| *** mgius has quit IRC | 00:09 | |
| vishy | ha flatdhcp appears to work | 00:27 |
|---|---|---|
| vishy | :) | 00:27 |
| openstackjenkins | Project nova build #1,069: SUCCESS in 3 min 11 sec: http://jenkins.openstack.org/job/nova/1069/ | 00:29 |
| openstackjenkins | Tarmac: Silence warning in case tests.sqlite doesn't exist | 00:29 |
| kpepple | vishy: woot | 00:30 |
| vishy | needs some testing | 00:34 |
| vishy | kpepple: I haven't tried it on multiple boxes yet so there may be bugs | 00:34 |
| vishy | kpepple: lp:~vishvananda/nova/ha-net | 00:36 |
| kpepple | vishy: not a worry ... the network availability / scalability problem is holding a lot of people back. if this fixes at least one of them, thats a win. | 00:36 |
| kpepple | vishy: will try ... | 00:36 |
| vishy | kpepple: there is a new flag on networks | 00:36 |
| vishy | multi_host | 00:36 |
| vishy | you can set during create with nova-manage create private 10.0.0.0/8 1 32 T | 00:37 |
| vishy | or default it with a flag | 00:37 |
| vishy | --multi_host=True | 00:37 |
| vishy | but if you do you have to run nova-network on all of the compute hosts | 00:37 |
| vishy | you can just use one per compute host, no separate network node | 00:38 |
| vishy | anyway I'm off, let me know if you try it out | 00:39 |
| kpepple | vishy: thx, will so | 00:39 |
| vishy | (btw it will give a different ip to each compute host) | 00:39 |
| *** vladimir3p_ has quit IRC | 00:47 | |
| *** stevezd has quit IRC | 00:49 | |
| *** mszilagyi has quit IRC | 00:49 | |
| *** jdurgin has quit IRC | 01:00 | |
| *** HugoKuo has quit IRC | 01:00 | |
| *** HugoKuo has joined #openstack-dev | 01:07 | |
| *** ohnoimdead has quit IRC | 01:07 | |
| *** lorin1 has quit IRC | 01:13 | |
| *** lorin1 has joined #openstack-dev | 01:14 | |
| *** johnpur has quit IRC | 02:50 | |
| *** openpercept has joined #openstack-dev | 03:18 | |
| *** mdomsch has joined #openstack-dev | 03:20 | |
| *** chomping has joined #openstack-dev | 03:30 | |
| *** chomping has quit IRC | 03:35 | |
| *** chomping has joined #openstack-dev | 03:35 | |
| *** chomping has quit IRC | 03:36 | |
| *** chomping_ has joined #openstack-dev | 03:36 | |
| *** chomping_ has joined #openstack-dev | 03:36 | |
| *** lorin1 has left #openstack-dev | 03:39 | |
| *** mszilagyi has joined #openstack-dev | 03:52 | |
| *** negronjl_ has joined #openstack-dev | 04:13 | |
| *** negronjl has quit IRC | 04:15 | |
| *** mszilagyi has quit IRC | 04:50 | |
| *** vladimir3p_ has joined #openstack-dev | 05:07 | |
| *** vladimir3p_ has quit IRC | 05:31 | |
| *** zaitcev has quit IRC | 05:59 | |
| *** po has joined #openstack-dev | 06:40 | |
| *** dragondm has joined #openstack-dev | 06:45 | |
| *** reidrac has joined #openstack-dev | 07:02 | |
| *** reidrac has left #openstack-dev | 07:03 | |
| *** reidrac has joined #openstack-dev | 07:03 | |
| *** reidrac has left #openstack-dev | 07:03 | |
| *** reidrac has joined #openstack-dev | 07:04 | |
| *** BK_man has quit IRC | 07:13 | |
| *** BK_man_ is now known as BK_man | 07:13 | |
| *** mancdaz1203 has quit IRC | 07:32 | |
| *** mancdaz has joined #openstack-dev | 07:33 | |
| *** mancdaz has quit IRC | 08:04 | |
| *** mancdaz has joined #openstack-dev | 08:05 | |
| *** dragondm has quit IRC | 08:19 | |
| ttx | soren: so, feedback from D2 release automation last week | 08:27 |
| ttx | soren: reupload_with_new_version.py is a bit brittle | 08:28 |
| soren | ttx: Yes! | 08:28 |
| ttx | like uscan seems to randomly fail | 08:28 |
| soren | Fascinating. On reliably networks? | 08:28 |
| soren | *reliable | 08:28 |
| ttx | on slow but reliable networks | 08:28 |
| soren | a.k.a. rural France ADSL? :) | 08:28 |
