Tuesday, 2010-09-28

*** pharkmillups has quit IRC00:01
*** silassewell has joined #openstack00:09
*** _anm has joined #openstack00:16
*** rnirmal has joined #openstack00:34
*** dendro-afk is now known as dendrobates00:36
*** joearnold has quit IRC00:38
*** jc_smith has quit IRC00:42
*** joearnold has joined #openstack00:44
*** sophiap has quit IRC00:53
*** miclorb has quit IRC00:59
*** Cybodog has joined #openstack01:02
*** rnirmal has quit IRC01:05
*** littleidea has quit IRC01:08
*** sophiap has joined #openstack01:09
*** sophiap_ has joined #openstack01:11
*** sophiap has quit IRC01:11
*** sophiap_ is now known as sophiap01:11
*** maplebed has quit IRC01:14
*** ooooPsss has quit IRC01:15
*** jessec has joined #openstack01:23
*** sophiap has quit IRC01:29
*** miclorb_ has joined #openstack01:32
*** hisaki has joined #openstack01:39
*** anm` has joined #openstack01:40
*** _anm has quit IRC01:41
*** anm` is now known as _anm01:41
*** hisaki has quit IRC01:44
*** Pentheus has quit IRC01:50
*** miclorb_ has quit IRC01:51
*** abecc has quit IRC01:54
*** miclorb has joined #openstack01:55
*** rnirmal has joined #openstack02:04
*** jessec has quit IRC02:07
*** sirp1 has joined #openstack02:30
*** Cybodog has quit IRC02:49
*** gundlach has quit IRC02:52
*** abecc has joined #openstack02:54
*** arcane has quit IRC02:59
*** arcane has joined #openstack03:23
*** jakedahn has joined #openstack03:31
*** mtaylor has joined #openstack03:32
*** ChanServ sets mode: +v mtaylor03:32
*** abecc has quit IRC03:37
*** sagi_ has joined #openstack03:44
*** kashyapc has joined #openstack03:46
*** Cybodog has joined #openstack03:53
*** kashyapc has quit IRC04:09
*** rnirmal has quit IRC04:19
*** kashyapc has joined #openstack04:26
*** hornbeck has quit IRC04:26
*** littleidea has joined #openstack04:35
*** jakedahn has quit IRC04:44
*** jakedahn has joined #openstack04:44
*** omidhdl has joined #openstack04:44
*** miclorb has quit IRC04:51
*** littleidea has quit IRC04:51
*** omidhdl has quit IRC04:51
*** littleidea has joined #openstack04:52
*** silassewell has quit IRC04:56
*** omidhdl has joined #openstack04:59
*** f4m8_ is now known as f4m805:00
*** cloud0 has joined #openstack05:01
*** hornbeck has joined #openstack05:26
*** joearnold has joined #openstack05:29
*** ibarrera has joined #openstack06:07
*** joearnold has quit IRC06:15
*** _anm has quit IRC06:19
*** allsystemsarego has joined #openstack06:19
*** guigui1 has joined #openstack06:21
*** jakedahn has quit IRC06:21
*** jakedahn has joined #openstack06:21
*** ArdRigh has joined #openstack06:46
*** stewart_ has joined #openstack06:57
*** stewart has quit IRC06:58
*** stewart_ is now known as stewart06:58
*** gaveen has joined #openstack07:00
*** brd_from_italy has joined #openstack07:01
*** whosjose has joined #openstack07:10
*** whosjose has left #openstack07:11
*** guigui1 has quit IRC07:36
*** anticw has quit IRC07:41
*** zheng_li has quit IRC07:51
*** gasbakid has joined #openstack07:57
*** guigui1 has joined #openstack08:05
*** zheng_li has joined #openstack08:15
sorenvishy, devcamcar: I asked a question on https://code.edge.launchpad.net/~devcamcar/nova/fix-iptables/+merge/36110 that hasn't been answered yet. Can you provide this feedback so that we can land this branch, please?08:51
*** omidhdl has left #openstack08:56
*** omidhdl has joined #openstack09:14
*** littleidea has quit IRC09:19
*** miclorb_ has joined #openstack09:43
*** rsoutar has joined #openstack09:54
*** ptremblett has joined #openstack10:04
*** rsoutar_ has joined #openstack10:12
*** rsoutar has quit IRC10:14
*** rsoutar_ is now known as rsoutar10:14
*** ArdRigh has quit IRC10:19
*** rsoutar_ has joined #openstack10:21
*** rsoutar has quit IRC10:23
*** rsoutar_ is now known as rsoutar10:23
*** omidhdl has left #openstack10:28
*** ctennis has quit IRC10:51
chmouelhey soren I am trying to checkout the ubuntu branch of nova but i am getting permission denied when doing a bzr branch lp:nova/ubuntu10:55
*** ooooPsss has joined #openstack10:55
*** brd_from_italy has quit IRC10:56
chmouelsoren: nevermind was my ssh agent stucked...11:00
*** gaveen has quit IRC11:00
*** ooooPsss has quit IRC11:01
chmoueland from which place nova-scheduler is supposed to come ?11:05
*** ctennis has joined #openstack11:07
*** guigui1 has quit IRC11:08
*** befreax has joined #openstack11:08
*** sophiap has joined #openstack11:11
*** brd_from_italy has joined #openstack11:12
*** miclor___ has joined #openstack11:12
*** miclorb_ has quit IRC11:12
*** gaveen has joined #openstack11:14
*** sophiap has quit IRC11:15
*** sophiap has joined #openstack11:16
sorenchmouel: It's on the trunk.11:35
chmoueli am trying to figure out how to merge trunk in ubuntu branch (bzr noob) like a git rebase11:36
sorenchmouel: Sorry about not posting instructions yet. I hit a bunch of blockers, fixed them, and then I got sidetracked on feature work.11:36
sorenDon't rebase.11:36
sorenJust grab trunk, and merge the ubuntu branch into it.11:36
chmouellike bzr merge lp:nova/ubuntu  in trunk dir?11:37
chmouelsoren: no worries I am trying to figure out stuff by myself and will post my notes of it somewhere...11:37
* soren will not be joining eventlet's fan club anytime soon.11:38
*** gaveen has quit IRC11:39
chmouelso now after bzr merge stuff bzr status show me bunch of new files modified etc... how do i get a clean branch back after I built package11:40
chmouelie: git reset --hard11:40
sorenbzr clean-tree11:54
sorenwhen you do a bzr merge, you need to commit the merge.11:55
sorenI believe this is the same as in git.11:55
sorenchmouel: Until you commit it, they'll appear as modified files (because, y'know... They are).11:56
chmouelah yeah it is but I wanted something like a rebase of the commit from ubuntu branch and not having to merge a new commit... well i guess i'll workout a workflow...11:56
chmoueli had to do something like that btw: on setup.py to build the package:11:57
chmouel-               'bin/nova-api-new',11:57
chmouel+               'bin/nova-scheduler',11:57
sorenWhy your reluctance to commit stuff?11:58
*** rsoutar has quit IRC11:58
chmoueli am worry to screw up stuff to be honest with not masterizing the SCM11:58
sorenThis is a DVCS. You're just comitting stuff locally. You can't break anything :)11:58
chmouelah ah yeah but if i want to propose the change back I may add all that ubuntu branch commit directly11:59
sorenIf you want to change stuff on the trunk, branch it and make your changes there.12:00
chmouelyep will get there ...12:01
*** abien has left #openstack12:01
*** ctennis has quit IRC12:04
sorenchmouel: Hm... I actually have a branch with fixes for that setup.py stuff.12:04
chmouelcool will wait for it...12:04
sorenIt's identical to your patch :)12:05
sorenNot sure why I never pushed it.12:06
chmoueli just installed  those built packages from trunk/ubuntu branch and I am still getting a 644 perm on /etc/sudoers.d/nova_sudoers which crash sudo12:06
sorenchmouel: Hm.. Yeah, you're right.12:07
chmouelsoren: you got that fix in your branch as well?12:11
sorenchmouel: Nope, working on it.12:12
chmoueli can help on that if you like12:12
sorenI'm done :)12:13
