| [10:08] <bac> hi gmb | |
| [10:20] <bac> gmb, the latest drop of juju deprecated some settings in environment.yaml in favor of usng --constraints. it makes the rewriting o the environments file done by your run_lp_tests script unecessary. i've made changes to it and to charmhelpers to account for juju changes but haven't completed the changes or pushed them anywhere yet. | |
| [10:21] <bac> i'll do that after breakfast. just wanted to give you a heads up so you don't duplicate effort. | |
| [10:45] <gmb> bac, Awesome, I was just looking at that. | |
| [10:45] <gmb> I'll let it lie, then. | |
| [10:45] <bac> gmb: great. i'll get to it post-oatmeal | |
| [12:10] <gary_poster> bac benji frankban gmb, call asap | |
| [12:10] <gmb> Booting FF now. | |
| [12:10] * bac enhordes | |
| [12:33] * frankban lunches | |
| [12:33] <bac> i have pushed run_lp_tests branch to lp:~yellow/+junk/run_lp_tests | |
| [12:33] <bac> that allows me to move the card, so the lane is unblocked now | |
| [12:34] * gmb moves bug 974586 card, blocks the lane again. | |
| [12:34] <_mup_> Bug #974586: bugtracker test logging failures <paralleltest> <Launchpad itself:Triaged> < https://launchpad.net/bugs/974586 > | |
| [12:41] * bac should read warthogs more. didn't know poolie was leaving. :( | |
| [12:55] * gmb -> lunch | |
| [12:56] <bac> how does one set up PYTHONPATH so that it'll load the code being tested instead of the system version? i would've expect this approach to work, but the local version (/home/bac/add-charm-helpers/helpers/python) is stuck in the middle and isn't getting loaded. http://paste.ubuntu.com/924798/plain/ | |
| [12:57] <benji> gary_poster: I'm back for the time being teathered to my phone. I doubt I | |
| [12:58] <benji> can do video confrencing, but other things seem to be ok. | |
| [13:00] <bac> so eggs in /usr/local/lib always get inserted before PYTHONPATH? | |
| [13:00] <gary_poster> cool benji. replying to jml now. his branch looks good | |
| [13:01] <gary_poster> bac, that's a policy; maybe a deb one | |
| [13:01] <bac> ergh | |
| [13:02] <benji> gary_poster: will you point me at his branch? | |
| [13:02] <bac> gary_poster: do you know a work-around? gmb added this to the makefile but it is ineffective: | |
| [13:02] <bac> PYTHONPATH=helpers/python python helpers/python/charmhelpers/tests/test_charmhelpers.py | |
| [13:03] <gary_poster> benji, sorry forgot you did not have IRC history. #launchpad-dev now fwiw. https://code.launchpad.net/~jml/testtools/forward-current-tags/+merge/101538 | |
| [13:04] <benji> thanks; I'm following along | |
| [13:12] <frankban> bac: if you only need to get rid of the system site packages to test an app locally, I think virtualenv is an easy tool to get that kind of isolation | |
| [13:14] <bac> frankban: but the package in question is the one being tested. i think requiring people to use virtualenv is a bit heavy handed. i can achieve the same results by hacking sys.path in the test. ugly. | |
| [13:16] <frankban> bac: oh, I thought it was only for your local environment | |
| [13:17] <gary_poster> bac, sorry, was talking on launchpad-dev. reviewing backlog... | |
| [13:17] <bac> np | |
| [13:18] <gary_poster> bac, system Pythons suck for isolation, sorry. sys.path hack might be best bet. -S will leave a clean sys.path initially, but will probably not be what you want | |
| [13:19] <bac> gary_poster: sys.path hack isn't so bad. mainly i was just surprised to see my PYTHONPATH stuck in the middle. :( | |
| [13:19] <gary_poster> bac, it's pretty common to make a requirement that virtualenv --whatever-that-option-is be required in these kinds of circumstances I think | |
| [13:19] <gary_poster> yeah | |
| [13:19] <gary_poster> reminds me of dark days trying to figure that stuff out :-P | |
| [13:26] <bac> i've pushed a new branch of charm-tools to lp:~yellow/charm-tools/trunk and have created a merge proposal. | |
| [13:26] <bac> until it is packaged we should try to work off of the ~yellow version | |