| ttx | randomly being FAIl PASS FAIL PASS PASS FAIL | 08:28 |
| soren | Ok. | 08:29 |
| ttx | like on the same tarball | 08:29 |
| soren | That particluar bit has lots of room for improvement for sure. | 08:29 |
| ttx | I thought it was "time for it to show up" but it sounds a bit more tricky | 08:29 |
| soren | Ah. | 08:29 |
| *** Vasichkin has joined #openstack-dev | 08:30 | |
| ttx | maybe some timeout | 08:30 |
| soren | I kind of like that it pulls it frmo Launchpad. That way, we're sure we don't cheat and use a different tarball. | 08:30 |
| ttx | tarball upload is a bit slow with my upload speed, and it does it 4 times | 08:30 |
| soren | Well, that's an artefact of this process being new. | 08:31 |
| soren | AS soon as we're confident it's in good shape, it needs to happen on the Jenkins box. | 08:31 |
| ttx | soren: hmm, with GPG keys being stored there ? | 08:32 |
| soren | That's already the case. | 08:32 |
| soren | We upload to the trunk ppa's from there. | 08:32 |
| soren | And the milestone-proposed ppa's. | 08:32 |
| ttx | ok | 08:32 |
| ttx | is there a way for reupload_with_new_version.py to only upload the tarball to the PPA once ? | 08:33 |
| ttx | it really needs to happen one time for each series ? | 08:33 |
| soren | Strictly speaking, no. | 08:34 |
| ttx | the whole "reupload_with_new_version.py nova-core/milestone-proposed nova 2011.3~d2" took 30 min. | 08:34 |
| soren | THere's a catch, though. | 08:34 |
| ttx | it would obviously go faster from Jenkins than from my home, but... | 08:34 |
| soren | If we're only to upload it once, we need to upload the first one, wait for its source to be published and then upload the rest (so that they can find the orig.tar.gz from the first upload). | 08:35 |
| soren | AT least I think that's the case. | 08:35 |
| soren | Let me ask around a bit. | 08:35 |
| ttx | hmm, might just be faster to reupload them from a fast location | 08:35 |
| ttx | soren: it's probably better for the script to reupload them, rather than poll for source code | 08:36 |
| soren | ttx: Yeah. | 08:38 |
| ttx | soren: ftr here is what I did: http://paste.openstack.org/show/1844/ | 08:40 |
| ttx | soren: I'm working on automating the middle part now | 08:41 |
| soren | wait! | 08:41 |
| soren | :) | 08:41 |
| soren | ttx: Before you do, take a look at lp-project-upload in ubuntu-dev-tools. | 08:42 |
| soren | ttx: I've never used it, but it sounds like just what we need. | 08:42 |
| * ttx took a look at TheKraken, the release automation tool from Aaron Bentley | 08:42 | |
| ttx | I'll compare now :) | 08:42 |
| ttx | soren: how do you want to handle the "# wait for stuff to build and be published" wait ? Poll for package publication ? Or have a two-step process with human verification in between ? | 08:44 |
| ttx | everything else can actually be done in one shot... especially if it's simple to trigger parameterized Jenkins jobs from other jobs | 08:44 |
| soren | ttx: Just checked with bigjools. No need to wait for the first package to get published. We can just include the tarball in the first upload and skip it on the rest. | 08:47 |
| soren | ttx: I think making the "wait for stuff" a manual step is the right choice. | 08:49 |
| soren | ttx: And not just because a technical implementation will be quite awkward. | 08:49 |
| soren | ttx: ...but that too :) | 08:49 |
| ttx | right | 08:49 |
| soren | ttx: I think it's a good idea to have a human actually see that it went as expected. | 08:50 |
| soren | ttx: You know.. | 08:50 |
| soren | ttx: I don't like that "reupload with new version" thing. | 08:50 |
| soren | ttx: Not just the implementation, but the idea. | 08:50 |
| ttx | so lp-project-upload is quite close, except that you can't set a description | 08:50 |