chmouelso well I was doing all of that so I can get the conf file samples from the packaging branch, any reason why those are not in trunk ?12:14
sorenNot particularly, no.12:15
chmouelwell i guess when the doc will be updated it will be easier for the one who does not know the syntax12:16
*** brd_from_italy has quit IRC12:22
*** gundlach has joined #openstack12:29
*** sophiap has quit IRC12:33
*** sophiap has joined #openstack12:33
*** kashyapc has quit IRC12:45
*** Podilarius has joined #openstack12:50
*** klord has joined #openstack12:51
*** _anm has joined #openstack12:54
*** befreax has quit IRC12:57
*** sophiap has quit IRC13:00
*** littleidea has joined #openstack13:04
*** miclor___ has quit IRC13:06
*** abecc has joined #openstack13:06
*** sophiap has joined #openstack13:11
chmouelsoren: I am trying to generate a certificate to use with eucaools so I have created a project and user admin and I am doing a nova-manage project zipfile admin admin13:11
chmouelwhich fail because it can't find the openssl.cnf13:11
chmouelI can see the template from the package in /var/lib/nova/CA/openssl.cnf.tmpl do you know how would i use that for generate that zipfile?13:12
*** guigui has joined #openstack13:16
*** jessec has joined #openstack13:18
chmouelor everything works using the environement stuff now and no need ssl thingies?13:20
*** devcamcar has quit IRC13:21
*** ChrisAM has quit IRC13:21
*** devcamcar has joined #openstack13:22
uvirtbotNew bug: #649806 in nova "AMQP concurrencty problem in API server" [Undecided,New] https://launchpad.net/bugs/64980613:26
uvirtbotNew bug: #649807 in nova "Redis concurrency problems in API server" [Undecided,New] https://launchpad.net/bugs/64980713:26
*** ChrisAM1 has joined #openstack13:27
*** ChrisAM1 is now known as ChrisAM13:28
*** vishy has quit IRC13:37
*** ChrisAM has quit IRC13:37
*** comstud has quit IRC13:37
*** xtoddx has quit IRC13:38
*** devcamcar has quit IRC13:38
*** anotherjesse has quit IRC13:39
*** jdarcy has joined #openstack13:39
*** vishy has joined #openstack13:39
*** xtoddx has joined #openstack13:40
*** anotherjesse has joined #openstack13:40
*** ppetraki has joined #openstack13:41
*** devcamcar has joined #openstack13:41
*** ChrisAM1 has joined #openstack13:41
sorenchmouel: Did you build packages and install them?13:44
sorenchmouel: Or are you running from a bzr checkout directly13:44
chmouelbuilt them and install13:44
sorensudo nova-manage project zipfile blah blah13:44
chmouelI think I have figred out on that zip file stuff, it just I was missed the sudo13:44
sorenshould do it, then.13:44
chmouelsince it could not read the /var/lib/nova/CA stuff13:44
chmouelbut after i get a server error when trying to  euca-upload-bundle -b mybucket -m file.manifest..13:45
*** mtaylor has quit IRC13:50
*** ChrisAM1 is now known as ChrisAM13:55
sorengundlach: I would love your feedback on bug #649806 and bug #649807.13:59
uvirtbotLaunchpad bug 649806 in nova "AMQP concurrencty problem in API server" [High,New] https://launchpad.net/bugs/64980613:59
uvirtbotLaunchpad bug 649807 in nova "Redis concurrency problems in API server" [High,New] https://launchpad.net/bugs/64980713:59
gundlachsoren: *just* finished reading the second one, actually13:59
*** gasbakid has quit IRC14:01
sorengundlach: I've been running some tests with those crude workarounds I suggested. It seems to work pretty well.14:01
sorengundlach: ...but some sort of actual pooling would be better, of course.14:02
*** f4m8 is now known as f4m8_14:02
gundlachsoren: yeah, it seems like that's what the thread safety error is saying14:02
gundlachsoren: fyi, the redis usage in API is supposed to go away, tho vishy says he may not be able to get to it before FF14:04
sorengundlach: Exactly, so it needs to be addressed.14:04
sorenUnless, of course, I decide to attack the redisectomy myself.14:04
soren(or someone else does)14:04
gundlachsince this is firmly in "bug" territory, maybe we wait on this until we know whether we get redis out or not.14:05
gundlachif we don't, then we can replace w/ a pool.14:05
sorenRight now, I can start and kill around 40 instances before things start to go sour.14:08
sorenNot sure what happens then. Smells like a libvirt deadlock.14:10
*** annegentle has joined #openstack14:11
*** hisaki has joined #openstack14:16
gundlachsoren: in case you hadn't seen: http://blog.eventlet.net/2010/03/18/safety/14:16
sorengundlach: Seen it. Tried it. Failed miserably.14:17
gundlachmm, darn :)14:17
sorenI mean /I/ failed miserably.14:18
sorenNot it.14:18
*** krish has joined #openstack14:18
sorenI just coulnd't really wrap my head around it just then.14:18
*** anticw has joined #openstack14:19
*** ar1 has quit IRC14:22
*** guigui has quit IRC14:23
*** krish has quit IRC14:24
*** krish_ has joined #openstack14:24
*** ctennis has joined #openstack14:24
*** ctennis has joined #openstack14:24
*** krish_ is now known as krish14:27
*** jfluhmann_ has joined #openstack14:28
*** brd_from_italy has joined #openstack14:29
uvirtbotNew bug: #649865 in nova "uploading bundle failure" [Undecided,New] https://launchpad.net/bugs/64986514:36
*** rnirmal has joined #openstack14:37
*** pharkmillups has joined #openstack14:39
* soren calls it a day14:39
sorenWell, sort of.14:39
sorenUntil the meeting, anyways.14:39
dendrobatesgundlach soren: after discussing the redis work with jaypipes, he asked if we could punt on it this release.14:43
gundlachwhat would punting mean14:43
gundlachkeeping redis in Austin?14:43
dendrobatesno not adding it back.14:43
dendrobatesto the abstraction layer14:43
gundlachthat works, as long as the RS API gets redis extracted.  vishy had agreed to do that, but a couple days ago said he's too slammed to take care of it.14:43
gundlachs/extracted/replaced with db/14:44
dendrobatessomeone needs to do it.14:44
gundlachright -- it will take work to get redis out, and it will take work to leave redis in, and everybody's busy... not sure what the best solution is14:44
gundlachwhat time is the meeting -- 4pm EST?  maybe we can hash it out then14:46
*** cloudmeat has joined #openstack14:46
dendrobates4pm cst14:48
dendrobatesI will be on a plane, so jaypipes will be leading it.14:48
*** oliver_ has quit IRC14:49
*** Pentheus has joined #openstack14:56
jaypipesgundlach: I can remove the redis depends.14:58
* jaypipes notes that breakfast is indeed the best meal of the day.14:58
gundlachjaypipes: great -- there's a bug out for it (https://bugs.launchpad.net/bugs/649188).14:59
uvirtbotLaunchpad bug 649188 in nova "RS API is still on Redis" [High,New]14:59
gundlachvishy: ^^14:59
jaypipesgundlach: yep, saw it.  I'll tackle it today.  got one more thing with glance to push and then I'm on it.14:59
gundlachsweet.  dendrobates: i now have no problem punting on redis -- we don't have to discuss at 4pm mtg15:00
jaypipesgundlach: yeah, we still do.  there's still the fakeldap driver dependency on redis...15:01
gundlachah! my mistake15:01
jaypipesgundlach: I may be able to get to that one tomorrow, though.15:01
jaypipesgundlach: but I want to check with todd since he was going to look at that piece...15:01
jaypipesxtoddx: ^^15:01
