| jimbaker | niemeyer, any preferences in terms of python tools for writing html? cheetah? | 00:06 |
|---|---|---|
| niemeyer | jimbaker: I would prefer something very simple | 00:09 |
| jimbaker | niemeyer, exactly. so please tell me what you prefer, and i will use it :) | 00:09 |
| niemeyer | jimbaker: If possible a single Python file we can embed in the project itself | 00:09 |
| niemeyer | jimbaker: Since our needs are pretty trivial.. just iterating over a list building a table with OK/FAILED + links | 00:10 |
| niemeyer | jimbaker: Do you know anything like that? | 00:11 |
| niemeyer | I'll get some food, biab | 00:12 |
| jimbaker | niemeyer, i don't know of an ultra simplistic tool. typically when it's framed like that, it tends to be done with just direct writes | 00:12 |
| SpamapS | OK/FAILED links? isn't that what jenkins does? | 00:16 |
| jimbaker | niemeyer, but in my experience that rarely works well. best to use a tool chain that ensures compliant output. i haven't used cheetah, but i know it's used by other people on the ubuntu server team | 00:16 |
| jimbaker | SpamapS, correct, we are re-implementing what jenkins does | 00:16 |
| jimbaker | SpamapS, regardless this is the task i have | 00:16 |
| SpamapS | eh? | 00:16 |
| jimbaker | for our functional testing | 00:17 |
| SpamapS | Right, subunit doesn't have something ? | 00:19 |
| SpamapS | I mean, jenkins is like, the industry standard. | 00:19 |
| SpamapS | And we have a.. really kick ass jenkins formula. :) | 00:19 |
| jimbaker | SpamapS, agreed that jenkins is the industry standard | 00:19 |
| SpamapS | https://wiki.jenkins-ci.org/display/JENKINS/NUnit+Plugin | 00:20 |
| SpamapS | I was kind of hoping we'd get to a point where we were graphing all that lovely test coverage | 00:20 |
| jimbaker | SpamapS, yes, output to standard nunitxml is pretty trivial. subunit supports this, through subunit2junitxml. however, the functional testing will not be done w/ trial | 00:21 |
| SpamapS | aww snap | 00:22 |
| jimbaker | SpamapS, this is actually a good thing. trial is pretty wretched for long running functional tests | 00:22 |
| jimbaker | SpamapS, niemeyer has written a new test runner that will execute shell scripts (this is the churn we discussed earlier on this channel) | 00:22 |
| SpamapS | Awright. Well, I don't really see us matching jenkins for frontend capabilities.. probably worth your time to produce nunint/junit xml and then just chug through it on jenkins. | 00:23 |
| jimbaker | SpamapS, however, it might make sense to output junitxml instead. perhaps that could be considered at a later time. | 00:23 |
| SpamapS | A new test runner that runs shell scripts? | 00:23 |
| SpamapS | Thats also what jenkins does. :) | 00:23 |
| SpamapS | Tho its not much of a tool for developers to run in an iterative sense.. | 00:24 |
| SpamapS | I didn't think the functional tests would be somethign you'd do in that manner. | 00:24 |
| SpamapS | Figured you'd just set jenkins jobs up to point at in progress branches and trunk and run tests on commit | 00:24 |
| jimbaker | SpamapS, that makes a lot of sense to do. again the test runner could readily output junitxml | 00:25 |
| jimbaker | SpamapS, the more important thing is that we want functional tests directly exercising ensemble commands | 00:26 |
| SpamapS | Hey, me too. :) | 00:27 |
| SpamapS | On all supported configurations | 00:27 |
| SpamapS | (that last bit is integration testing.. but .. details.. ;) | 00:28 |
| jimbaker | SpamapS, cool. so consider this as an intermediate step | 00:28 |
| jimbaker | trust me, it will be very easy to have this integrated into jenkins in some subsequent step | 00:28 |
| SpamapS | Alright, well yeah, I'd think the time spent outputting HTML would be better spent outputting junitxml that jenkins can read and digest. | 00:29 |
| SpamapS | On the tests.. | 00:29 |
| SpamapS | I wrote a few silly bash scripts that deploy/relate/query the formulas.. and james page came up with a cool way to have formulas embed tests ... | 00:29 |
| SpamapS | we should have a pow-wow once you've got the test runner doing some stuff. | 00:29 |