| soren | It's wrong for all the reasons I stated in my EP2011 talk. | 08:51 |
| ttx | and my script would also take care of downloading and verifying the tarball... the actual "create-release" part is 5 lines of code so simple there is no good reason to reuse it | 08:52 |
| ttx | hmm | 08:52 |
| ttx | soren: but how would you avoid that ? | 08:52 |
| soren | No clue. | 08:53 |
| soren | oh, brb. | 08:53 |
| *** chemikadze has quit IRC | 08:55 | |
| *** chemikadze has joined #openstack-dev | 08:57 | |
| soren | back | 08:58 |
| soren | ttx: Yeah, no clue. | 08:59 |
| soren | ttx: Admittedly, a lot of the stuff we're doing is making sure that this will work as expected, but still. | 09:00 |
| *** mancdaz has quit IRC | 09:18 | |
| *** mancdaz has joined #openstack-dev | 09:19 | |
| *** chemikadze has quit IRC | 10:22 | |
| *** chemikadze has joined #openstack-dev | 10:28 | |
| *** chomping_ has quit IRC | 10:37 | |
| *** chomping has joined #openstack-dev | 11:04 | |
| *** chomping has quit IRC | 11:09 | |
| *** markvoelker has joined #openstack-dev | 12:02 | |
| *** HugoKuo_ has joined #openstack-dev | 12:44 | |
| *** HugoKuo has quit IRC | 12:48 | |
| *** mdomsch has quit IRC | 13:06 | |
| *** jaypipes has joined #openstack-dev | 13:13 | |
| *** kbringard has joined #openstack-dev | 13:24 | |
| *** kbringard has quit IRC | 13:26 | |
| *** kbringard has joined #openstack-dev | 13:27 | |
| *** kbringard has quit IRC | 13:41 | |
| *** kbringard has joined #openstack-dev | 13:41 | |
| *** openpercept has quit IRC | 13:48 | |
| *** vladimir3p_ has joined #openstack-dev | 13:49 | |
| *** openpercept_ has joined #openstack-dev | 13:50 | |
| *** openpercept_ has quit IRC | 13:50 | |
| *** vladimir3p_ has quit IRC | 13:51 | |
| *** dprince has joined #openstack-dev | 14:05 | |
| *** cweidenk1ller is now known as cweidnekeller | 14:15 | |
| *** cweidnekeller is now known as cweidenkeller | 14:16 | |
| *** cweidenkeller has quit IRC | 14:18 | |
| *** jkoelker has joined #openstack-dev | 14:19 | |
| *** cweidenkeller has joined #openstack-dev | 14:20 | |
| *** bcwaldon has joined #openstack-dev | 14:28 | |
| *** cp16net_ has joined #openstack-dev | 14:31 | |
| *** cweidenkeller has quit IRC | 14:39 | |
| *** cweidenkeller has joined #openstack-dev | 14:39 | |
| *** vladimir3p has joined #openstack-dev | 14:40 | |
| po | hi ! can someone explain me why vm need to contact 169.254.169.254 when launched ? | 14:53 |
| ttx | po: that's how the EC2 metadata server works | 14:54 |
| kbringard | that's where it gets stuff like, the ssh keys and whatnot | 14:54 |
| kbringard | hah, ttx beat me to it | 14:54 |
| po | ttx: 169.254.169.254 is nated to the controler ip, isn't it ? | 14:54 |
| ttx | po: so a vm that is designed to run on EC2-like envs (like ubuntu cloud images) will try to get data from 169.254.169.254 | 14:54 |
| ttx | it's nated to the nova-api, yes | 14:54 |
| ttx | though we might simplify in the future so that it actually points to the host. | 14:55 |
| po | can't it be disabled ? | 14:55 |
| po | i mean, if i pack my own vm image | 14:56 |
| ttx | po: of course it can. | 14:56 |
| ttx | po: hrm. | 14:56 |
| ttx | po: depend on what you mean by "it" I guess | 14:56 |
| ttx | the calls to 169.254.169.254 can definitely not be sent from your VM and the thing will continue to work | 14:57 |
| ttx | unless your VM actually depends on that data (like the Ubuntu cloud images do) | 14:57 |
| po | ok, that's why my ubuntu server image can't boot | 14:58 |
| *** dragondm has joined #openstack-dev | 15:03 | |
| *** reidrac has quit IRC | 15:12 | |
| tr3buchet | vishy: I was out yesterday. What did you need? | 15:29 |