*** anm_ has quit IRC15:05
*** ctennis has quit IRC15:09
*** ctennis has joined #openstack15:11
*** anm_ has joined #openstack15:13
*** kashyapc has joined #openstack15:14
*** kashyapc has joined #openstack15:26
*** sophiap has quit IRC15:27
*** sophiap has joined #openstack15:28
*** annegentle_ has joined #openstack15:30
*** mtaylor has joined #openstack15:31
*** ChanServ sets mode: +v mtaylor15:31
*** krish_ has joined #openstack15:33
*** annegentle has quit IRC15:33
*** krish has quit IRC15:36
*** gondoi has joined #openstack15:40
*** hisaki has quit IRC15:43
*** dpackard has joined #openstack15:57
*** dpackard has quit IRC15:58
*** brd_from_italy has quit IRC16:10
*** dendrobates is now known as dendro-afk16:13
*** SandCube has joined #openstack16:16
*** zheng_li has quit IRC16:18
*** ibarrera has quit IRC16:18
*** joearnold has joined #openstack16:18
*** joearnold has joined #openstack16:20
*** krish__ has joined #openstack16:27
jaypipesgundlach: how are you and eday testing the rackspace API?  I see the test cases in /nova/tests/api/rackspace, but I don't see them hooked in to the test suite runner...16:27
gundlachi'm using nosetests16:27
jaypipesgundlach: every time I run nosetests, it complains of no module named gflags, even when I run it in a venv :(16:28
gundlachtry 'python /usr/bin/nosetests' instead from inside venv16:28
gundlachi had the same problem16:28
jaypipesgundlach: ok.  would have been nicer to use the existing test system...just a nit.16:29
gundlachi agree, i'm sure that's the plan but they just haven't been hooked in yet.16:29
jaypipesgundlach: I didn't even realize there were unit tests until I saw them in the tests folder..16:29
gundlachyep, this shouldn't be a permanent arrangement :)16:29
gundlachi could take care of that after FF assuming no other bugs crop up16:29
jaypipesgundlach: cool :)16:30
*** krish_ has quit IRC16:30
gundlach(i'm not actually sure how to get regular unit tests running under the existing Twisted-based test framework, but i'll figure it out)16:30
jaypipesgundlach: just add the import line to the run_tests.py file.16:30
gundlachreally?  that's it?16:31
gundlachsomehow i thought i tried that and it vomited.  will have to try that again :)16:31
*** littleidea has quit IRC16:31
jaypipesgundlach: any other suggestions? http://codepad.org/cpZrO2SV16:32
gundlachpip install nose within your venv, maybe?16:33
gundlachyeah, looks like i installed 0.11.4 in the venv that i have in my nova/ dir (so i don't have to recreate it for every branch)16:33
*** SandCube has quit IRC16:34
jaypipesgundlach: hmm, ok.16:35
jaypipesgundlach: do you have to run nosetests in sudo? :(16:37
gundlachsorry you're having trouble -- you could add the line to run_tests.py instead if that's all it takes?16:38
jaypipesgundlach: well it asks for a sudo password halfway through...16:38
gundlachbefore running any tests?16:38
jaypipesno, during the test run.16:38
gundlachhmm which test?16:38
gundlach[if you give it the wrong credentials which test fails?]16:39
*** jakedahn has quit IRC16:39
*** jakedahn has joined #openstack16:39
jaypipesgundlach: well, unfortunately, instead of using/modifying the existing base test case in /nova/test.py, you used a new one, so I don't think that will work.16:39
jaypipesgundlach: nosetests is currently hung...lemme paste the output. one sec.16:39
jaypipesgundlach: http://codepad.org/w5h7GwM716:40
jaypipesgundlach: you can see near the bottom where I had to type in a sudo pass...16:40
gundlachi didn't use a new base test case -- i just inherited from the standard unittest.TestCase, didn't I?16:41
jaypipesgundlach: which I did. I don't think that is the source of the hang...16:41
edayjaypipes, gundlach: I was actually hoping to move tests the other way. Remove run_tests.py and make each test case independant so nosetests can run them16:41
jaypipesgundlach: there is a BaseTestCase in /nova/test.py16:41
gundlacheday: ++16:41
jaypipeseday: I wouldn't mind that.16:41
jaypipeseday: but if we have a test runner, we should improve and modify it, not just create new test suites with no docs as to where they are.16:42
*** littleidea has joined #openstack16:42
gundlachjaypipes: i agree, the api unit tests need to be integrated -- running nosetests to test them was a halfway solution until they were integrated.16:43
jaypipesgundlach: again, I have no problem with nosetests (I use it a lot and I like it).  Just wish there was some notification as to how to run these without issues :)16:43
gundlachjaypipes: not sure what to tell you, as it works on my box :/16:44
*** joearnold has quit IRC16:44
*** littleidea has quit IRC16:44
jaypipesgundlach: I created unit tests for the image service yesterday not even realizing there was a unit test in the tests/api/rackspace/ subdirectory (though it doesn't do anything.)16:44
jaypipesgundlach: sorry, but "it works on my box" isn't a particularly useful way to get people working on this code ;)16:45
*** pvo has joined #openstack16:46
*** ChanServ sets mode: +v pvo16:46
gundlachjaypipes: http://paste.openstack.org/show/33/16:46
gundlachjaypipes: let me rephrase: we've got features that haven't been written yet, and we've got a goal of a testing harness for the API to be finished.  neither are finished, so the API test system isn't in a complete state.16:47
jaypipesgundlach: is that paste showing the *correct* output?16:47
gundlachjaypipes: were you able to get the tests running successfully by importing them into run_tests.py?16:47
jaypipesgundlach: no16:48
jaypipesgundlach: I didn't even know they were there.16:48
gundlachjaypipes: yes -- one failure i think is from cerberus's upstream branch16:48
jaypipesgundlach: k, cool.16:48
jaypipesgundlach: is there a reason for all the output to stderr/out?  or just not completed yet?16:48
*** maplebed has joined #openstack16:48
gundlachjaypipes: i'm not aware of that either -- that started a couple days ago, so i assume ~cerberus/nova/servers_api turned on some debugging flag16:49
gundlach_cerberus_ might now16:49
jaypipesgundlach: l16:49
jaypipesgundlach: k :)16:49
jaypipesgundlach: my typing sucks today.16:49
_cerberus_A lot of the stdout junk was me testing stuff in my branch, yes16:49
gundlachjaypipes: so when you importing api and api.rackspace into run_tests.py it barfed?16:49
*** burris has quit IRC16:50
jaypipesgundlach: no, I have not done that yet.  I was just saying that, if you put the test cases in /nova/tests and used the existing BaseTestCase, all you should have needed to do was add the import line.16:50
gundlachahh i see16:50
jaypipesgundlach: I'll see what happens :)16:50
gundlachi vaguely remember that BaseTestCase's tornado-related stuff made eventlet unhappy, but i could have made that up16:51
jaypipesgundlach: is the /nova/tests/api_integration.py source file bogus?16:51
jaypipesgundlach: seems to be...16:51
gundlachno idea, it's been there since before i started the RS API.  i think it's frem EC216:51
gundlachapi_unittest.py is just EC2 i believe16:51
jaypipesgundlach: hmm, do you have /etc/sudoers/nova_sudoers file?16:51
jaypipesgundlach: so you do not have nova installed locally?16:52
jaypipesk, checking to see if that's the issue..16:52
* jaypipes doesn't have it installed locally either..16:52