| jimbaker | SpamapS, sounds like a plan. i definitely want to hear more about james' ideas | 00:32 |
| jimbaker | SpamapS, we could readily add your bash scripts | 00:34 |
| SpamapS | jimbaker: I need to distill them down to smaller nuggets that can be encapsulated and run wherever.. one challenge is setting up the env w/ an environments.yaml and ssh key and such | 00:35 |
| jimbaker | SpamapS, yeah, that's a much more interesting problem. one thing that we can do w/ building out environments.yamls is do things like running several envs in parallel | 00:36 |
| jimbaker | SpamapS, i do have a variant of sshclient that ignores possible man-in-the-middle, but it would be much better if we managed this properly | 00:37 |
| jimbaker | SpamapS, another thing i did play with was making sshclient much more robust for extremely long waits. at the very least, we might want to have something like a wait-for-ensemble ENVIRONMENT which successively looks at the milestones in an env coming up | 00:39 |
| jimbaker | (it might also be a useful ensemble subcommand in the future, don't know) | 00:40 |
| SpamapS | Not sure what you mean | 00:43 |
| SpamapS | I wrote a wait4state perl script that just runs status | 00:43 |
| niemeyer | jimbaker: This looks nice: https://github.com/defunkt/pystache | 00:43 |
| SpamapS | but /win 10 | 00:45 |
| SpamapS | doh | 00:45 |
| jimbaker | niemeyer, looks good to me, thanks! | 00:45 |
| jimbaker | SpamapS, but /win... ? | 00:46 |
| SpamapS | irssi | 00:46 |
| SpamapS | hey quick python question | 00:46 |
| SpamapS | do lists keep an internal length? | 00:46 |
| niemeyer | jimbaker: It's an incarnation of this: http://mustache.github.com/ | 00:46 |
| SpamapS | I'm profiling something that does len(giant_list) a lit and it seems to me that removing these calls makes it a lot faster (and just keeping track of the length in a different variable) | 00:47 |
| niemeyer | SpamapS: They do | 00:47 |
| niemeyer | SpamapS: Are you doing tons of those calls? | 00:47 |
| SpamapS | yes | 00:47 |
| niemeyer | SpamapS: function calls in Python are expensive in general.. you may just be seeing that cost | 00:48 |
| jimbaker | niemeyer, yeah, mustache is a fine approach | 00:48 |
| SpamapS | 5 for every line of a 500,000 line file.... | 00:48 |
| SpamapS | So I'm not crazy.. its not a massive expense per call.. but it adds up | 00:48 |
| jimbaker | SpamapS, doesn't seem to be so much even for cpython | 00:48 |
| niemeyer | SpamapS: Yeah, definitely.. you'd notice 2.5M Python function calls for sure | 00:48 |
| SpamapS | http://paste.ubuntu.com/680217/ | 00:49 |
| jimbaker | so there, we have two contrasting viewpoints | 00:49 |
| SpamapS | now to find out where the other 8.8s are spent. :-P | 00:49 |
| niemeyer | jimbaker: You mean you don't find them expensive? | 00:50 |
| jimbaker | niemeyer, i just don't think it's that bad if it's just 2.5M | 00:51 |
| niemeyer | jimbaker: | 00:52 |
| jimbaker | niemeyer, functional call overhead in cpython is obviously quite expensive, for a variety of reasons | 00:52 |
| niemeyer | def f(): pass | 00:52 |
| niemeyer | start = time() | 00:52 |
| niemeyer | for i in range(2500000): f() | 00:52 |
| niemeyer | print time()-start | 00:52 |
| niemeyer | jimbaker: This takes 1 second in my machine | 00:52 |
| niemeyer | jimbaker: Doing absolutely nothing | 00:52 |
| niemeyer | jimbaker: There's more than a function call there, but gives an idea | 00:52 |
| jimbaker | niemeyer, i'm sure that's the case :) | 00:52 |
| niemeyer | jimbaker: That's what I meant by "you'd notice" | 00:53 |
| jimbaker | it's just that 0.5s as the case w/ SpamapS's perf or 1s for your case is still a relatively small number compared to what i have seen in the past, that's all ;) | 00:53 |
| jimbaker | but if run in a command, or in the direct path for rendering a page, sure, it's extremely noticeable | 00:54 |
| SpamapS | That, and things like .append, .sort, etc, are the only function calls.. | 00:54 |
| SpamapS | this script lags the same algorithm in perl by about 15% | 00:54 |
| SpamapS | I was wrong btw, it doesn't do 5 for every line | 00:54 |
| SpamapS | just 1 len for every line | 00:54 |
| SpamapS | but yeah, 500,000 is 0.504s | 00:55 |