| *** mancdaz has quit IRC | 15:31 | |
| kbringard | interesting, it looks like cloud-config supports using multiple metadata urls (back to po's question) | 15:32 |
| *** mancdaz has joined #openstack-dev | 15:35 | |
| mtaylor | ttx, soren: I looked at TheKraken a while back - but unfortunatley it's designed to make a release from a branch - and not using setup.py sdist, IIRC, which made me want to kill someone | 15:56 |
| ttx | mtaylor: yes, I just ripped from it | 15:57 |
| *** KAM has joined #openstack-dev | 15:59 | |
| *** cp16net_ has quit IRC | 16:09 | |
| *** bcwaldon has quit IRC | 16:14 | |
| *** mdomsch has joined #openstack-dev | 16:16 | |
| *** peads has joined #openstack-dev | 16:20 | |
| *** cp16net_ has joined #openstack-dev | 16:24 | |
| *** cp16net_ has quit IRC | 16:27 | |
| vishy | tr3buchet: I was trying to understand a few things in the networking refactor | 16:36 |
| vishy | i think i figured it out | 16:36 |
| vishy | tr3buchet: I'm not convinced by the use of super in the flatdhcp mixin | 16:37 |
| vishy | * FloatingIp mixin | 16:37 |
| vishy | also the check on auto_assign_floating_ip was a little messed up | 16:37 |
| vishy | tr3buchet I have a branch up for ha-networking lp:~vishvananda/nova/ha-net | 16:39 |
| vishy | testing it today on multiple machines | 16:39 |
| tr3buchet | vishy: hmm | 16:39 |
| tr3buchet | how are you not convinced by the use of super? | 16:40 |
| *** jdurgin has joined #openstack-dev | 16:40 | |
| *** KAM has left #openstack-dev | 16:41 | |
| *** zaitcev has joined #openstack-dev | 16:44 | |
| *** zul has quit IRC | 16:52 | |
| *** bcwaldon has joined #openstack-dev | 16:55 | |
| *** lorin1 has joined #openstack-dev | 16:57 | |
| vishy | the super call in FloatingIp mixin depends on inheritance structure and ordering | 16:57 |
| vishy | which i think is very fragile | 16:57 |
| *** mdomsch has quit IRC | 17:02 | |
| comstud | the mro is kinda interesting | 17:07 |
| *** ohnoimdead has joined #openstack-dev | 17:08 | |
| comstud | we had played around double inheritance in standalone test programs to verify it | 17:08 |
| comstud | although I'm not familiar with the code that actually landed | 17:08 |
| *** deshantm_laptop has joined #openstack-dev | 17:09 | |
| *** mdomsch has joined #openstack-dev | 17:12 | |
| *** erik-s has quit IRC | 17:13 | |
| comstud | taking a look, the super() in FloatingIp works fine as long as you've got a class that inherits from FloatingIp along with another class... and FloatingIp is first in the inheritance list. | 17:13 |
| *** erik-s has joined #openstack-dev | 17:14 | |
| comstud | I guess I should say: s/first/not last/ | 17:15 |
| comstud | i can't argue about it being 'fragile', tho :) | 17:17 |
| *** CatKiller has quit IRC | 17:17 | |
| *** AhmedSoliman has joined #openstack-dev | 17:20 | |
| *** po has quit IRC | 17:21 | |
| *** AhmedSoliman has quit IRC | 17:22 | |
| *** AhmedSoliman has joined #openstack-dev | 17:23 | |
| *** CatKiller has joined #openstack-dev | 17:23 | |
| *** jaypipes has quit IRC | 17:30 | |
| *** jaypipes has joined #openstack-dev | 17:31 | |
| *** CatKiller has quit IRC | 17:37 | |
| *** CatKiller has joined #openstack-dev | 17:49 | |
| *** CatKiller has quit IRC | 17:50 | |
| *** CatKiller has joined #openstack-dev | 17:51 | |
| *** negronjl_ has quit IRC | 17:54 | |
| vishy | jaypipes: IOError: [Errno 13] Permission denied: '/var/log/glance/api.log' | 17:58 |
| vishy | in apt-get. Perhaps this is the error that chuck reported | 17:59 |
| jaypipes | vishy: https://bugs.launchpad.net/glance/+bug/784837 | 17:59 |
| uvirtbot | Launchpad bug 784837 in glance "glance-api.log permission error when using upstart script" [High,Fix released] | 17:59 |