gundlachheh, would be funny if tests failed only if nova were installed16:53
jaypipesgundlach: indeed :)16:53
*** pharkmillups has quit IRC16:55
*** krish__ has quit IRC16:59
*** zheng_li has joined #openstack16:59
*** krish__ has joined #openstack17:00
jaypipesgundlach: got it.  I was running nosetests in the rootdir, and the rpc_unittest.py file was failing miserably (since it depends on the other test runner)17:01
jaypipesgundlach: if I run nose in /nova/tests/api/ all is good.17:01
gundlachah, my mistake -- i run it from nova/tests/api17:01
gundlachsorry -- were i more leet i would have been able to paste my actual console session including my pwd, instead of just the test output17:01
gundlachdo you know how to make screen's Copy mode copy into the system clipboard?17:01
jaypipesgundlach: middle mouse button.17:02
gundlachdoesn't work in screen, which can't scroll with the mouse afaik?17:02
*** pvo has quit IRC17:02
jaypipesgundlach: if it's like xterm...should just be able to highlight an area, and middle mouse button will paste that.17:02
gundlachright, but my output was more than one window-ful.17:03
gundlachnm :)17:03
gundlachanyway, glad you figured it out and sorry for the trouble17:03
jaypipesgundlach: no worries mate17:03
gundlach_cerberus_: https://code.launchpad.net/~gundlach/nova/servers_api/+merge/3689417:05
*** jakedahn has quit IRC17:06
*** jakedahn has joined #openstack17:06
*** krish_ has joined #openstack17:07
*** krish__ has quit IRC17:10
_cerberus_gundlach: I'll take a look17:11
*** comstud has joined #openstack17:14
*** ChanServ sets mode: +v comstud17:14
*** littleidea has joined #openstack17:15
*** littleidea has quit IRC17:15
*** joearnold has joined #openstack17:16
*** burris has joined #openstack17:18
*** dizz|away has quit IRC17:19
*** kevnfx has joined #openstack17:22
*** jc_smith has joined #openstack17:25
*** pvo has joined #openstack17:25
*** ChanServ sets mode: +v pvo17:25
edaygundlach: for multi-page screen copies, I usually save to a /tmp file and just open to where I need it (possibly directly in the browser for copy/paste)17:27
*** Ryan_Lane has joined #openstack17:27
Ryan_Lanesoren: is there any way other than ppas to add additional apt sources to a vm using vmbuilder? we have a repository, but it isn't on launchpad...17:28
*** pharkmillups has joined #openstack17:33
gundlacheday: perfect, i forgot you could write to a file.  thanks!17:34
gundlach(there should be a /dev/clipboard that you can open for writing -- then you could copy to the clipboard :) )17:34
edaygundlach: that would be cool :)17:35
*** anm_ has quit IRC17:37
*** anm_ has joined #openstack17:38
gundlacheday: hah -- turns out cygwin has a /dev/clipboard :)17:44
*** littleidea has joined #openstack17:46
*** kevnfx has quit IRC18:02
*** kashyapc has quit IRC18:03
*** rlucio has joined #openstack18:03
*** krish__ has joined #openstack18:05
*** krish_ has quit IRC18:08
*** dendro-afk is now known as dendrobates18:22
_cerberus_eday: Mind checking out my merge prop? https://code.launchpad.net/~cerberus/nova/servers_api/+merge/3690418:25
eday_cerberus_: already was :)18:25
_cerberus_Ahh, thanks'18:25
*** krish_ has joined #openstack18:30
*** krish__ has quit IRC18:32
*** alekibango has quit IRC18:36
*** krish__ has joined #openstack18:38
*** krish_ has quit IRC18:39
*** krish_ has joined #openstack18:46
*** krish__ has quit IRC18:47
*** burris has quit IRC18:48
*** joearnold has quit IRC18:48
*** jakedahn has quit IRC18:48
*** ptremblett has quit IRC18:49
*** jakedahn has joined #openstack18:49
*** burris has joined #openstack18:49
vishyi just made an ugly ugly patch18:51
*** alekibango has joined #openstack18:57
*** krish__ has joined #openstack19:00
*** krish_ has quit IRC19:04
sorenvishy: Ooh, our first entry for our hall of shame?19:08
* eday denies the patch before it even makes it to a merge proposal19:09
*** pvo has quit IRC19:09
sorenRyan_Lane: Sure. So, you start by writing a patch... :)19:09
sorendendrobates: I'm perfectly cool with dropping Redis completely for Austin.19:10
vishysoren: so multi-region19:10
vishysoren: we're sharing users and projects19:11
vishysoren: so creating and destroing network and default security group on project creation doesn't work19:11
alekibangohi. what is going on?19:11
*** ewanmellor has joined #openstack19:11
dendrobatessoren: good, cause it looks like that is going to happen19:11
vishysoren: so I just hacked it in to api, which is a bit ugly19:11
ewanmellorMorning all19:12
*** jakedahn has quit IRC19:12
ewanmellorDoes anyone have a .deb for Squeeze for eventlet?19:12
*** krish__ has quit IRC19:12
sorenewanmellor: morning?!? Where are you?19:12
*** jakedahn has joined #openstack19:13
ewanmellorI'm in Cambridge, UK.  I'm just used to talking to Yanks, that's all ;-)19:13
sorenewanmellor: Ah :)19:13
ewanmellorsoren: You're my .deb God -- I don't suppose you've packaged eventlet have you?19:13
sorenewanmellor: Of course I have.19:14
sorenewanmellor: In Debian, even.19:14
sorenNot the right version, though.19:14
sorenAt least I thought I did.19:14
* soren checks19:14
ewanmellorI couldn't find it.19:14
alekibangowhat would dropping redis mean?19:14
sorenNot using it anymore.19:14
ewanmellorUnless it's disguised as python-event or python-gevent.19:15
sorenewanmellor: Err... I'm terribly sorry. I lied. I could have sworn I put it in Debian, too, but it seems not.19:15
* soren digs through e-mail from ftp-master.19:15
alekibangosoren: source packages might be enough :)19:15
sorenGrab 'em from Ubuntu.19:15
ewanmellorsoren: Is there a particular version I need?19:16
edaysoren, dendrobates: Dropping redis now? I'm in support of this, but I didn't see anything about it19:16
dendrobateseday: it came up yesterday, jaypipes can cover it in the meeting.19:16
sorenewanmellor: 0.9.12. It's in the nova-core PPA, and will be in Ubuntu as soon as I get my act together.19:16
dendrobatesI'll be on a plane, but I'm about to send a meeting reminder.  jaypipes will lead it in my absence19:17
ewanmellorsoren: http://ppa.launchpad.net/soren/nova/ubuntu lucid main ?19:17
ewanmellorOr have you made another PPA?19:17
sorenewanmellor: That may have it, too. I was referring to http://launchpad.net/~nova-core/+archive/ppa (whose repo is at http://ppa.launchpad.net/nova-core/ppa/ubuntu).19:18
ewanmellorsoren: Perfect, thanks.  I'll look there.19:18
*** krish__ has joined #openstack19:21
*** sophiap has quit IRC19:26
*** sophiap has joined #openstack19:26
*** krish__ has quit IRC19:27
*** krish__ has joined #openstack19:28
sorenewanmellor: Thanks for the (unintended, but helpful nonetheless) reminder. Just uploaded eventlet 0.9.12 to Ubuntu.19:28
sorenewanmellor: My notes reveal that I was expecting mtaylor would upload it to Debian. Why I expect that and why it hasn't happened are two questions to which I have no answer.19:29
sorenMaybe I made it all up.19:29
mtaylorsoren: hrm. for some reason I thought you were going to19:30
ewanmellorsoren: No problem -- thanks for the packages!  It's working a lot better now ;-)19:30
sorenmtaylor: I don't blame you. 5 minutes ago I thought I already did it!19:31
