IRC logs for #baserock for Thursday, 2015-06-25

*** zoli__ has joined #baserock03:40
*** zoli__ has quit IRC04:09
*** zoli__ has joined #baserock04:20
*** petefoth has joined #baserock05:39
*** zoli__ has quit IRC06:23
*** rdale has quit IRC07:04
*** zoli__ has joined #baserock07:06
*** zoli___ has joined #baserock07:08
*** zoli__ has quit IRC07:08
*** a1exhughe5 has joined #baserock07:22
*** mariaderidder has joined #baserock07:46
*** fay_ has joined #baserock07:53
*** bashrc_ has joined #baserock08:02
*** ssam2 has joined #baserock08:05
*** ChanServ sets mode: +v ssam208:05
* pedroalvarez thinks the performance of Mason is not acceptable08:07
pedroalvarezIt used to be08:07
rjekpedroalvarez: In what way?08:08
pedroalvarezIt seems too slow to me08:08
*** jonathanmaw has joined #baserock08:10
rjekpedroalvarez: What is too slow?  The builds?08:11
* rjek has noticed it taking longer and longer to build Baserock on the few recent occations he's done it08:11
rjekperhaps we need finer-grained strata to allow for more parallelisation.08:12
SotKhow many workers does its distbuild network have?08:12
SotKif it only has one, then there will be no parallelisation at all08:12
rjekIn that case, slowness would not be mason's fault :)08:14
rjekOr it would be mason's fault, depending on how you look at it.08:14
rjekIdeally for CI you want answers in minutes08:15
pedroalvarezrjek: the builds08:15
pedroalvarezI'm just saying that a local build in my devel machine would take less.08:15
SotKlooking at the log, it only seems to be using one worker even if it has more than one08:16
pedroalvarezi might have more ram and cpu though08:16
ssam2jmac had a devel system in this cloud recently and decided that a VM in his laptop was much faster08:16
pedroalvarezit only has one worker, so that's fine08:16
rjekpedroalvarez: Enterprise™ Cloud™ in being slow unshocker :)08:16
ssam2it only has one worker08:16
rjekISTR the IO being pretty poor there.08:16
SotKhow much ram and cpu does it have?08:18
*** rdale has joined #baserock08:18
ssam22GB ram, 2 VCPUs08:18
SotKI guess that explains why builds are faster on people's devel machines too then08:19
rjekThat seems surprisingly puny.08:19
rjek Timing buffered disk reads:  36 MB in  3.10 seconds =  11.61 MB/sec08:20
rjekThat won't be helping, either.08:20
tlsamason used to build 1 system, now it builds 4.  Also more stuff has been added to the systems.  I think that's why its got slower08:25
radiofreehow long is it taking to build a devel system these days08:26
tlsaradiofree: it builds all 4 at once, I think so you can't tell.  Also it depends on how much is reusable from the last build08:27
tlsaalso, it always took longer than a local build because of the devel system test.  Dunno if it still does that, but it used to deploy the devel system after it built, and then get the deployed devel system to do a build08:29
*** gary_perkins has joined #baserock08:31
pedroalvarezno, sadly mason in has never done the devel system test08:31
tlsaoh, right08:32
perrylargh...i kept testing a distbuild and wondering why it wasn't catching an error i'd purposely thrown in; then i realise /usr/bin/morph isn't set to use the version i'm working on...i need more coffee09:04
pedroalvarezperryl: I've been testing distbuild recently using `scripts/run-distbuild`09:06
pedroalvarezI do recommend it09:06
perryl:D i'm using it myself, it's very handy!09:06
pedroalvarezit is09:07
perrylor...i was using it, now that i've changed morph to use my working version i keep getting a "permission denied" error when i try and run it :(09:07
perrylmaybe if i also remember to set permissions for my new morph to executable it might :)09:08
SotKperryl: don't forget to do the same for /usr/bin/distbuild-helper :)09:09
perrylSotK: set permissions?09:11
*** edcragg has joined #baserock09:11
* perryl debates going back to bed and starting the day over again09:11
SotKset permissions and make it use the version you're working on09:12
perrylcool! i hadn't thought to change that one :)09:12
* SotK has lost too much time forgetting that one in the past :)09:15
*** lachlanmackenzie has joined #baserock09:22
*** a1ex has joined #baserock10:02
*** a1exhughe5 has quit IRC10:06
*** zoli___ has quit IRC10:48
*** pacon has joined #baserock10:56
perrylhmm, it appears that in order to create a database via sqlite3.connect you need to first create the directory that the database resides in10:59
rjekThat wouldn't surprise me; SQLite's written mostly in ANSI C, which is directory-ignorant.10:59
perryli wouldn't have considered it except the python docs mentioned sqlite3.connect('example.db') without a path, which got me thinking11:00
rjekah :)11:00
