IRC logs for #baserock for Thursday, 2014-11-27

*** cosm [] has quit [Ping timeout: 258 seconds]00:15
*** cosm [] has joined #baserock00:31
*** petefoth [] has joined #baserock07:23
*** zoli_ [] has joined #baserock07:35
*** zoli_ [] has quit [Changing host]07:35
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock07:35
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]08:30
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock08:51
*** mariaderidder [] has joined #baserock08:57
*** mariaderidder [] has quit []08:58
*** bashrc [] has joined #baserock09:01
*** tiagogomes [~tiagogome@] has joined #baserock09:02
*** jonathanmaw [] has joined #baserock09:40
*** Krin [] has joined #baserock10:01
*** petefoth_ [] has joined #baserock11:02
*** petefoth [] has quit [Ping timeout: 245 seconds]11:05
petefoth_ is now known as petefoth11:05
*** rdale [] has quit [Ping timeout: 265 seconds]11:48
*** wdutch [] has quit [Quit: Quit]11:52
*** wdutch [] has joined #baserock11:55
*** mariaderidder [] has joined #baserock12:24
*** locallycompact [] has joined #baserock12:37
perrylout of curiosity, how long does the build-deploy cycle take to complete?14:08
perrylor is that like asking 'how long is a piece of string'?14:09
persiaA bit like that, yes.14:09
persiaIf you're working on an edge component, it can be a matter of minutes.14:09
richard_mawit is really, it depends on how the size of the system you want to build, how much has been previously built, and how large you want to make disk images if you're deploying those14:09
persiaIf you're modifying something deep, like experimenting with a new C compiler, and invalidating all your caches, it can be hours or days.14:10
persia(e.g. moving from gcc/glibc to clang/musl)14:10
paulsher1oodpersia: not days, normally. full rebuild took ~ 10 hours for me last time i tried it14:11
richard_mawfor a dinky little ARM board it would be14:11
persiapaulsher1ood: I've taken > 3 days to build a devel system, and had runs I interrupted that ran nearly that long and remained incomplete.14:11
paulsher1oodpersia: ok i believe you  :)14:12
persiaNote that some of the slowness bugs I had have been fixed.  In practice, I suspect you are closer to the right maximum14:12
paulsher1oodat the other extreme, on my macbook, my normal build-and-deploy-system-with-new-kernel-to-self takes 10-15 minutes14:12
persiaYou could probably be even faster if you were to do build-and-deploy-to-self with a leaf python app.14:14
paulsher1oodwell yes, most of the time is building the kernel14:14
straycatdeploy is faster now at least14:18
paulsher1oodyup. what happened there?14:19
jmacspdar: Have there been any changes to ceph in Baserock recently? I'm getting an error saying the boost library is missing14:19
pdarOh yes, are you using a new master?14:20
straycatliw's fix to not transmit zeros iirc14:20
jmacspdar: Maybe - I'm working on an odd branch atm14:22
jmacsThis is why I need a tool to say "tell me the origin of this component" :)14:22
straycatso, now we just need to rewrite everything in k, and we'll be fine, style guide here:
paulsher1oodjmacs: what do you mean by 'origin'? 14:23
pdarI think radiofree said it might be something to do with the change to glib14:24
jmacsI'm not sure; details on where the source came from, where it's been cached, and the git ref14:24
persiaHow is this different from repo/ref?14:25
radiofreepdar: did building the newer version work?14:25
pdarradiofree: it worked great yep thanks :)14:26
pdarjmacs: if you change to the most recent version of boost-tarball v1.56 the boost libraries should compile fine14:49
pdarjmacs: but, this stops them form working with the old version of ceph thats set in defenitions14:50
*** locallycompact [] has quit [Read error: Connection reset by peer]14:51
jmacsHow do I do that? Can I map from boost version numbers to a ref to put in ceph-service.morph easily?14:53