| jimbaker | SpamapS, did you try it w/ pypy? that would be an interesting exercise | 00:55 |
| SpamapS | jimbaker: I wanted to but didn't know how to even get pypy ;) | 00:55 |
| jimbaker | functional call overhead w/ pypy is extremely reduced | 00:55 |
| niemeyer | Yeah, should kick ass | 00:55 |
| SpamapS | lol.. lucid, karmic, and jaunty.. :-P | 00:56 |
| jimbaker | http://pypy.org/download.html - just give it a try, i don't know where they are on latency | 00:56 |
| niemeyer | Takes 16ms in Go.. | 00:56 |
| niemeyer | The 2.5 | 00:56 |
| jimbaker | niemeyer, that might be close to pypy for the empty case, factoring out startup. which may or may not be applicable, given the domain | 00:57 |
| jimbaker | (i too should give it a try ;) ) | 00:58 |
| niemeyer | Maybe.. haven't tried that kind of thing there yet.. would be curious, though | 00:58 |
| niemeyer | Given their latest work it must be pretty amazing | 00:58 |
| jimbaker | bbl | 01:01 |
| SpamapS | pretty hard to find pypy binaries.. :-P | 01:01 |
| _mup_ | Bug #838568 was filed: push-review must be tested again <Ensemble:In Progress by niemeyer> < https://launchpad.net/bugs/838568 > | 01:54 |
| === daker_ is now known as daker | ||
| niemeyer | Alow! | 13:57 |
| hazmat | jimbaker`, jinja2 is pretty nice re templating engines | 14:16 |
| _mup_ | ensemble/local-provider r353 committed by kapil.thangavelu@canonical.com | 14:24 |
| _mup_ | remove utility wait_for_node, just make the nesc. method public on zookeeperconnect, allow get open port to take a host option. | 14:24 |
| jimbaker` | hazmat, jinja2 is fine. but let's go with mustache | 15:00 |
| hazmat | jimbaker`, wasn't clear that mustache has any advantage over say.. string.Template | 15:01 |
| hazmat | er.. actually string formatters | 15:05 |
| hazmat | mustache looks fine though | 15:06 |
| jimbaker` | hazmat, mustache does one very important thing, which is that it performs html encoding | 15:08 |
| jimbaker` | when compared to string.Template. obviously the mustache {{}} itself is familiar in both templating systems | 15:09 |
| niemeyer | jimbaker`: and loops, I guess | 15:10 |
| niemeyer | Lunch time.. biab! | 15:53 |
| SpamapS | niemeyer: can we get push-review from a PPA or something? | 16:14 |
| SpamapS | niemeyer: I *really* like the automation it promises. :) | 16:14 |
| _mup_ | Bug #839695 was filed: ensemble.formula.tests.test_bundle.BundleTest.test_executable_extraction fails on empty home dir <Ensemble:New> < https://launchpad.net/bugs/839695 > | 16:21 |
| SpamapS | hazmat: re the bug linked above.. | 16:27 |
| SpamapS | hazmat: its checking for 493 as the stat.. but.. in my bzr co of ensemble, its 509 .. I believe this is because umask has changed on ubuntu.. | 16:28 |
| SpamapS | but I could be wrong | 16:28 |
| SpamapS | indeed.. its just a matter of how you checked out the branch | 16:29 |
| SpamapS | I think the appropriate check is that the stat *matches* the sample.. not that it is exactly 509.. basically we want to make sure exec was preserved | 16:33 |
| hazmat | SpamapS, sounds good | 16:41 |
| SpamapS | hazmat: http://pastebin.com/DaaJ1Mp0 | 16:45 |
| SpamapS | that is currently building in my PPA.. if it passes all tests, I'll change debian/rules to fail the build on test fail and submit a FFe for revision 336 to upload to Ubuntu. | 16:46 |
| SpamapS | (with the 3 patches I've submitted for review applied.. ;) | 16:47 |
| niemeyer | SpamapS: Absolutely! Working on that already.. should be apt-gettable today | 16:55 |
| jcastro | hey m_3 | 18:16 |
| jcastro | http://blog.carbonfive.com/2011/09/01/deploying-node-js-on-amazon-ec2/ | 18:16 |
| _mup_ | Bug #839794 was filed: Final test for the tool! <Ensemble:In Progress by niemeyer> < https://launchpad.net/bugs/839794 > | 18:17 |
| jcastro | would now be a good time to ask how the node standalone formula is coming along? | 18:17 |
| jimbaker` | jcastro, i like how that blog post begins: "After nearly a month of beating my head against the wall that is hosted node.js stacks — with their fake beta invites and non-existent support..." | 18:45 |
| m_3 | jcastro: hey | 18:46 |
| m_3 | jcastro: yeah, the node formula was part of a talk I gave on IRC a few weeks ago | 18:47 |