rjeksqlite probably passes that string directly to fopen()11:01
perrylbit annoying that there's no given error message for that case; it just hangs indefinitely :(11:04
* perryl wonders if python debugger may have helped her solve this quicker11:04
rjekIs there a mkdir -p function in Python's stdlib you could call before hand?11:04
perrylthere's os.mkdir but i don't know if it allows multilevel directory creation11:06
perrylit'd be good if there was, rather than a million "if os.path.exists(..........)" checks11:07
CTtpollardperryl: os.makedirs?11:08
perrylCTtpollard: just seen that, i'll give it a run through! :)11:09
*** zoli__ has joined #baserock11:13
perrylwhoo, that works! thanks all!11:15
*** a1ex has quit IRC11:19
*** a1exhughe5 has joined #baserock11:22
*** pacon has quit IRC12:18
ssam2it turns out CloudFoundry is pretty cool. it's like a free software version of Heroku12:22
ssam2i managed to set up Etherpad-lite on Pivotal's private Cloud Foundry thing in less than half an hour12:23
ssam2the downside is it's a 60 free trial only. but the upside is that perhaps in 60 days we can have figured out how to set up our own Cloud Foundry ;)12:24
ssam2*60 day free trial12:25
paulsher1oodssam2: w00t! next step, running on Baserock? :)12:48
*** paulsher1ood is now known as paulsherwood13:46
paulsherwoodpedroalvarez: in terms of hardware, what do folks need at a minimum to try out a baserock openstack deployment?14:16
pedroalvarez6G ram, a couple of cores, and you will be able to instanciate a vm with 512M of ram14:17
pedroalvarezmore VMs == more resources14:18
rjekpedroalvarez: Surely you can't be saying the OpenStack overhead is 5.5GB?14:18
pedroalvarezno, I think that 3G *might* be enough.. but I believe it starts crasing if you open horizon14:19
paulsherwoodpedroalvarez: what if i wanted a cluster? can one machine do that?14:19
* paulsherwood notices that the only machine he can get onto with enough memory may be a Moonshot... 14:20
richard_mawpaulsherwood: by my understanding of the definition of cluster, you can't by definition.14:20
paulsherwoodrichard_maw: you may be right :)14:20
pedroalvarezpaulsherwood: are you interested on testing openstack deployments?14:22
pedroalvarezI might have some hardware you can borrow14:22
paulsherwoodpedroalvarez: yup!14:36
pedroalvarezpaulsherwood: any experience deploying to kvm?14:41
paulsherwoodhow hard can it be, really? :)14:53
pedroalvarezwell, deploying is easy, but I don't know how hard is going to be to see the kvm host for a mac user14:55
pedroalvarezpaulsherwood: you can follow this video, and I can grant you access to the same machine I was using for deployments
paulsherwoodok, cool! i'd like to try it the ybd way if possible :)14:58
*** a1exhughe5 has quit IRC15:43
rdalemaster seems to have gone from binutils-redhat to binutils-tarball, which is a step backwards15:48
rdale^is a^seems to be15:48
* paulsherwood wonders what was the reason for that change?15:48
pedroalvarezrdale: why?15:48
rdalethere's no git history and only one version15:49
pedroalvarezthe reason for that change was that we were building binnutils from a git repo, but with all the contents of a tarball commited15:49
pedroalvareznow is clearer that we are using a tarball15:49
paulsherwoodso it's the same thing?15:49
rdaleah i see15:49
paulsherwoodwe need at some point to make lorry deal with shallow clones15:49
paulsherwoodis there a story for that already? would be nice to be able to drop these several tarball things15:50
pedroalvarezpaulsherwood: maybe, but this time, using a tarball is not related to the size of the git repo15:50
paulsherwoodoh, really? why then?15:53
Kinnisonbecause some projects have a large delta between a release tag and a dist tarball15:54
*** zoli__ has quit IRC15:54
pedroalvarezthe main reason to use tarballs in build essential is that they need many dependencies to build from git that are not present when building build essential chunks15:55
pedroalvarez(correct me if I'm wrong)15:56
tiagogomes_you are not wrong15:57
rdaleah i see - seems a bit like the language bootstrapping issues discussed yesterday16:01
*** mariaderidder has quit IRC16:28
*** jonathanmaw has quit IRC16:36
*** ssam2 has quit IRC16:59
*** bashrc_ has quit IRC17:02
*** edcragg has quit IRC17:05
*** zoli__ has joined #baserock17:55
*** lachlanmackenzie has quit IRC18:16
*** gary_perkins has quit IRC18:54
*** zoli__ has quit IRC19:04
*** edcragg has joined #baserock19:20
*** pacon has joined #baserock22:01
*** pacon has quit IRC23:23
*** pacon has joined #baserock23:24
*** pacon has quit IRC23:26
*** edcragg has quit IRC23:48

Generated by 2.15.3 by Marius Gedminas - find it at!