17:07:32 #startmeeting XenAPI 17:07:33 Meeting started Wed Dec 12 17:07:32 2012 UTC. The chair is johngarbutt. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:07:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:07:37 The meeting name has been set to 'xenapi' 17:07:54 Hi all, lets start the XenAPI meeting 17:08:00 #link http://wiki.openstack.org/Meetings/XenAPI 17:08:10 Usual page has the agenda 17:08:24 let's start with blueprints 17:08:30 #topic blueprints 17:08:53 anyone got any progress to report on blueprints/features? 17:10:05 over in citrix we are working on refactoring the volume drivers in nova 17:10:06 https://blueprints.launchpad.net/nova/+spec/xenapi-volume-drivers 17:10:11 #link https://blueprints.launchpad.net/nova/+spec/xenapi-volume-drivers 17:10:26 seems to be getting there 17:10:52 anyone else? 17:10:55 o/ 17:11:08 just a hello, that is. 17:11:15 ehlo 17:11:15 And a hello from me also. 17:11:24 cool, hi people 17:11:35 we have pending reviews for this blueprint: 17:11:49 (in quantum) 17:11:52 #link https://blueprints.launchpad.net/quantum/+spec/xenapi-ovs 17:12:03 #link https://review.openstack.org/#/c/15022/ 17:12:32 this is good work done by internap, so if people can take a look, but maybe more for a quantum specific meeting 17:12:44 johngarbutt: I'll see if we can find some folks to take a look at that. 17:12:53 awesome, thanks. 17:13:14 there is this follow on one two #link https://review.openstack.org/#/c/15023/ 17:13:41 cool, shall we move on to Docs stuff? 17:13:49 #topic docs 17:14:05 so we have this doc: http://wiki.openstack.org/HypervisorSupportMatrix 17:14:16 I have been updating it now and then 17:14:31 are there any major things that are missing 17:15:01 We're very keen to know if there are major items missing from the matrix 17:15:11 I spotted networking and storage features, i.e. support for particular vendors seems to be missing 17:15:25 i have a question in terms of things 17:15:36 does stuff now run in vpx machines with xenserver or on ubuntu nodes? 17:16:14 ubuntu VM on top of XenServer, or any other linux PV guest for that matter 17:16:32 is it hard to setup contra kvm ? 17:17:09 it is a bit harder, because of the extra VM, but devstack does deal with both 17:17:32 maybe you raise a good point, we need some docs on how to get started with Ubuntu packages an XenServer 17:17:52 johngarbutt: so you create a vm inside the xenserver? 17:17:54 automated or ? 17:18:01 #action raise doc bug for XenAPI in Ubuntu docs 17:18:23 depends what you are doing, devstack will create an Ubuntu VM for you, if you run it on your XenServer 17:19:03 we should probably take this offline onto #openstack later or the ML? 17:19:08 http://wiki.openstack.org/XenServer/Development 17:19:22 that contains part of getting a domU set up for nova-compute 17:19:25 johngarbutt: are we trying to push more xenserver or xcp? Or that should be the exact same? 17:19:41 does the feature lag time between the two matter? 17:19:47 And here: https://github.com/openstack-dev/devstack/blob/master/tools/xen/README.md 17:20:04 #link http://docs.openstack.org/folsom/openstack-compute/install/apt/content/introduction-to-xen.html 17:20:21 XenServer vs XCP shouldn't matter in that list 17:20:35 which feature lag time were you thinking about? 17:20:40 only sad thing is that xenserver doesn�'t have ceilometer support :) 17:21:07 I thought it did have something, armandomi at Internap is a guy who knows 17:21:15 not sure it is documented yet... 17:22:21 johngarbutt: got time afterwards to chat a bit ? 17:22:37 zykes-: what do you mean there? 17:22:55 afraid not today, but certainly should arrange some time soon 17:23:08 johngarbutt: I thought the flow was opensource xen -> xcp -> xenserver in terms of feature dev and flow. 17:23:27 but maybe I have some of that reversed 17:24:03 just chat about xenserver vs kvm 17:24:40 I see you mean, at the moment it is more Xen -> XenServer -> XCP 17:24:50 but I think there is a hope to change that 17:25:25 and get more people involved at the XCP level, patches are already welcome, just not great process around there yet 17:25:34 zykes-: shoudln't ceilometer just take nova message formats? 17:25:47 what isn't supported from the xen side? 17:26:22 not sure it works like that yet, I herd discussion on people moving that way, but not been tracking it closely enough 17:26:25 pvo, The confusion may have come from a recent XCP which was produced post-XS6.0 but pre-XS6.1 with some XS6.1 features included in it. 17:26:28 I think there is interest on our side to get more involved at the XCP level as well. 17:27:05 BobBall: maybe so. I know I need to get more visibility into it. Been kinda distracted by this whole "launch a product thing" :) 17:27:18 pvo: cpu pulling i think 17:27:32 Quite understandable! :) 17:27:57 zykes-: this is cpu polling of the VMs? 17:28:15 I'm actually interested in that as well. Are you using xen/xcp/xenserver and need that? 17:28:34 pvo: atm i'm on kvm, but concidering xen 17:28:51 zykes-: is there a bug opened for it yet? 17:29:41 don't think so 17:29:50 the stuff they poll off of libvirt is what you need 17:30:32 zykes-: gotcha. 17:30:45 Armando told me there was a XenAPI RRD stuff, but I can't see it in the code now I am looking myself 17:30:51 can probably model it the way we do bw. 17:31:03 johngarbutt: we had a lot of issues with the rrd code for polling bw 17:31:09 BobBall can attest 17:31:31 thanks for the heads up! I think bobball mentioned that 17:31:36 how well does SAN disks fare with OpenStack and Xen ? 