| [13:27] <gary_poster> gmb, you up for reviewing lpsetup branch from frankban? | |
| [13:38] <gary_poster> bac, actually, did you see than Sean actually committed code with his fix? | |
| [13:39] <bac> gary_poster: i did not. | |
| [13:39] <gary_poster> I'm tempted to use it, rather than ours, and see if it works | |
| [13:39] <bac> that's good. | |
| [13:39] <gary_poster> bac, but a review from you would be appreciated by him and me | |
| [13:41] <bac> gary_poster: ok. i'll add comments to the bug since he didn't use a MP | |
| [13:41] <gary_poster> bac, makes sense | |
| [13:42] <bac> i'm a bit partial, but i think my approach was cleaner. :) | |
| [13:43] <gary_poster> bac, heh, you are probably right :-) | |
| [13:46] <bac> gary_poster: so he has committed code but not made a release. are you still going to make an interim egg and package it in LP? | |
| [13:51] <gary_poster> bac, yes, that's what I was thinking. | |
| [13:53] <gmb> gary_poster, frankban: Just finishing lunch etc., but I'll be happy to take that review in the next 20 minutes or so. | |
| [13:53] <gary_poster> thx gmb | |
| [13:53] <frankban> thanks gmb | |
| [13:56] <frankban> gary_poster: when you have a minute, I'd like to discuss about lxc and dns servers | |
| [13:58] <gary_poster> ack frankban. should be available in 3 | |
| [13:58] <frankban> thanks gary_poster | |
| [14:00] <gary_poster> frankban, in https://talkgadget.google.com/hangouts/_/extras/canonical.com/goldenhorde & ready whenever you are | |
| [14:10] <gmb> frankban, Is there a diff for your lpsetup changes somwhere? | |
| [14:11] <frankban> gmb: https://code.launchpad.net/~frankban/lpsetup/create-scripts/+merge/101536 | |
| [14:11] <gmb> Ta | |
| [14:25] <bac> gmb, frankban: please see my comment on frankban's MP. | |
| [14:26] <gmb> bac, I'd already made the same note :) | |
| [14:26] <bac> okey dokey | |
| [15:03] <frankban> thanks bac, I've just replied | |
| [15:11] <bac> frankban: we need to agree as a group whether our related work will use the LP style guide or not. IIRC "launchpad related" projects have continued to use it. from a ease-of-programming perspective that is what i would advocate, meaning i'd prefer to have one set of guidelines and use it across projects where practical. the exception to that rule is for projects that will be community-supported and only have a tangential relationship to LP. | |
| [15:14] <gary_poster> bac, frankban, given that lpsetup has PEP 8 everywhere, I'd suggest keeping PEP8 for this MP, with bac raising the issue on the dev list. Depending on that resolution, we can have a future style-only branch. | |
| [15:14] <gary_poster> for now, let's go with consistency within lpsetup | |
| [15:15] <gary_poster> well, and for the future too; we can switch all at once | |
| [15:17] <bac> frankban, gary_poster: i hadn't reviewed the whole branch and didn't realize that was the case. our style guide also says "stay consistent with existing code" or something like that. | |
| [15:17] <gary_poster> right bac, cool. still I agree with your assertion that we should collectively agree | |
| [15:18] <gary_poster> about this kind of policy | |
| [15:18] <gary_poster> bac, if we have not already agreed...? | |
| [15:18] <bac> with open("worms") as can: | |
| [15:18] <gary_poster> heh | |
| [15:18] <bac> colour = random.choice(palette) | |
| [15:18] <bac> paint() | |
| [15:19] <frankban> :-) | |
| [15:19] <gmb> There should be a law against people phoning me whilst I'm reviewing. | |
| [15:20] <gary_poster> bac, :-) | |
| [15:30] <gmb> frankban, I'm really sorry, I've been looking at this branch for ages and I can't find anything I don't like. | |
| [15:30] <gmb> r=me | |
| [15:30] <gmb> Apologies. | |
| [15:31] <frankban> gmb: what a shame :-) however thanks for trying | |
| [15:31] <gmb> :) | |
| [15:35] <benji> gary_poster: my connectivity looks like it might be back, but it's intermittent; I'm going to take an early lunch and hope that it's stable by the time I'm done | |