| m_3 | jcastro: I'll brush it off, promote it to principia, and blog about it | 18:47 |
| jcastro | m_3: do you have time to post a response to the guy? | 18:47 |
| jcastro | it reads to me like he needs ensemble hard core | 18:47 |
| m_3 | jcastro: yeah, I'll do it during a boring talk this afternoon | 18:49 |
| jcastro | oh sorry, forgot you were travelling | 18:49 |
| m_3 | jcastro: conf wraps up today, so I'll be able to put time on it over the weekend | 18:50 |
| === niemeyer_ is now known as niemeyer | ||
| === daker is now known as daker_ | ||
| negronjl | jcastro: ping | 19:51 |
| jcastro | negronjl: pong | 19:55 |
| negronjl | jcastro: in a meeting right now.. I'll ping you later | 19:56 |
| * jcastro nods | 19:56 | |
| _mup_ | ensemble/local-unit-deploy r354 committed by kapil.thangavelu@canonical.com | 19:56 |
| _mup_ | factor out unit environment retrieval into separate method. | 19:56 |
| _mup_ | ensemble/local-unit-deploy r355 committed by kapil.thangavelu@canonical.com | 20:14 |
| _mup_ | user/environment qualified lxc container name, tweak upstart job | 20:14 |
| _mup_ | ensemble/local-unit-deploy r356 committed by kapil.thangavelu@canonical.com | 20:28 |
| _mup_ | additional test for qualified container name | 20:28 |
| hazmat | SpamapS, does upstart do env expansion on the exec stanza? | 21:01 |
| SpamapS | hazmat: yes | 21:11 |
| niemeyer | Okay.. lpad is out, lbox is out.. back to reviews and porting formulas | 21:24 |
| niemeyer | But I'll step out for now.. may do some extra time tomorrow for helping on the queue | 21:25 |
| * niemeyer bbl | 21:26 | |
| _mup_ | ensemble/local-unit-deploy r357 committed by kapil.thangavelu@canonical.com | 21:33 |
| _mup_ | have upstart respawn and run the container in the foreground, doc strings for all container deployment methods. | 21:33 |
| * robbiew needs a drink | 21:36 | |
| hazmat | excellent idea | 21:44 |
| SpamapS | hazmat: upstart for containers, but not for the agents?! | 21:49 |
| hazmat | SpamapS, for the agent | 21:51 |
| hazmat | unit agent only | 21:51 |
| hazmat | inside a container | 21:51 |
| hazmat | its not truly restartable.. probably shouldnt' be respawn auto | 21:52 |
| * hazmat returns with a bloody mary | 21:52 | |
| hazmat | SpamapS, one nice fallout of the local dev stuff, it should be pretty easy to use this for arm servers, with an external zk address. | 21:53 |
| SpamapS | task tsk.. we need our idempotency. :) | 21:53 |
| SpamapS | :) | 21:53 |
| * hazmat reruns tests to ascertain potency | 21:53 | |
| SpamapS | oh I thought you were saying that its not re-startable because of some problem with idempotency | 21:56 |
| hazmat | SpamapS, the real issue is loss of events while the agent was down, in practice its probably not an issue, except when it is | 22:09 |
| SpamapS | err.. events? | 22:10 |
| hazmat | no good reason not to go ahead with it i suppose | 22:10 |
| SpamapS | I thought we just watched to see what the state was | 22:10 |
| SpamapS | if we miss 4 changes, who cares, just make the new state true. | 22:10 |
| hazmat | SpamapS, if the agent is down, watches won't be delivered | 22:10 |
| hazmat | and extant watches need to be restablished as new watches potentially | 22:11 |
| SpamapS | if we need that, we need a queue, not a data store | 22:11 |
| SpamapS | But I see what you're saying | 22:11 |
| hazmat | we're delivering a queue/event interface to formulas, but disconnections mean establishing a delta between state on reconnect to inform the formula of the events between the watch restablishment | 22:12 |
| SpamapS | what we need is to keep track of what state the agent thinks it is in, so when it comes back up it can do the appropriate transitions. | 22:12 |
| SpamapS | Thats not a queue.. thats a watched data store. | 22:12 |
| hazmat | yup | 22:12 |
| SpamapS | Convenient, yes, but not a queue. :) | 22:12 |
| hazmat | its really a sync between a remote and local data store | 22:12 |
| hazmat | to compute the delta | 22:12 |
| SpamapS | I see the problem.. very interesting | 22:13 |
| hazmat | in truth its probably a very rare issue in practice for a respawn | 22:13 |