petefoththe virtualbox-ssh write extension looks for carious network related parameters (HOST_IPADDR, NETMASK, NETWORK_CONFIG. can these be passed in in the samre way as others, i.e. in the clkuster definitio file or as key/value pairs of arguments in the call to `morph deploy’? 14:57
paulsher1oodjmacs: would a quick tutorial help?15:00
pdarjmacs: if you take the commit id of the boost v1.56 at and swap that for the one in ceph-service.morph15:01
* paulsher1ood assumes jmacs has seen the kernel update tutorial
paulsher1oodyou can put a short sha, branch, or tag in the ref: field15:02
jmacsYes, I was wondering how pdar found that page just given the information "v1.56"15:02
straycatpetefoth, any of the env vars can be passed on the command line to morph deploy15:03
jmacsSo "boost_1_56_0" would work as a ref15:04
jmacspaulsher1ood: No, I haven't seen that video, I'll watch it now.15:04
petefothstraycat: thanks. Are they passed as network config stanzas (as decribd in simple-network.configure) or is some other form?15:04
jmacsGot it anyway. Click on the tag, then on the "tagged object" field.15:05
straycatpetefoth, not sure i understand, the vars themselves are just strings interpreted by the extension15:07
petefothstraycat: Yes I understand that (I think :). I’m particulalry flummoxed by the NETWORK_CONFIG variable: I suspect - from reading  simple-network.configure - that it is more than a simple string: possibly its is one or more network config stanzas as handled in the `parse_network_stanza` function. I need to describe - in the file - how these stanzas get passed to the virtualbox-ssh write extension (not the15:22
petefothsimple-network.configure extension)15:22
pedroalvareznote, simple-network.configure extension is not being used by any of the systems15:26
pedroalvarezit was removed from all the systems with systemds15:26
KinnisonYes, which is sad15:27
petefothpedroalvarez: noted thanks! I wont bother documenting that one just yet then :) I guess that virtualbox-ssh *does* get used sometimes (or does everyone here use KVM or OpenStack?15:27
Kinnisonpetefoth: You're going to have to ask jjardon for help with how to write up what simple-network.configure *used* to be used for15:27
pedroalvarezpetefoth: yeah, I believe thet some people use virtualbox-ssh15:28
* SotK uses virtualbox at home but has never done a virtualbox-ssh deployment15:29
jmacsI've been using virtualbox-ssh extensively15:29
petefothKinnison: OK :) But I still need to know how to setup networking using the vbox-ssh wirte extension. 15:29
* paulsher1ood uses virtualbox-ssh exclusively15:29
Kinnisonpetefoth: You'll need to have jjardon help you and then you can document it15:30
Kinnisonpetefoth: when he did the systemd networking stuff he dropped simple-network.configure and made apparently no moves to document the replacement15:30
KinnisonI could be missing something though15:30
petefothjmacs: please can you talk (or email or irc me through the vbox-ssh stuff)15:30
pedroalvarezis needed to configure the network when doing a vbox-ssh deployment?15:31
pedroalvarezis mandatory?15:31
petefothKinnison: you’re missing that I was using simple-network as a (probably confusing) example. What I care about at the moment is the vbox-ssh networking stuff15:31
KinnisonAt the time *I* wrote the virtualbox networking stuff, it was using simple-network.configure15:32
petefothpedroalvarez: I don’t know, and I can’t work it out from looking at the code! That’s why I’m shouting ‘Heeeeeelp’ :)15:32
pedroalvarezyou don't need to document the networking stuff to document the vbox deployment15:32
paulsher1oodpetefoth: i assume you've seen
paulsher1ood(which has an example)15:33
petefothpaulsher1ood: thanks - I should hav elooked there instead of looking through the code :)15:34
jmacspaulsher1ood: I will only have copied NETWORK_CONFIG from somewhere - I'll try and remember where15:34
* pedroalvarez realises that in every example of vbox deployment we configure the network, and wonders if it's something needed in virtualbox15:34
* Kinnison notes that the virtualbox-ssh deployment extension expects NETWORK_CONFIG, expects it to define eth0 and eth1, and expects that one will be attached to a host-only network15:34
