IRC logs for #baserock for Friday, 2015-05-01

*** bfletcher has quit IRC03:15
*** ssam2 has joined #baserock06:47
*** ChanServ sets mode: +v ssam206:47
*** mike has joined #baserock06:57
*** mike is now known as Guest2652306:57
*** mariaderidder has joined #baserock06:57
*** zoli_ has joined #baserock06:59
*** straycat has quit IRC07:13
*** zoli_ has quit IRC07:29
*** Albert_ has joined #baserock07:38
*** bashrc has joined #baserock08:04
*** jonathanmaw has joined #baserock08:19
*** gary_perkins has joined #baserock08:21
*** bfletcher has joined #baserock08:43
*** mariaderidder has quit IRC08:44
edcraggmorning, it seems the recent gcc update breaks on armv8... https://admin.codethink.co.uk/pnopaste/?275309:02
radiofreeD:09:03
rjekBoo09:03
franredtiagogomes, have you tested on reboot https://gerrit.baserock.org/#/c/517/?09:04
tiagogomesfranred no09:04
tiagogomesI want to finish with my patches first09:05
franredumm, tiagogomes have you tested your patches rebooting the VM multiple times?09:06
tiagogomesfranred multiple times? No, just a reboot for one node09:06
* pedroalvarez thinks some patchs can be merged09:07
* franred too09:08
radiofreeedcragg: i wonder if it's something to do with building it with gcc 4.x?09:12
radiofreethough that worked fine on x86 at least09:12
tiagogomesThat's the type of thing why I advised to not update gcc09:14
pedroalvarez:)09:14
pedroalvarezWe have had this sort of problems before for other archs, but we always managed to solve them, moving us forward09:15
edcraggradiofree: there are quite a few places in the morphologies' build steps which select for armv7 and set gcc flags etc, where armv8 is totally missing09:15
radiofreeedcragg: :\09:16
pedroalvarezI agree, we should have tested it before the merge, but, let's fix this!09:16
radiofreei've got the patches fedora applies to gcc5.1 here if you want to take a look through them09:16
edcraggi'll try to add them in and see if it fixes it, trouble is it's unknown how long we have access to the hardware, but that's another story09:16
radiofreethey don't look entirely relevant sadly09:17
*** lachlanmackenzie has joined #baserock09:17
radiofreei can't read rpm spec files, but i suppose we're at least doing target=aarch64-linux-gnu ?09:17
edcraggradiofree: yes, i read something about that, apparently related to the name of the toolchain?09:18
radiofreeah, probably09:18
radiofreeinterestingly they seem to have one generic set of config flags for "arm"09:18
radiofree--with-cpu=cortex-a8 --with-tune=cortex-a8 --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-abi=aapcs-linux"09:18
mwilliams_ctedcragg: could you repaste that on paste.baserock.org or somewhere else public?09:19
edcraggradiofree: for armv7, we set --target=armv7a and --host=armv7a09:20
edcraggalso, yes, those settings are included, and i think i've found the right ones for armv809:20
radiofreeedcragg: yeah, i'd expect us to be using armv8-a?09:20
edcraggyep09:20
radiofreenot sure what fedora is doing here then09:20
edcraggradiofree: if you replace a8 with a9, we have the same thing09:22
edcraggstrata/build-essential/stage1-gcc.morph09:22
*** wschaller_ has joined #baserock09:22
Zarathanks for reviewing https://gerrit.baserock.org/#/c/551/ , straycat :)09:50
pedroalvarezLooks like trove is broken :/09:51
Zara:009:51
pedroalvarez`su -c git 'some commands'` asks for a password09:51
pedroalvarez(being the root user)09:51
Kinnisonsu -c 'some commands' git09:52
pedroalvarezsu -c ' some commands' - git ?09:52
Kinnisonthe - only if you want the shell to behave as a login shell09:52
pedroalvarezKinnison: same result09:53
Kinnisondoes the git user exist?09:53
franredKinnison, yes09:54
pedroalvarezWe have done something recently with PAM, but I don't know if that is related09:54
*** straycat has joined #baserock09:54
Kinnisonhmm09:55