mtaylorsoren: one of these days you and I are going to figure out which of us is doing what19:31
sorenmtaylor: Apparantly, I'm doing python-gflags. I forgot all about that.19:32
mtaylorsoren: oh lovely19:32
sorenmtaylor: I put you as Original-Maintainer in the Ubuntu eventlet package. I'd be a liar (again) if you don't take that up on Debian :)19:32
*** joearnold has joined #openstack19:33
mtaylorsoren: hehe. ok. I'll take it up19:35
*** hazmat has joined #openstack19:35
*** jakedahn has quit IRC19:35
Ryan_Lanesoren: lol. that was the answer I was hoping not to hear :)19:37
_cerberus_pylint question: the routes lib, by default, maps requests for resources with a specific id to methods expecting an "id" param, which conflicts with both our override declaration and pisses pylint off. However, it's also default behavior for a 3rd party library. What's the correct choice here?19:40
*** krish_ has joined #openstack19:42
eday_cerberus_: we need to ignore those. I thought there was a .pylintrc rule to ignore them19:43
eday_cerberus_: are you running it with pylintrc included?19:43
_cerberus_Whoops :-D19:43
*** mtaylor has quit IRC19:44
_cerberus_eday: actually, the pylintrc I'm looking at doesn't seem to define such an exclusion19:46
*** krish__ has quit IRC19:46
eday_cerberus_: ok, you can ignore the line by appending # pylint: disable-msg=...  (where ... is the error code)19:47
_cerberus_eday: thanks19:48
sorenvishy: https://code.launchpad.net/~soren/nova/network-fixes/+merge/36700 ping (since you don't seem to be subscribed)19:52
*** krish_ has quit IRC19:55
vishysoren: is there a reason to start dhcp before an instance is started?19:56
vishysoren: is it just in case there are running instances already?19:57
sorenvishy: In case there's already an instance running that's waiting for it.19:57
sorenvishy: Precisely.19:57
vishysoren: ok, looks good19:57
vishyi'll approve19:57
sorenvishy: I changed another few things today, don' tknow if you noticed.19:57
sorenvishy: Notably s/os.kill/_execute('sudo kill')/19:58
vishyyeah saw that19:58
*** ptremblett has joined #openstack19:58
vishyi always run as root, so i didn't notice that problem19:58
sorenvishy: I guessed :)19:59
vishybtw: the lucid version of libvirt still fails with security groups19:59
sorenvishy: Even with aa-complain /usr/sbin/libvirtd?20:00
vishyinstalling the maverick packages and making sure gawk ebtables and iptables are in works20:00
vishysoren: yeah20:00
sorenvishy: With kvm?20:00
sorenvishy: Or user-mode-linux?20:00
vishysoren: kvm20:00
sorenWhat sort of errors are you seeing?20:00
sorenIn your honour, I've been doing my testing under Lucid.20:00
sorenIt works fine for me.20:00
devcamcarcerberus: i'm fixing something in our ec2 api that i think could help the rs api as well20:00
*** dendrobates is now known as dendro-afk20:01
devcamcarcerberus: i'm adding a field to our db schema called "api_id"20:01
vishyhmm...i'll have to roll back to the old version to find out20:01
sorenvishy: Just a vague recollection?20:01
devcamcarcerberus: it looks like you guys are having to do an extra layer of mappings20:01
*** pvo has joined #openstack20:01
*** pvo has joined #openstack20:01
*** ChanServ sets mode: +v pvo20:01
vishyi think it was the same error i was seeing regarding ebiptables20:01
vishyno driver for ebiptables or some such20:01
devcamcarcerberus: and i think if each api sets this id as it wants and pushes it to database, that we can bypass a layer of mappings20:02
gundlachdevcamcar: i don't know that i agree with that...20:03
gundlach1) doesn't that segment our data into two groups, depending on which API you used?20:03
gundlach2) why push API info down into the DB layer instead of keeping it encapsulated and hidden in the API layer?20:03
devcamcargundlach: no, it just lets each api set the api specific string representation as it sees fit20:03
devcamcarso ec2 uses something like i-abc123 for an instance id20:04
gundlachright, so if i create an instance with ec2, can i later fetch data about it with the RS api?20:04
devcamcarand that needs to be stored - currently its based on the PK, which is bad20:04
devcamcargundlach: yes20:04
gundlachah, ok, i must be misunderstanding20:04
devcamcargundlach: i'm not building any walls between anything20:04
devcamcarits just a string20:04
devcamcarfor reference:20:05
_cerberus_devcamcar: Sounds interesting. re: the notes on my merge, the branch isn't done, either. I hadn't gotten to handling the mapping of ids yet. I was just attempting to merge so I can continue to work on it without generating an encyclopedic diff20:05
devcamcarright now in this merge there is an ec2_id, but we decided that we don't want api specific stuff in the database20:05
devcamcarso i'm going to repush in a few and renamed ec2_id to api_id20:06
gundlachok walk me through this: in RS i create a server, whose id is the integer 2.  nova saves that into api_id as a string "2" i guess.  then i try to read my list of instances via the EC2 api.  what will i get?20:06
devcamcarwhich means you guys can push the rackspace id there as well20:06
_cerberus_devcamcar: re: api stuff in the database, I added a model for our Rackspace auth tokens that may need to get pushed somewhere else20:06
devcamcargundlach: so in your case, pk=2, api_id="2", if i then do an ec2-describe-instances i will just see a "2" where normally you would see "i-abc123", and you can use "2" in place of "i-abc123"20:07
gundlachok, got it.  now the reverse: i create an instance via EC2 and get "i-abc123" or some similar string stored into api_id.  then i try to read my servers via the RS API.  what are their IDs?20:08
*** krish_ has joined #openstack20:08
devcamcargundlach: that is the part that i'm not clear on - i don't have much experience with the rs api yet20:08
xtoddx_cerberus_: have you looked at https://code.launchpad.net/~xtoddx/nova/authserver?  has a session token model.  You should probably pull out whatever is useful for you from that.20:09
gundlachyeah; the problem is that IDs are integers that can be reused between users [aka, each user gets his own id space].  i guess there's nothing stopping us from changing that so we have a single id space across all users, and new entities don't monotonically increase their IDs20:09
gundlachdevcamcar: but i don't know how to map strings to ids :)  would it be an option to make all EC2 ids strings that only contained digits?20:10
_cerberus_xtoddx: I hadn't, but I will now. Thanks20:10
*** theron has joined #openstack20:10
edaygundlach, devcamcar: Can't we just used unique 64-bit ints and have i-[hex of id] for ec2 and just use the plain ID int for rs?20:12
xtoddxgundlach: or you can set api-id to username:id or something20:12
edayI don't see why each RS instance needs it's own namespace20:12
gundlacheday: i guess it would only matter if some code had come to rely on that, but i doubt it.20:13
gundlach[separate id spaces]20:13
edaygundlach: I can't imagine any code that would, it should just treat them as ints20:13
*** Mike has joined #openstack20:13
devcamcargundlach: you can still do that if you want, its just up to the rackspace api to choose how to populate the api_id field20:13
devcamcari think it provides a good common ground20:14