petefothIf that still works I can use that for the file15:34
* paulsher1ood wonders who wrote 'ditch the verbosity and just do it wi with a script'15:34
jmacspetefoth: I'm just using the example from
jmacsWith a modified static address for eth015:35
petefothjmacs: thanks that’ll be useful too. 15:35
paulsher1oodthe 'ditch verbosity' is more verbose than the no-frills page15:35
pedroalvarezaaaaaah! NETWORK_CONFIG  is also present in the vbox write extension, not only in the simple-network configure extension15:35
petefothpedroalvarez: exactly!15:36
pedroalvarezsorry pete15:36
petefothpedroalvarez: NP! stupid questions and vaguely enlightened answers all work together to get me to the right place in the end :)15:37
* petefoth notes that the correct answer to his question was ‘Read the wiki, dummy!’ :)15:38
pedroalvarezafter a quick look to the vbox write extension, I think it may be broken15:38
petefothpedroalvarez: not *my* problem - I’m just doucmenting how it is *meant* to work15:39
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]15:39
* paulsher1ood wonders who Robert is, and what made him decide to write when exists15:39
pedroalvarezpaulsher1ood: not sure, but I think that it was bashrc >15:44
* petefoth looks in his irc logs15:44
paulsher1oodmaybe, maybe not15:44
*** genii [~quassel@ubuntu/member/genii] has joined #baserock15:46
paulsher1oodbashrc: assuming you wrote the vm-script page, i don't see what it gives over the no-frills page, which pre-dates vm-script?15:47
* paulsher1ood looks harder15:47
bashrcI think there's more in the vm-script, but I'll have to check. I was trying to get to a state where I could compile zfs15:48
bashrcbut if they're very similar then just merge them into one15:49
paulsher1oodvm-script assumes presence of emacs, debian/ubuntu i think15:49
paulsher1oodbut i agree with the idea of having a script to do all of this15:49
paulsher1oodbashrc: i'll think a bit more15:50
bashrcmy impression is that the barrier to entry is pretty high. You have to jump around various pages to get something running15:51
paulsher1oodthe no-frills page covers everything in vm-script, but it's vbox specific (can run on macos or linux). vm-script is kvm and linux specific iiuc15:52
paulsher1oodso probably worth having two pages, one to cover vbox, one for kvm. and maybe standardize the style15:53
paulsher1oodbashrc: sadly i agree with you about the barrier to entry seeming high15:53
petefothand keep the pages ‘editor-agnostic’ 15:55
bashrcif you seek more adoption then maybe have one or more downloadable scripts which can be used to make a working VM so that users can get going quickly15:55
richard_mawthere's an argument for vagrant15:55
paulsher1oodscripts would be simpler, richard_maw :)15:57
robtaylorjjardon: is the git repo for the current tzdata 'coordinator'16:02
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock16:30
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]16:33
straycatwhat's the ip of the shared artifact cache?16:35
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock16:44
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]16:45
straycatwell, i just screwed up the dev machine i was using for import, at least i can deploy a new one in 2 minutes.16:49
pedroalvarezI've recovered devel systems in the past uncompressing a system artifact in /16:50
pedroalvarezwhich was pretty hacky but it worked :)16:51
*** tiagogomes [~tiagogome@] has quit [Quit: Leaving]16:51
* paulsher1ood realizes we should rename linux-jetson-tk1 to linux@jetson-tk1 etc16:55
Kinnisonis jetson-tk1 the name of the ref?16:56
* Kinnison thought the point of the @ was chunkname@shasum but is perhaps missing something16:56
paulsher1oodi thought the point was to differentiate chunk@foo from chunk@bar, where bar and foo may be refs or other things16:57
KinnisonSo they're meant to be conceptually entirely different things?16:57
* Kinnison thought the post-@ stuff was parameterisation16:58