| vishy | jaypipes: hmm, i wonder if we need matching fixes in nova | 18:00 |
| jaypipes | vishy: likely. zul will be back on channel later. he's taking care of his kid right now... | 18:01 |
| jaypipes | vishy: let's hit him up on this. | 18:01 |
| *** mgius has joined #openstack-dev | 18:03 | |
| *** mdomsch has quit IRC | 18:07 | |
| *** AhmedSoliman has quit IRC | 18:16 | |
| heckj | tr3buchet - got a sec? Ran into an issue with multi-nic in trunk | 18:57 |
| vishy | heckj: what is it? | 19:04 |
| vishy | I've been fooling | 19:04 |
| vishy | \me cheers! | 19:04 |
| heckj | heh | 19:04 |
| * vishy cheers! | 19:04 | |
| vishy | that is | 19:04 |
| vishy | ha-networking works | 19:04 |
| heckj | Ran into an issue with positional arguments while invoking "network create" with nova-manage | 19:05 |
| vishy | yeah they changed | 19:05 |
| vishy | you need to add a private | 19:05 |
| vishy | or 'label' | 19:05 |
| heckj | Should we be adding a label in there to work around? | 19:05 |
| vishy | yes | 19:05 |
| heckj | so adding "label" up front on that method was intentional. Ok - wasn't 100% sure | 19:05 |
| heckj | What's the implication of a label? If I just make it something stupid for an all-in-one setup, will that bite me later? | 19:06 |
| heckj | (already sent a note to the openstack list - so you'll see this question there too...() | 19:06 |
| vishy | doesn't matter | 19:10 |
| vishy | the label is just so you can tell the different networks apart | 19:11 |
| vishy | it will bridge the vm into all the networks you have defined | 19:11 |
| heckj | vishy: thanks - got it. Made a tweak to my copy of nova.sh and it's installing cleanly now | 19:11 |
| vishy | yeah i just updated nova.sh btw | 19:11 |
| vishy | (well the novascript one) | 19:12 |
| vishy | the cloudbuilders one is unmodified | 19:12 |
| * heckj looks | 19:12 | |
| heckj | I've been riffing off the cloudbuilders on | 19:12 |
| heckj | one | 19:12 |
| *** scotticus has left #openstack-dev | 19:16 | |
| tr3buchet | heckj: yes | 19:17 |
| tr3buchet | looks like vish set you straight | 19:18 |
| heckj | tr3buchet: vishy answered my Q - didn't know about the network create command change needed | 19:18 |
| heckj | Yep! | 19:18 |
| *** mrmartin has joined #openstack-dev | 19:18 | |
| tr3buchet | good deal | 19:18 |
| vladimir3p | vishy: is there any quick doc/spec on ha-networking? | 19:44 |
| vishy | vladimir3p: it is actually pretty simple | 19:44 |
| vishy | vladimir3p: but it won't work on windows yet | 19:44 |
| vishy | vladimir3p: there is an extra flag on networks called multi_host | 19:45 |
| vishy | you set it to true and then you run a nova-network on every compute host | 19:45 |
| vishy | all of the compute hosts will get an ip on the network | 19:45 |
| vladimir3p | vishy: will you assign IPs from fixed range to nodes similarly to how it is done for single networking node? | 19:47 |
| vishy | yes | 19:48 |
| vishy | everything else works the same | 19:48 |
| vladimir3p | vishy: today config of br100 on CloudCtr/networking is slightly different from br100 on other nodes | 19:48 |
| vladimir3p | ah, ok | 19:48 |
| vishy | all computes gateway for their own vms | 19:48 |
| vishy | that way there is no spof for network | 19:48 |
| vishy | in fact, in my current testing I'm running all services on all hosts | 19:49 |
| vladimir3p | vishy: got you. in this case it is not really HA, but more like a multi-node network | 19:49 |
| vishy | Well it is HA for the cluster | 19:50 |
| vladimir3p | vishy: we can probably call it HA, because failure on single node will not affect others (from netw perspective) | 19:50 |