| [15:48] <gary_poster> benji ack good luck | |
| [16:20] <benji> gary_poster: I take it that I should re-implement our worker tagging test result on top of lp:~jml/testtools/forward-current-tags. Correct? | |
| [16:21] <gary_poster> benji, yes, that's my suggestion. Please feel free to argue otherwise, as usual | |
| [16:21] <benji> gary_poster: does that mean that it's usual for me to argue? ;P | |
| [16:22] <gary_poster> benji, :-P no | |
| [16:22] <benji> heh | |
| [16:22] <benji> that looks like a good approach; I'll see where it takes me | |
| [16:25] <gary_poster> cool thx | |
| [16:55] <gmb> gary_poster, It's just been pointed out to me that the first Monday of UDS is a UK national holiday. Is it okay with you if I take the coming Monday (16th) in lieu of it, since I'm going to be swimming the nerd sea on that day? | |
| [16:56] <gary_poster> gmb, sure, though what is the nerd sea? :-) | |
| [16:56] <gmb> gary_poster, The UDS lobby, of course :) | |
| [16:56] <gmb> (I hasten to add, it was Jono Bacon's name for it, not mine :)) | |
| [16:57] <gmb> gary_poster, Thank you. I'll add it in CanonicalAdmin now. | |
| [16:57] <gary_poster> lol gmb, cool | |
| [17:00] <gmb> Ahaha. | |
| [17:00] * gmb accidentally builds LP tree on Precise. | |
| [17:00] <gmb> Hilarity ensues. | |
| [17:00] <gary_poster> in order to work around an ec2 test problem with lightweight checkouts, I have to specify --pqm-public-location. Because I'm trying to have fun with sed, I end up with the general command... | |
| [17:00] <gary_poster> ec2 test --pqm-public-location=`bzr info | sed -e '/push branch/!d' -e '/push branch/s/\s*push branch: //'` | |
| [17:01] <gary_poster> evilness is fun | |
| [17:01] <gary_poster> so is making a noun into a noun | |
| [17:01] <gmb> Wow. | |
| [17:01] <gary_poster> evilnessitude might be even better | |
| [17:02] <gmb> gary_poster, That's disturbingly like something I would do :) | |
| [17:02] <gary_poster> heh | |
| [17:02] <gary_poster> ok, test started, sed played with... | |
| [17:02] <gary_poster> time for some lunch | |
| [17:03] * gmb knocks off for the evening. Night all. | |
| [17:05] <benji> night, gmb | |
| [17:27] <gary_poster> night gmb. approved Monday. | |
| [17:50] <bac> gary_poster, with an 8 cpu ec2 machine, should we expect a slave to come up within an hour? | |
| [17:51] <gary_poster> bac, give or take. two hours is usually too long. You should be able to look at the juju log on the slave instance to see the output of setuplxc | |
| [18:19] <gary_poster> bac, I self-approved this, but in case you are curious: https://code.launchpad.net/~gary/launchpad/trivial/+merge/101619 | |
| [18:19] <gary_poster> tests have been running for a couple of hours now. | |
| [18:20] <gary_poster> bac, I'd like to get a parallel test run or two or three to see how we are doing. Did you get one running? | |
| [18:21] <bac> memcache change looks good. thanks for the thorough documenting | |
| [18:21] <bac> gary_poster: i just started one a little bit ago and am waiting for the slave to come up | |
| [18:22] <bac> i think gmb's script is working now it was just too aggressive in only giving the slave 1 hour to come up | |
| [18:22] <gary_poster> bac, memcache: cool, thanks. waiting for slave: great. Please lemme know when tests have run: I'd like you to collect the output of testr failing after a test run. | |
| [18:22] <bac> gary_poster: will do | |
| [18:23] <gary_poster> thanks | |
| [18:23] <bac> sadly i don't think my run got your memcache fix | |
| [18:26] <gary_poster> no, it won't | |
| [18:57] <bac> gary_poster: my attempt failed, at least the automatic part. | |
| [18:57] <gary_poster> bac, :-/ how so? | |
| [18:57] <bac> while waiting for the slave to come up, a call to 'juju status' got | |
| [18:58] <bac> CalledProcessError: Command '['juju', 'status']' returned non-zero exit status 1 | |
| [18:58] <bac> 2012-04-11 14:47:17,862 INFO Connecting to environment... | |
| [18:58] <bac> 2012-04-11 14:47:19,807 INFO Connected to environment. | |