paulsher1oodnot necessarily. but for example we could have linux at a specific ref with docker config, same without. 16:58
* Kinnison repositions the @foo stuff from "variant-by-parameterisation" to "entirely separate stuff" box in his head16:59
paulsher1oodi think it's still variant by parameterisation? it's certainly just a way of distinguishing variants of chunk. i'd hope anything after the @ was usefulmeaningful to humans - could be a tag, or not17:00
KinnisonIf there's no useful constraint on what can differ between foo@bar and foo@baz then they're entirely separate things in the conceptual model17:01
KinnisonHence my confusion, I thought it was just a way to affect which SHA was built from17:01
* Kinnison admits he didn't follow the conversation very closely as was occupied elsewhere at the time :-(17:02
jmacsSo I've managed to build the current ceph-service-x86_64-generic.morph, and I've set the versions of both boost and ceph to the same as the building version in my branch of baserock/sam/chef, but the baserock/sam/chef world still refuses to build boost17:02
paulsher1oodKinnison: the first use case was we had several chunk definitions called 'u-boot'17:02
paulsher1oodwe now have u-boot@jetson, u-boot@wandboard. people can interpret some meaning from that. and parsing can know that they are not the same chunk17:04
KinnisonRight, but in what way is that different from calling them u-boot-jetson u-boot-wandboard etc?17:05
* Kinnison is trying to determine the semantic value of the @....17:05
paulsher1oodthe @ gives ability to infer that they are both variants of u-boot. we already use - overmuch to achieve that17:06
KinnisonSo it's purely a convention for humans?17:06
paulsher1oodexcept that i could, for example easily script to identify which 'chunks' occur more than once17:07
Kinnisonpaulsher1ood: only on the assumption that the convention is followed uniformly17:07
*** zoli_ [] has joined #baserock17:07
*** zoli_ [] has quit [Changing host]17:07
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock17:07
* Kinnison would likely have advocated '+' over '@' if he'd known it was purely for humans17:07
KinnisonNo matter, 'tis done now17:07
KinnisonNo need to repaint17:08
paulsher1oodoh.. i'd be ok switching, not sure what the difference is though17:08
KinnisonI read '@' as "at"17:08
paulsher1oodradiofree read it as an email address ;)17:08
Kinnisonthus it implies what is before is a name and what is after is some kind of location or temporal identity17:08
Kinnisonemail addreses fit my parse mechanic17:09
KinnisonBut so does jeff@12pm17:09
Kinnisonor u-boot@2b53415417:09
Kinnisonanyway, I'm just needing to readjust my expectations, and that's fine17:10
* Kinnison has to do that all the time :-)17:10
jjardonrobtaylor: nice, thanks17:14
robtaylorjjardon: np17:14
* robtaylor is with Kinnison on this, u-boot@jetson reads as 'u-boot at jetson' to me also17:16
radiofreeis there a .morph domain name?17:17
rjekif you pay for it17:17
radiofreemaybe we can git send-email patches directly to the files17:18
jmacspdar: Any idea why I'm still getting the 'Boost thread library not found' issue? I'm not convinced it's built it yet, but I don't know how to check17:21
straycatheh, so it looks as though the version of setuptools i upgraded us to doesn't play nicely with a particular dependency needed to install openstack packages >.>17:21
* paulsher1ood still prefers | over @ and +, but would be happy to adopt + if this will lay it to rest :)17:22
pdarjmacs: Can I come see?17:23
jjardonKinnison: to be fair, AFAIK none of my changes broke any system already in definitions, and no one told me about specific network configurations I was not aware of when I submitted the patches. Saying that Id be happy to take a look when I have some time if someone put a system in definitions I can test against17:23
jmacspdar: Sure17:23
Kinnisonjjardon: I raised the issue the moment I saw the patches and was told it was "handled"17:23
Kinnisonpaulsher1ood: I'm fine with '@' being the token you chose17:24
