pre-installed with the os we want or using a simple local tftpboot server should be sufficient 14:37:05 mburns, yes, that is ideally, though haven't had the chance to use it yet for installing vms in a test 14:37:10 mburns, eedri Unless we are planing to get a colo so place how do we pxe boot systems at remote locations unless I am missing something pxe can do. 14:37:35 RobertM: networking is completely internal to the machine (libvirt network) 14:37:53 * ewoud back 14:38:08 run tftpboot services on real hardware and have libvirt network pxe boot from physical host 14:38:12 never leaves the machine 14:38:13 * RobertM is missing something 14:38:32 * mburns misses having a whiteboard to explain... 14:38:47 etherpad? 14:38:50 RobertM: 1 physical host running any OS (rhel, fedora, whatever) 14:39:06 mburns, Maybe you can explain it this afternoon in person. 14:39:09 on top of that machine you run vms for engine and vdsm hosts 14:39:21 you also run a tftp server 14:39:50 I have do those kind of setups inside a LAN but I don't know how we could do that over a WAN 14:39:52 the network definition of the vms tells it to pxe from the physical server 14:39:56 do=done 14:40:13 RobertM: yeah, we can discuss this afternoon if you want 14:40:16 * eedri also doesn't have experience over WAN 14:40:31 but basically, there is no WAN, you run the test completely within 1 physical host 14:41:20 If we want to get like 10 core-duo and stack them in someone basement / a colo I can it being done pretty easyly but tiring to do it over the WAN I am having trouble. 14:41:43 RobertM: ok, lets discuss later today, then we'll summarize for people on the list 14:41:52 mburns, That works for me. 14:41:52 we have to run anyway 14:41:55 Yep 14:41:59 +1 for contining this issue on the list 14:42:07 ok anything else? 14:42:16 puppet status? 14:42:31 can we defer that? or take it on list? 14:42:36 sure 14:42:49 thanks 14:43:09 #action mburns and RobertM to discuss jenkins test deployment and report back to list 14:43:19 #action take puppet status to list 14:43:23 ok, thanks all 14:43:25 * mburns has to run 14:43:28 #endmeeting