| heckj | vishy: is that in trunk, or in a branch you're working on? Went to look for multi_host flag and didn't find it. | 19:50 |
| vishy | heckj: branch | 19:50 |
| vishy | proposing it for review | 19:50 |
| heckj | Ah, grazie | 19:50 |
| vishy | https://code.launchpad.net/~vishvananda/nova/ha-net/+merge/67078 | 19:52 |
| vishy | i think with a little tweaking to dnsmasq, you can also make all of the vms use a hardware gateway for fixed ips | 19:52 |
| vishy | it means floating ips and fixed ips will have asymmetric routes to the vms, but I think it will work. Testing that on a real cluster sometime soon | 19:53 |
| *** jtran has joined #openstack-dev | 20:06 | |
| *** dprince has quit IRC | 20:26 | |
| *** jtran has quit IRC | 20:28 | |
| *** winston-d has quit IRC | 20:37 | |
| *** lorin1 has quit IRC | 21:04 | |
| *** mrmartin has quit IRC | 21:18 | |
| *** kbringard has quit IRC | 21:35 | |
| *** bcwaldon has quit IRC | 21:45 | |
| *** bcwaldon has joined #openstack-dev | 21:47 | |
| *** bcwaldon has quit IRC | 21:52 | |
| heckj | vishy: I'm being dense. Where is the code assigning an IP address per network host into the database for each network? | 22:00 |
| heckj | vishy: nevermind. Just had to ask the question to see it | 22:01 |
| *** nati has joined #openstack-dev | 22:28 | |
| jaypipes | mtaylor: ping | 22:28 |
| devcamcar | can anyone provide details about the current level of keystone integration into nova? | 22:30 |
| *** markvoelker has quit IRC | 22:32 | |
| vishy | heckj: you got it? | 22:33 |
| *** jaypipes has quit IRC | 22:33 | |
| vishy | devcamcar: the only integration is the middleware at the moment | 22:33 |
| vishy | which just automatically creates keystone users in nova | 22:33 |
| devcamcar | vishy: i'm basically running trunk right now just trying to get up to speed on this stuff | 22:33 |
| devcamcar | vishy: is it even possible to add projects and users with nova-manage right now? | 22:34 |
| *** jkoelker has quit IRC | 22:34 | |
| vishy | no | 22:36 |
| vishy | well yes if you don't use keystone | 22:36 |
| vishy | with keystone you add them via keystone | 22:36 |
| vishy | then the middleware automatically creates them in nova | 22:36 |
| devcamcar | vishy: using your latest nova.sh, what does this mean: echo 'no openstack cli automation yet' | 22:37 |
| vishy | uh... | 22:37 |
| vishy | you meant he cloudbuilders one? | 22:37 |
| heckj | vishy: yeah, thanks - just took soem time to read through your ha-net branch and figured out what all it was doing. Got lost a few times on the way | 22:37 |
| devcamcar | vishy: yea | 22:37 |
| vishy | devcamcar: I think it is because we removed the rc exporting and stuff | 22:38 |
| vishy | heckj: could probably use a little cleanup and refactoring. I was just trying to make it work first of all | 22:39 |
| heckj | vishy: more like I'm just not familiar with all the twisty passages in that code | 22:39 |
| vishy | heckj: this is one of the issues with adding new features but trying not to break the existing usage. | 22:40 |
| vishy | heckj: it started out pretty clean and simple (well as clean as simple as networking can be, which is pretty dirty and complicated :) ) | 22:40 |
| *** jaypipes has joined #openstack-dev | 22:45 | |
| *** AhmedSoliman has joined #openstack-dev | 22:47 | |
| *** jaypipes has quit IRC | 22:51 | |
| *** lorin1 has joined #openstack-dev | 22:52 | |
| *** nati has quit IRC | 22:58 | |
| *** nati has joined #openstack-dev | 22:58 | |
| *** jaypipes has joined #openstack-dev | 23:01 | |
| *** Vasichkin has quit IRC | 23:19 | |
| *** AhmedSoliman has quit IRC | 23:21 | |
| *** nati has quit IRC | 23:24 | |
| *** deshantm_laptop has quit IRC | 23:26 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!