| hazmat | for a net split its more significant | 22:13 |
| hazmat | or any long running outage | 22:13 |
| SpamapS | yeah, so at the moment, ensemble is not partition tolerant. ;) | 22:13 |
| * SpamapS disappears | 22:16 | |
| hazmat | SpamapS, have a good weekend | 22:16 |
| jimbaker` | how do i push a branch that's a branch of lp:~ensemble/ensemble/ftests ? i tried this, but it's not right: bzr push --stacked-on lp:~ensemble/ensemble/ftests --remember lp:~jimbaker/ensemble/ftests/generate-html | 22:55 |
| jimbaker` | niemeyer, ^^^ ? | 23:02 |
| _mup_ | ensemble/local-unit-deploy r358 committed by kapil.thangavelu@canonical.com | 23:09 |
| _mup_ | use ensemble home instead of units directory for all unit deployments | 23:09 |
| hazmat | jimbaker`, i think you just push to ~jimbaker/ensemble/generate-html | 23:09 |
| hazmat | its on the merge side you pick the series branch afaik | 23:10 |
| * hazmat checks | 23:10 | |
| jimbaker` | hazmat, ok, i was wondering why it was wanting to use the wrong series | 23:10 |
| hazmat | yup | 23:10 |
| hazmat | jimbaker`, so on merge the target branch is ensemble/ftests | 23:11 |
| hazmat | there all branches of the same project for the user project namespace | 23:11 |
| hazmat | s/their | 23:11 |
| _mup_ | Bug #839969 was filed: Checkout Ensemble branch, run bash ftests, and generate summary in HTML <Ensemble:In Progress by jimbaker> < https://launchpad.net/bugs/839969 > | 23:15 |
| hazmat | jimbaker`, is the ftest stuff using jenkins for test running? | 23:18 |
| jimbaker` | hazmat, no | 23:18 |
| hazmat | just wondering if we could submit tests for branches in dev | 23:18 |
| hazmat | via an api | 23:18 |
| hazmat | w/ the security integration tests get about 30% slower out of the box on average... | 23:19 |
| jimbaker` | tests will get merged as normal i would think. at some point, we will have jenkins integration, it would be simple/trivial to do | 23:19 |
| hazmat | jimbaker`, i guess i'm confused what's running the tests? your not using jenkins or buildbot? | 23:19 |
| jimbaker` | i'm using two new tools: butler.py, which i just wrote; and niemeyer's churn, which executes bash scripts | 23:21 |
| jimbaker` | a butler used to manage the buttery, which might store the results of the churning... | 23:21 |
| jimbaker` | and of course jenkins is presumably a butler. whatever ;) | 23:22 |
| hazmat | not sure i understand why.. but okay | 23:22 |
| jimbaker` | hazmat, at some point jenkins is presumed. twisted trial however is definitely not going to be used | 23:22 |
| jimbaker` | the reason is that we need something that is focused on running ensemble commands | 23:23 |
| hazmat | jimbaker`, i assume the test runner would be doing normal tests and ftests | 23:23 |
| hazmat | is that not accurate? | 23:23 |
| jimbaker` | that's not accurate | 23:23 |
| jimbaker` | it's only running ftests, which are to be only be written using bash | 23:23 |
| jimbaker` | that's niemeyer's decision, and it makes sense | 23:24 |
| hazmat | from a purpose of ease of test dev for users probably | 23:24 |
| jimbaker` | such test scripts could potentially use resources that are written in python, eg to validate results or to wait for resources | 23:24 |
| hazmat | still i seems like conflating issues, an ftest suite and test suite, and the test bot | 23:25 |
| jimbaker` | hazmat, perhaps, but it's hopefully what was requested by niemeyer. again i think it's pretty simple, and it will integrate well w/ jenkins eventually, just need to emit junitxml instead of the summary html file | 23:26 |
| hazmat | jimbaker`, fair enough.. we need a test runner for the ftests that does reports | 23:26 |
| jimbaker` | and that's the merge proposal :) | 23:27 |
| hazmat | bcsaller, do you have a moment to catch up? | 23:32 |
| bcsaller | hazmat: sure do | 23:33 |
| hazmat | bcsaller, ~hazmat/ensemble/local-unit-deploy is the latest in the pipeline | 23:47 |
| bcsaller | hazmat: thanks | 23:47 |
| niemeyer | jimbaker`: lbox propose --for lp:ensemble/ftests | 23:59 |
| niemeyer | Sorry, single dash | 23:59 |
| niemeyer | jimbaker`: lbox propose -for lp:ensemble/ftests | 23:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!