17:31:59 I'm aware of a polling issue for net devices in XS6.0 but I believe they are resolved in XS6.1 17:32:00 if you do iscsi you are fine, roughly 17:32:14 johngarbutt: more like fc :) 17:32:23 We are in the process of adding NFS 17:32:54 no idea about fc I am afraid, XenServer supports that stuff, but not the OpenStack+XenServer combo at the moment 17:33:04 depends exactly what you want to do with your SAN 17:33:24 what is it that decides that johngarbutt ? 17:33:27 a xenapi layer ? 17:34:12 well I was more meaning if you want to store all VM disks on the SAN, or if you want to let OpenStack manage your SAN like Volumes that get attached 17:34:30 johngarbutt: both :) 17:34:38 OK... 17:34:59 do you want to do the first one using boot from volume, that is the best way I guess 17:35:15 the issue is getting your SAN vendor supported inside Cinder, does it work with KVM today? 17:35:58 the second issue is making that support work with how XenAPI models storage 17:35:59 johngarbutt: HP is working with it :) 17:36:14 and some others (brocade + emc i believe as well) 17:36:30 interesting, I would have to look at that, last time I looked they were all iSCSI based 17:36:31 it should be a pretty HOT topic i think since alot of enterprises already have fiberchannel 17:36:34 :p 17:37:12 its also a point where they can look at cheaper soutions, but lets get back on topic and make sure we talk more later, sounds interesting 17:37:20 any more on docs stuff? 17:37:29 did we spot anything missing on that matrix 17:37:42 johngarbutt: I don't think so, atm 17:37:47 #action look at fiberchannel support in cinder, see if it needs to go on matrix 17:37:51 cool 17:38:02 lets move on QA 17:38:07 #topic QA 17:38:23 tomorrow is bug squash day, any big ones people are worries about 17:38:43 seems like a good chance to triage the Xen bugs, kick out some old ones, and see if there are some easy things we can fix 17:38:59 Fraid I have to jump now - will catch up on the minutes tomorrow. 17:39:10 OK 17:39:32 johngarbutt: will review that list and see if we can throw some folks into that tomorrow. 17:39:42 can't commit yet though 17:39:42 ofc you can buy cheaper stuff, but if you have existing gear already in place then you get the idea..... you can simply use what you have! 17:40:01 johngarbutt: https://bugs.launchpad.net/nova/+bug/1061045 is pretty bad for us. 17:40:03 Launchpad bug 1061045 in nova "Failure to set root password leaves instance in ERROR" [High,Confirmed] 17:40:05 just for giggles, I am sat in the Cambridge office wearing black tie and dinner suit ready for the christmas party later! 17:40:35 ouch 17:41:06 there was some talk about this, we need to be able to report some errors but not put the instance into the Errored state 17:41:28 yea, I think that was the gist of what Johannes mentioned in the bug 17:41:59 ah yes, I see his point, we could avoid the need for the error state 17:42:18 i don't understand the API 500 side of that 17:42:22 because the API doesn't wait 17:43:11 oh, I guess nova-compute takes so long to return the manager recons the compute is dead 17:43:20 maybe it should be doing a cast not a call in the compute api 17:43:25 would have to check that 17:43:56 set_admin_password is a call 17:43:59 building is a cast 17:44:03 that would do it 17:44:15 the bug report is not clear 17:44:22 it must be referring to set_admin_password 17:44:25 not building a new instance 17:44:40 I think so, we should add a node about that 17:45:11 OK, so in other QA news... 17:45:29 we are rebuilding our internal QA stuff, just running devstack and tempest 17:45:43 and using it to test everything we have added recently 17:45:54 NFS, boot from volume, live migration, etc 17:46:09 it would be ace to have more public tests 17:46:16 in the same vain as smokestack 17:46:34 i.e. hooking into gerrit but external 17:47:02 or maybe we could try get the CI team to get the existing system to work with XS too? 17:47:12 anyone got time to work on that at the moment? 17:47:23 does it seem important to you too? 17:47:40 johngarbutt: we're look at some of those. 17:47:57 our CI team is interfacing with OS CI team to get some of our tests integrated, like SmokeStack 17:48:06 I should put you in touch with those guys on our side. 17:48:11 awesome, that would be good 17:48:36 we are already running devstack with XenServer quite frequently, and had a few ideas to improve things 17:48:48 #action pvo to get john in touch with rax CI team 17:48:52 with XS 6.1 we should be able to run XenServer in a VM, which could help 17:49:02 johngarbutt: ya, looking forward to that. 17:49:21 OK moving on to next part 17:49:29 (we covered bugs... oops) 17:49:33 #topic AOB 17:50:01 are people good with this meeting time? 17:50:10 is weekly good for people? 17:50:29 ya, works for me 17:50:36 wiki says its every 2nd Wed of the month 17:50:41 will try keep it shorter 17:50:48 yes, I was thinking move to weekly 17:50:51 so we want to move? 17:50:51 ok 17:50:56 works for me 17:51:28 that we I am less likely to forget, and we can keep them short at meaningful, hopefully 17:51:45 awesome, afraid I should run off now 17:51:54 sounds good. 17:51:54 thanks for coming along 17:51:58 thanks john 17:52:01 np 17:52:04 #endmeeting