*** Mike is now known as Guest1980020:14
devcamcarthe strings do have to be unique though20:14
devcamcarso username:id ++20:14
devcamcarthere is currently a unique constraint on the field20:14
edayalso, we probably don't want anything that is auto-incrementing - this is the death of distributed apps. we should be using unique IDs (UUIDs if space allows or some custom algorithm)20:15
gundlacheday: re 64-bit ints: maybe ec2 uses i-[hex of userid]-[hex of id], and rsapi just uses the id?  that lets us retain id space per user, if that helps scaling20:15
edaygundlach: sure20:15
*** rnirmal has quit IRC20:16
jaypipeseday: ++ on uuids.20:16
gundlachdevcamcar: maybe to make this clear, you could make the api_id field an int (the lowest common denominator)?20:16
gundlachor long, rather20:16
devcamcareday: we do need to make the actual row id's themselves proper UUIDs instead of auto-increment ints, but that is a different issue20:17
gundlachso that any consumers can know how to translate that to/from the format they require20:17
devcamcargundlach: the lowest common denomintator is a string, and if you guys need to denorm an int into a string i think thats fine20:18
*** Guest19800 has quit IRC20:18
gundlachdevcamcar: if it's a string, then some future code that forgets that some people have to translate to an int might store e.g. "abc"20:19
edaydevcamcar: hmm, I would thing we want the canonical type to be int, string->int conversions can break, but int->string never fails20:19
gundlachif it's an int, then all code is forced to turn it into an int, which all APIs can be sure to consume20:19
devcamcarthe whole point is that ec2 api doesn't store ints here though, so i'm trying to find a middle ground20:19
edaydevcamcar: the backend API can store ints, and just format it as "i-%x" % instance_id20:20
edayfor ec220:20
_0x44sirp1: You want to look this over before I suggest merging? http://bazaar.launchpad.net/~chris-slicehost/glance/implement_teller_parallax_lookup/revision/520:21
devcamcari must enforce uniqueness on the field as well, how would rackspace id work in that context?20:21
gundlachdevcamcar: rackspace can switch to a single id space20:22
gundlachrather than one per user20:22
edaydevcamcar: string or int, we just need enough random bits to be unique. uuid is "guaranteed" but is 128 bits. We can probably get away with a 64-bit int though20:22
eday64-bit gives (if we do time-based), 4096 IDs per microsecond20:23
vishysoren: i proposed my ugly hack to initiate commentary20:24
gundlacheday: you OK with exposing the create time of the entity via the ID?20:24
sorenvishy: link?20:25
edaygundlach: I personally don't care, but others may. I don't like secrets20:26
gundlachsame as i feel.20:26
gundlachlet's proceed with it, and if there are complaints we can change the algorithm :)20:26
edaygundlach: we can always scramble it in a non-reversable way that keeps the uniqueness properties20:26
gundlachwe can?  i don't have enough math to be able to say that with confidence.  cool.20:26
edaygundlach: I don't know off the top of my head, but I'm sure there is a way :)20:27
edaygundlach: or we can just use crypto-level random 64-bit ints and hope for the best, haha20:27
sirp1_0x44: have you merged trunk into your topic branch (to pull in jaypipes directory re-org)?20:27
_0x44sirp1: Yes, I branched this yesterday.20:28
gundlachyeah, i was just trying to figure out how big the brute force dictionary for a 64 bit key was :)  anyway, it's bigger than the secret we're trying to protect.20:28
_0x44sirp1: http://0x44.pastebin.com/JkDvA9TP20:28
*** littleidea has quit IRC20:29
devcamcarwell this all sounds good, but i think you guys are missing that point that i can't convert a UUID to the format "i-abcdefgh"20:29
*** ctennis has quit IRC20:29
*** littleidea has joined #openstack20:29
devcamcari dont want to make our ec2 api users type a 34 digit string everytime they want to terminate an instance20:29
gundlachdevcamcar: is there a length limit on the ec2 id?20:29
devcamcargundlach: not that i know of, but that doesn't mean i want to punish users with a 34 digit string20:29
gundlachdevcamcar: how long would it be in base 26?20:30
*** littleidea has quit IRC20:30
edaydevcamcar: we're not proposing using a proper UUID, just a 64bit int that would be "i-00000000"20:30
sorengundlach: 64 bits?20:31
gundlachsoren: yeah but i convert to base 26 i'm asking how many digits it could take.  i think it's log base 26 of 2^64 but i'm not sure20:31
edaydevcamcar: are there length restrictins on the ec2 string?20:31
sorengundlach: 14 characters.20:31
*** kevnfx has joined #openstack20:31
sorengundlach: It is.20:31
edaydevcamcar: also, is it case sensitive?20:31
gundlachdevcamcar: 16 char string isn't horrible20:31
devcamcari assume it is case sensitive20:32
*** jtimberman has quit IRC20:32
sorenWhat does EC2 use? 7 characters, or so?20:32
edaydevcamcar: then base64 would be fine too20:32
* soren doesn't remember.20:32
devcamcarso whats the problem with you guys storing a string?  this is not meant to be an integer field. ec2 id is specifically s string20:32
sirp1_0x44:  jaypipes change hit trunk about 6 hoours ago, it moved parallax and teller under a new "glance" directory: http://bazaar.launchpad.net/~hudson-openstack/glance/trunk/revision/520:32
gundlachdevcamcar: yeah, but we're trying not to tie to a specific API in the DB, remember?20:32
devcamcarso touching the data outside of python, for instance, for reporting, becomes impossible without another layer of transform20:32
devcamcarright, which is why i proposed a string :)20:33
_0x44sirp1: Ah, then no, I haven't yet. :\20:33
edaydevcamcar: because it's ec2-specific. the canonical instance ID will need to be a 64-bit int20:33
devcamcari could also argue that an int is rackspace specific in this case20:33
gundlachdevcamcar: yes, but an int can be converted to a string automatically, guaranteed20:33
edaydevcamcar: so rather than keep an extra field for ec2->int mapping, we can just make that a quick str format20:33
devcamcarthe canonical id is not the api_id20:33
devcamcarit is the row_id20:33
*** mtaylor has joined #openstack20:33
*** ChanServ sets mode: +v mtaylor20:33
vishysoren: https://code.launchpad.net/~vishvananda/nova/move-default-group/+merge/3692320:33
gundlachan arbitrary string can't be converted to an int20:33
*** jtimberman has joined #openstack20:34
gundlachand you're right that using an int instead of as tring will require another transform in ec2 reporting tools -- but keeping it as a string would require another layer of transformation in any RS API reporting tools :)20:34
edaygundlach: (well, it can, it just gets really large... it all bits... ;)20:34
devcamcarso the rs api requires the int?20:34
_0x44sirp1: Merging now20:34
devcamcarif thats the case i'm happy to do the transform20:34
gundlacheday: yeah, i don't think that's what we're going for :)20:34
gundlachdevcamcar: yes20:34
pvoeday: gundlach: we'd talked about doing a vm id combined with the region to make it unique within a region to avoid uuids20:34
gundlachdevcamcar: the current RS API spec gives all IDs as integers20:34