| [18:58] <bac> [('SSL routines', 'SSL3_READ_BYTES', 'ssl handshake failure')] | |
| [18:58] <bac> 2012-04-11 14:47:22,699 ERROR [('SSL routines', 'SSL3_READ_BYTES', 'ssl handshake failure')] | |
| [18:58] <bac> this is not uncommon | |
| [18:58] <bac> but, i can manually wait for the slave to come up and force the build by hand | |
| [19:00] <bac> it has been up 66 minutes so perhaps it'll be happy soon | |
| [19:02] <gary_poster> hm | |
| [19:02] <gary_poster> ok | |
| [19:14] <bac> ok, setuplxc is done. took about 80 minutes | |
| [19:19] <bac> gary_poster: you ever see juju status report 'relation-errors'? | |
| [19:19] <bac> relation-errors: | |
| [19:19] <bac> buildbot: | |
| [19:19] <bac> - buildbot-slave | |
| [19:20] <gary_poster> bac, rarely, yes | |
| [19:20] <bac> well it is my lucky day | |
| [19:20] <gary_poster> trying to think of something helpful.... | |
| [19:20] <gary_poster> bac, usually the logs are the place I go | |
| [19:21] <bac> yeah, i see some stuff in charm.log | |
| [19:22] <bac> nothing too informative, though | |
| [19:23] <gary_poster> bac, also try master's log | |
| [19:23] <gary_poster> /var/lib/buildbot/masters/master/twistd.log | |
| [19:25] <bac> hmm, at :06 the bb master received SIGTERM and shut itself down | |
| [19:25] <bac> i don't think the slave was even up yet | |
| [19:27] <bac> i was able to restart the master by hand | |
| [19:32] <bac> and manually started a slave. go juju go. | |
| [19:38] <bac> gary_poster: my hand starting stuff did work. it looks like the slaves started as the wrong user. | |
| [19:38] <bac> /usr/bin/xvfb-run: 181: /var/lib/buildbot/slaves/slave/lucid-devel/build/bin/test: Permission denied | |
| [19:39] <bac> which confuses me as the log shows: | |
| [19:39] <bac> sudo /usr/local/bin/launchpad-lxc-test -t stories/gpg | |
| [19:50] <gary_poster> bac, the permission denied with xvfb-run usually indicates a bad command passed to xvfb-run | |
| [19:50] <gary_poster> caused by crazy bash escaping | |
| [19:50] <gary_poster> we've had careful escaping to try and prevent this | |
| [19:50] <bac> gary_poster: such as where? | |
| [19:51] <gary_poster> so, the trigger for the problem is that we are trying to pass an argument with spaces to xvfb-run's -a argument IIRC | |
| [19:51] <bac> gary_poster: i'm using stock yellow charms. are you suggesting it is an error in one of the hooks? | |
| [19:52] <gary_poster> bac, no, the problem is arguably in ssh, but we are not fixing that today. :-P . Today's breakage of the workaround we implemented is in one of these: | |
| [19:52] <gary_poster> launchpad-lxc-test | |
| [19:52] <gary_poster> or lxc-start-ephemeral | |
| [19:52] <gary_poster> I don't think there would be another | |
| [19:54] <gary_poster> bac, to see a related behavior, change lxc-start-ephemeral to begin with set -ex (IIRC) | |
| [19:55] <gary_poster> that will spit out the commands lxc-start-ephemeral sends | |
| [19:55] <gary_poster> bac, could you give me the url of your master | |
| [20:00] <bac> gary_poster: looking | |
| [20:01] <bac> gary_poster: ec2-23-21-33-127.compute-1.amazonaws.com:8010 | |
| [20:17] <gary_poster> bac, ack. Maybe we should pair on getting a test run manually tomorrow. If we run into any problems, you can get a feel for where the bodies are buried | |
| [20:17] <bac> gary_poster: that's a good idea | |
| [20:17] <bac> if this instance is no good to you i'll tear it down now | |
| [20:21] <gary_poster> bac yeah tear down | |
| [20:21] <gary_poster> sorry for frustrations (and slow turn around; other stuff going on too) | |
| [20:22] <bac> ok | |
| [20:48] <bac> gary_poster: sean says he'll push a new py-memcached tonight | |
| [20:48] <gary_poster> bac, heh, ok cool | |
| [20:49] <bac> twas the right decision | |
| [20:49] <gary_poster> cool bac | |
| [20:49] <bac> besides we haven't seen it yet | |
| [20:53] <gary_poster> :-) true | |
| [20:54] <gary_poster> I also haven't seen my testresults yet... | |