paulsher1oodok, thanks17:24
Kinnisonpaulsher1ood: I would argue that '|' is a very dangerous metacharacter17:24
Kinnisonpaulsher1ood: esp. if someone wrote a definitions processor in shell :-)17:24
DavePageKinnison: Ceci n'est pas un pipe?17:24
KinnisonDavePage: well done, here's an apple17:25
richard_mawKinnison: I'd argue that's a good reason to put it in, so people don't do something so silly as to use shell and not sanitise their input17:25
Kinnisonrichard_maw: heh17:25
*** jonathanmaw [] has quit [Quit: Leaving]17:27
petefoth1maybe on the wiki, we could have downloadable scripts (for VBox, for KVM) for those who want to do stuff quickly, along with a narrative pages for each explaining what each script is doing and why for those who want to understand before they get started? 17:31
DavePageYou could have the narrative as comments within the script!17:31
petefoth1Wite the scripts sensibly and the wiki page can *be* the script17:31
paulsher1oodshow me the code :)17:31
petefoth1DavePage: indeed17:32
petefoth1paulsher1ood: show the code on the wiki page 17:32
jjardonKinnison: I explicity said (being aware that I could brake something) that the changes will only handle ethernet devices and  will configure them with DHCP: but we can always revert the patches if you think they are not ready17:37
paulsher1oodpetefoth1: i meant, why not write the code17:37
Kinnisonjjardon: I just think that rather than hardcoding the DHCP config in systemd's chunk you should have altered simple-network.configure to create the relevant configuration files :-)17:37
robtaylorpaulsher1ood: | makes sense in this case for a mathmo =) ('|' reads 'given that')17:38
petefoth1paulsher1ood: I'm thinking about how to write something that displays the content of some file in a wiki page. I have no idea how to do that, but I am sure someone has solved that problem already17:40
petefoth1is that the 'code' you suggest that I write? or...?17:41
jjardonKinnison: sure, but Im not sure when I will have time to do that, thats why my suggestion about the revertion if things are broken for people. Will try to spend some time over the weekend though17:41
petefoth1or just have a link on your quick-start page to the script file in g.b.o17:42
petefoth1with informative comments in your script - which may be there already17:43
Kinnisonjjardon: I think we should keep moving forwards, and just note this as a failure on my part to be engaged enough to catch the issue before merge :-)17:43
paulsher1oodpetefoth1: you mentioned having downloadable scripts (for VBox, for KVM) 17:43
petefoth1Ah yes. probabaly based on what bashrc wrote on the vm-script page17:44
* jjardon likes going forward ;)17:44
jjardonis it ok to lorry this?
*** Krin [] has quit [Quit: Leaving]17:50
jmacsIs there an easy way to tell which release of baserock a branch is based on?17:56
jmacsbaserock/sam/chef in this case17:56
paulsher1oodjmacs: have you built it?17:57
paulsher1oodtechnically it's not 'based on' a release.17:57
*** bashrc [] has quit [Quit: Lost terminal]17:57
paulsher1oodin a running baserock system, /baserock has metafiles describing every component17:58
jmacsAh, haven't tried actually running it yet17:58
paulsher1oodjmacs: you could just untar it for example17:58
*** mariaderidder [] has quit [Quit: Ex-Chat]17:59
paulsher1oodbut the ref info is in the definition anyway?17:59
paulsher1oodwhat are you actually trying to work out, if i may ask?17:59
jmacsI don't think I kept the deployed image unfortunately17:59
paulsher1oodjmacs: if you've built it, it'll be cached17:59
jmacsSam's branch won't build ceph-service, and I'm trying to figure out what's changed between sam branching off and 14.40.118:00
paulsher1oodgit diff?18:00
jmacsI imagine that'll be quite big18:01
paulsher1ooddoes 14.40.1 built it?18:01
paulsher1oodyou could git bisect, then18:01
paulsher1oodor maybe not... please hold18:02
jmacsI thought git-bisect required a linear set of commits18:02
jjardonrobtaylor: as you found it, can I get a +1 from you as well?