pvoif we are able to version the api int eh future20:35
_0x44One crazy thing about the RS public API is that we don't refer to the images internally by integer id, only by UUID string20:35
devcamcari have to research if the ec2 tools can support longer ids20:35
*** krish__ has joined #openstack20:35
sorenvishy: Yikes! @ security_group_destroy_all20:36
vishyyou don't have to convert if you have api_id vs id20:36
vishysoren: hehe, i needed a quick way to clean them up after tests20:36
vishysoren: I'm thinking the flag idea might work if FLAGS.cleanup_db_on_project_destroy:20:37
vishysoren then delete security groups and disassociate networks20:37
_0x44pvo: Why would we want to avoid UUIDs? In our current system they're guaranteed unique because they eventually hit Nucleus and should get sorted out there.20:37
pvo_0x44: for the reason devcamcar mentioned. How do I reference that to anyone? Its guaranteed I'll have to email it or msg it.20:38
pvoif it is unique per region with an id and a name, there are other bits that can be identified with it.20:38
pvoor even just region:id20:38
*** krish_ has quit IRC20:39
*** littleidea has joined #openstack20:39
pvouuids don't tell you much about the instance. if the urn had a region:id:name or just region:id you have a bit more info about it and wouldn't need to look any info up about an vm if it isn't in your region20:39
_0x44UUIDs shouldn't tell you much about the instance. They're identifiers, not overloaded metadata objects.20:40
pvoex. if I hit the IRE instence and pass it a urn of PHX:12345:somesillyname, don't bother with the request20:40
pvoa url is both20:40
_0x44If we're looking for something with universality like a URL we should use a URL instead of making our own format.20:41
*** miclorb_ has joined #openstack20:41
jaypipesgundlach, _cerberus_: so, when's the last time you ran nosetests with --all-modules to actually run the /nova/tests/api/rackspace/auth.py file?  I'm showing lots of failures.20:42
_0x44pvo: We should register openstack as a valid namespace if we want to do something like that.20:43
_cerberus_jaypipes: I've been running it individually without nosetests and everything passes20:43
*** cloudmeat has quit IRC20:43
pvo_0x44: we could.20:43
*** cloudmeat has joined #openstack20:44
jaypipes_cerberus_: hmmm.20:44
sorenvishy: I'll need to look at it tomorrow. i'm having trouble forming opinions on it right now.20:44
*** theron has quit IRC20:44
_0x44Then the id could be urn:openstack:nova:region:####:name, and we assume urn:openstack:nova during customer communication to nova, and urn:openstack:swift for swift.20:44
jaypipes_cerberus_: http://codepad.org/R9Rua5Fe20:45
sorenOpenStack is the software, not a service provider.20:45
openstacksoren: Error: "is" is not a valid command.20:45
sorenopenstack: It wasn't meant as a command, so there.20:45
openstacksoren: Error: "It" is not a valid command.20:45
sorenopenstack: ditto20:45
openstacksoren: Error: "ditto" is not a valid command.20:45
gundlachjaypipes: i'm not familiar with all-modules; i thought nosetests ran every test it could find by default20:46
jaypipesgundlach: no, it runs tests objects in all files with "test" in them.20:46
*** Cybodog has quit IRC20:47
_0x44soren: URN namespaces don't seem to have to identify service-providers.20:47
jaypipesgundlach: which is why, for instance, /nova/tests/api/wsgi_test.py works.20:47
gundlachah, gotcha20:47
_cerberus_jaypipes: I've been running everything inside of venv, but http://paste.openstack.org/show/35/20:47
_cerberus_That's in my servers_api branch20:47
jaypipes_cerberus_, mtaylor, gundlach: we *really* need to hook Tarmac into running these API tests...20:47
_cerberus_perhaps it's broken in trunk atm20:48
jaypipes_cerberus_: I'm in a venv.20:48
*** cloudmeat has quit IRC20:48
*** allsystemsarego has quit IRC20:48
_0x44soren: But I agree, making every id a URL would probably be a better idea in the long term?20:48
jaypipes_cerberus_: you can tell by the prompt in the codepad paste...20:48
*** Cybodog has joined #openstack20:49
_cerberus_jaypipes: indeed20:49
pvo_0x44: I guess my thinking is if the vms were identified by a urn, it can be easily distinguished by a user as to which endpoint to ask for metadata about it instead of querying every endpoint you know about to find it.20:50
soren_0x44: I've always envisioned it eventually being a url.20:50
devcamcarall: slightly different suggestion - since disk space is cheap and the requirements of the api specific id's are fairly significantly different, i think it makes sense to store them separately now that i understand the rackspace requirements a bit better20:50
sorenRight now, we don't have the luxury if such experimentation with the API design, though.20:51
sorenSoon, but not right now.20:51
zulsoren: i was thinking can you make the /var/lib/nova/nova.sqllite in the nova-common.postinst20:51
pvosoren: agreed20:51
devcamcarforcing them to share a specific id in this way seems counterproductive, but i suggest we discuss this in the context of multi region at the next design summit20:51
sorenzul: Possibly.20:51
jaypipes_cerberus_: maybe I don't have your current work?  is it merged yet?20:51
devcamcarfor now i can just store an ec2_id and quickly move past this, and i have no problem with there also being a rs_id field in there as well20:51
sorenzul: Hm... Actually, no.20:51
jaypipes_cerberus_: though all this test does is talk to the mock router and db...20:51
_0x44pvo: You'd get the same thing from a URL and have the added bonus that it's parsable with urllib ;)20:52
_cerberus_jaypipes: no, nothing is merged at the moment. I'm working on some pylint/pep8 related cleanup as suggested by eday, and then I'm going to repush for approval20:52
zulsoren: why not?20:52
sorenzul: I've thought about it before, but only just now realised why it's not a good diea.20:52
sorenzul: ...because it shouldn't hard code the use of sqlite.20:52
_0x44pvo: err urlparse20:52
gundlachjaypipes: lp:~cerberus/nova/servers_api20:52
sorenzul: The real problem is that, for some reason, it gets created with root:root ownership.20:52
sorenzul: I'm assuming that's what you're meaning to fix, right=?20:53
zulsoren: agreed..then it should create the the right owner20:53
pvo_0x44: true, but api.dfw.rackspacecloud.com/servers/12345 seems it would contain a lot of extra info20:53
sorenzul: I really want to understand why it gets created that way to begin with. Everything setuid()'s to nova, so something must be instantiating the db before calling setuid. That sounds like a bug anyway.20:54
_cerberus_jaypipes: I'm also having some really annoying pylint issues. I just realized I was using a too-new version, so I rolled back, and now I'm getting this http://paste.openstack.org/show/34/20:54
*** ctennis has joined #openstack20:54
*** ctennis has joined #openstack20:54
jaypipes_cerberus_: no idea :(20:55
soren_cerberus_: Try clearing out your .pyc files?20:55
zulsoren: its a bug imho20:55
sorenzul: Precisely. Creating the db in postinst is a workaround. I'd rather actually fix it. If possible, that is.20:56
*** jdarcy has quit IRC20:56
zulsoren: ack20:56