paulsher1oodjmacs: hence maybe not. 18:03
paulsher1oodyou can work out where a branch diverged from master, though18:03
straycatGiven, I don't expect the import tool to be able to import openstack18:04
straycatI'm also not sure whether upgrading setuptools may have implications for the existing openstack stuff18:04
straycatfranred, Have you tried rebuilding openstack stuff recently?18:04
franredjjardon, looks ok to me18:04
franredstraycat, yesterday18:05
straycatOkay cool18:05
franredbut I haven't rebase my branch since 2 weeks ago18:05
straycatOkay less cool18:05
paulsher1oodjmacs: 14.2618:05
paulsher1oodthe magic was...18:06
paulsher1oodgit merge-base origin/baserock/sam/chef origin/maste18:06
jjardonpushed! thanks franred paulsher1ood 18:06
paulsher1oodgit merge-base origin/baserock/sam/chef origin/master (sorry)18:06
jmacsmerge-base, that rings a bell now18:06
paulsher1oodgit describe --abbrev=0 6523c60a6e918d982ac378beef48e0d3dfa0603e18:06
paulsher1oodjmacs: so merge-base is where sam forked, and the describe command finds the latest tag before that18:07
straycatfranred, What's your branch?18:07
franredstraycat, ^^18:08
jmacsThanks paulsher1ood!18:08
paulsher1oodjmacs: yw18:09
robtaylorjjardon: hsure, +118:13
paulsher1oodstraycat: that setuptools comment is too deep for me, would it make sense to write up the problem and send to the list?18:14
* paulsher1ood wonders if we can get the arm mason working off the same trove as the x86 one, so they keep in step better?18:15
jjardonmmm, taking a look around, maybe it would be better to pass the configuration directly in the native format, so no script is needed, similar how coreos does:
straycatpaulsher1ood, maybe, I'll do some more investigation first18:20
paulsher1oodstraycat: ok18:21
radiofreewhy does binutils never seem to get cached?18:55
radiofreewhen doing a deploy with --no-git-update and no network, it complains about that18:56
radiofree"Expansion of upstream:binutils-redhat for pullpat yielded...."18:56
radiofreeand again :\18:58
radiofreeany help here?18:58
*** zoli__ [] has joined #baserock19:05
*** zoli_ [~zoli_@linaro/zoli] has quit [Read error: Connection reset by peer]19:05
*** zoli__ [] has quit [Client Quit]19:05
*** tiagogomes [] has joined #baserock19:14
*** tiagogomes [] has quit [Client Quit]19:14
paulsher1oodthat sounds like a bug, tbh20:56
paulsher1oodi've never understood why deploy needs network at all20:56
KinnisonIt needs network because it needs to make the build graph to know the artifact to deploy20:57
paulsher1oodmeh :)20:57
KinnisonAnd it needs network because we're not caching a bunch of the data the server works out for us20:57
KinnisonThat's the issue20:57
paulsher1oodsounds fixable, at least :)20:58
KinnisonSam was looking at it on the train back from London earlier this week20:58
Kinnisondunno how far he got20:58
paulsher1oodsounds like the demo idea won't work, radiofree 21:01
paulsher1oodbut not to worry21:01
paulsher1ood(i can't be sure of network)21:01
KinnisonIf you have all the git repos cached locally then it will work21:12
paulsher1oodi think he's trying to deploy after build.21:24
*** genii [~quassel@ubuntu/member/genii] has quit [Remote host closed the connection]22:46

Generated by 2.15.3 by Marius Gedminas - find it at!