*** dabo has joined #openstack20:56
*** littleidea has quit IRC20:56
sorenRelease meeting in #openstack-meeting in two minutes!20:58
jaypipeschrist there are so many levels of abstraction in nova...20:59
sirp1_cerberus_:  i figured out what the issue with the daemon library‚Ķ. it relies on a module called lockfile, and the version in PyPI appears to incompatible because it defines a function called LinkFileLock rather than a class20:59
vishyjaypipes: i think we could use 2 or 3 more21:00
jaypipesvishy: it's driving me nuts to be honest...21:00
mtaylorjaypipes: yes. totally agree21:02
mtaylorI'm going to try to carve out time to get that finished over the next coupla weeks21:02
sorenzul: Hm.... When did you last see this problem?21:02
*** rnirmal has joined #openstack21:02
zulsoren: the db problem?21:02
sorenzul: Yeah.21:02
anotherjessemtaylor / jaypipes suggestions for improvements?21:03
zulsoren: this morning when i was trying to install openstack from lp:nova/ubuntu21:03
sorenzul: I /think/ the api server may have been at blame, but it's been completely replaced, and I can't seem to reproduce it.21:03
mtayloranotherjesse: yes21:03
sorenzul: Oh, that's ancient.21:03
mtaylorjaypipes: well, the 'have hudson/tarmac run api tests'21:03
zulsoren: oh21:03
zulsoren: ill try it again tonight21:03
sorenzul: You're meant to branch trunk and merge lp:nova/ubuntu into that.21:04
zulsoren: ok...nice to know :)21:04
sorenzul: I'll update lp:nova/ubuntu just for giggles, but it was never meant to be used on its own.21:04
jaypipesmtaylor: ya, I gotcha.21:04
*** metcalfc has joined #openstack21:04
zulsoren: thanks21:04
*** dabo has left #openstack21:04
*** burris has quit IRC21:06
*** krish__ has quit IRC21:27
*** _0x44 is now known as rootkit21:28
*** jakedahn has joined #openstack21:30
*** miclorb_ has quit IRC21:31
jaypipesewanmellor: got a sec to join us in #openstack-meeting?21:31
gundlachach i missed the meeting21:33
*** rootkit is now known as _0x4421:34
*** burris has joined #openstack21:35
*** krish__ has joined #openstack21:43
*** abecc has quit IRC21:44
*** annegentle_ has quit IRC21:52
sorenxtoddx: https://launchpad.net/bugs/649806 https://launchpad.net/bugs/649807 <--- Sound familiar?21:52
uvirtbotLaunchpad bug 649806 in nova "AMQP concurrencty problem in API server" [High,New]21:52
xtoddxi had a branch where i undid the authmanger singletone and had no problems21:52
sorenxtoddx: Yup. I did the same for rpc.Connection and everything was suddenly smooth.21:53
uvirtbotLaunchpad bug 649807 in nova "Redis concurrency problems in API server" [High,New]21:53
xtoddxsoren: yea, with ldap and with redis21:53
xtoddxand de-singletonizing was the right call21:53
xtoddxi actually had problems putting them in a pool, but having them be an instance when needed wasn't a problem21:53
edayxtoddx: it works fine, but it means creating a new tcp connection/amql handshake (which is like 3RTTs) for each API request that needs a response21:54
sorenxtoddx: Yup. Connections seem to be cleaned up nicely and everything, but we should probably use eventlet's pooling thingamajig nonetheless.21:54
edayxtoddx: pooling and reuse of cached connections is the way to go21:54
soreneday: Ouch, 3 roundtrips?21:54
soreneday: I didn't realised.21:54
xtoddxyea a pool for anything that takes multiple trips to auth is a good idea21:54
xtoddxeday: would there be a possibility of connection timeouts in a pool?21:55
* jaypipes wonders why we don't just use Celery... http://ask.github.com/celery/21:57
*** ArdRigh has joined #openstack21:58
edaysoren: AMQP is a feature rich but chatty protocol21:58
edaysoren: it's ok for persistent conenctions, but short lived ones have an overhead21:58
* Daviey recently toyed with celery, and was more impressed with it than i expected to be21:58
jaypipesDaviey: so, so much better than carrot.21:59
jaypipesDaviey: plus, the @task decorators and webhooks would be perfect in Nova... but I digress ;)22:00
*** gondoi has quit IRC22:02
alekibangojaypipes: maybe because AMQP is a hot buzzword --  and rabbitmq is pretty nice, reliable and scaleable implementation...22:02
alekibangoi never heard about celery :)22:02
alekibangoah, it uses rabbitmq :)22:03
alekibangoic now22:03
creihtcelery is more about queueing tasks rather than a general message queue22:04
alekibangoi would need that celery for another project of mine :) nice catch22:04
*** pvo has quit IRC22:05
Daviey(it integrates into Django really well fwiw)22:05
alekibangoits really lovely - for build server i need22:06
sorenzul: Merged trunk into ubuntu branch, just for you :)22:06
* soren calls it a day22:07
Davieybetter than calling it a fish, i suppose.22:08
*** cyonyx has joined #openstack22:14
*** littleidea has joined #openstack22:15
burrissay I'd like to setup my own Hudson CI server and start with the same jobs as on hudson.openstack.org, are the Hudson configs shared anywhere?22:23
creihtmtaylor: -^22:24
*** cloudmeat has joined #openstack22:29
*** kevnfx has quit IRC22:32
*** ppetraki has quit IRC22:34
*** cyonyx has quit IRC22:37
*** krish_ has joined #openstack22:41
*** pharkmillups has quit IRC22:43
*** krish__ has quit IRC22:44
mtaylorburris: no - but I certainly can share them... can you email me (I'm about to run out) and remind me to do that?22:47
*** joearnold has quit IRC22:47
_cerberus_mtaylor: I've asked a few times, but no one seems to know. Are you aware of this issue with pylint 0.19? http://paste.openstack.org/show/34/22:48
mtaylor_cerberus_: nope. never seen that before22:49
_cerberus_Fair enough22:49
*** krish_ has quit IRC22:52
*** krish_ has joined #openstack22:55
*** littleidea has quit IRC22:57
eday_cerberus_: was it working for you before? if so what changed to break it?22:59
_cerberus_eday: Turns out I was using a newer version by mistake, so I attempted to roll back to 0.1923:00
edaywas the old version removed first? might be a conflict there23:00
edayis this on an ubuntu box, or your mac?23:01
_cerberus_eday: both, and as an extreme measure, I completely wiped out all of my site-packages on the ubuntu slice and started over23:01
_cerberus_This time using the pip-requires as the basis23:02
edayfor ubuntu, try apt-get install pylint23:02
edaythats the one I'm using23:02
_cerberus_Hmm, ok23:02
_cerberus_Yeah, that seems to be fine. Wonder if it's patched from the version available through pypi23:03
* eday shrugs23:04
_cerberus_Thanks for the help either way23:04
edayi always hit ubuntu/nova ppas befoer trying pypi for things.. easier to upgrade :)23:04
*** littleidea has joined #openstack23:09
*** joearnold has joined #openstack23:12
*** krish_ has quit IRC23:15
*** rnirmal has quit IRC23:20
*** pharkmillups has joined #openstack23:22
*** abecc has joined #openstack23:24
*** pharkmillups has quit IRC23:25
*** krish_ has joined #openstack23:27
*** joearnold has quit IRC23:28
*** abecc has quit IRC23:31
*** pvo has joined #openstack23:32
*** pvo has joined #openstack23:32
*** ChanServ sets mode: +v pvo23:32
*** pvo has quit IRC23:39
*** krish__ has joined #openstack23:40
*** ArdRigh has quit IRC23:43
*** ArdRigh has joined #openstack23:43
*** krish_ has quit IRC23:43
*** jc_smith has quit IRC23:59

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!