richard_mawPAM is the most likely culprit09:55
straycatrichard_maw, biff09:55
pedroalvarezotoh, `sudo -u git 'some command'` seems to work09:59
pedroalvareznote: this is breaking trove because trove-setup script needs to run things as git09:59
pedroalvarezgit user*09:59
Kinnisonpedroalvarez: I have a meeting now, but I'll come help if you've not solved it by the time I'm done10:00
richard_mawpedroalvarez: we only went with `su` way back then because we didn't have sudo10:01
* richard_maw recalls from the systemd mailing list that `su` may start a new logind session, especially with `su … -`10:02
pedroalvarezI understand. It may make sense to move to `sudo`, but I think that we should fix this anyway10:02
richard_mawdo we have the following in /etc/pam.d/su ?10:07
richard_maw# This allows root to su without passwords (normal operation)10:07
richard_mawauth       sufficient pam_rootok.so10:07
*** flatmush1 has quit IRC10:12
*** flatmush has joined #baserock10:12
richard_mawthanks for the review straycat, I'm currently in the middle of putting my devel system back together, so I'll take a look at it when I can10:13
jonathanmawrichard_maw: I did not add 'su' to /etc/pam.d/su10:15
*** edcragg has quit IRC10:16
jonathanmawso the pam changes must have broken it10:16
jonathanmawalso, confirmed that adding "auth sufficient pam_rootok.so" to /etc/pam.d/su works without password10:20
*** straycat_ has joined #baserock10:23
straycat_straycat: ping10:23
*** straycat_ has quit IRC10:23
straycatrichard_maw, cool10:23
jjardonpedroalvarez: Hi, thanks for the review in https://gerrit.baserock.org/#/c/558/ ! . I wonder if you have any kind of problem testing it? I did not have problems booting the USB stick in my laptop, but other users reported problems in https://gerrit.baserock.org/#/c/556/10:29
jjardonrichard_maw: Hi, only curious: did you explore the possibility of use dracut to generate the initramfs image? I might be wrong but seems to be the tool used upstream for that kind of thing10:32
pedroalvarezjjardon: I'd like to know how radiofree tested the change10:33
pedroalvarezjjardon: adding initramfs shouldn't break anything10:33
*** flatmush1 has joined #baserock10:34
*** flatmush has quit IRC10:34
pedroalvarezjjardon: and btw, regarding the 4 steps you have posted, I think that you can't forget step 2, because it's going to complain if  it's not built when you deploy cluster10:35
radiofreepedroalvarez: i'm no sure how just adding an initramfs will work though?10:37
radiofreeyou've not changed the kernel parameters or anything?10:37
radiofreei dded the image to a usb, it failed at the mount rootfs stage10:37
radiofreesince i'm assuming it's still trying to mount /dev/sda?10:37
*** zoli__ has joined #baserock10:40
jjardonpedroalvarez: yeah, thats what I though, so the problem is somewhere else10:41
*** flatmush1 has quit IRC10:42
*** flatmush has joined #baserock10:42
pedroalvarezradiofree: if you did so, your kernel args shouldn't have any /dev/sda hardcoded10:46
pedroalvarezhang on10:46
* pedroalvarez checks one important thing10:46
pedroalvarezfalse alarm10:48
pedroalvarezthat should just work, really10:48
pedroalvarezI'm going to try it myself10:48
pedroalvarezrichard_maw: after some investigation with Kinnison and franred, we also think that we are missing some files in /etc/pam.d/, so we are trying to make shadow install those files at build dime10:53
pedroalvarezs/dime/time10:54
pedroalvarezfranred: https://gerrit.baserock.org/#/c/548 biff10:57
richard_mawjjardon: dracut is not the only available tool for making initramfs images. Debian/Ubuntu have their own thing too. We didn't go with dracut because for our purposes, it's *way* overkill.10:59
franredpedroalvarez, my point is... how are we going to test it? do I need to compile it for the calxeda box?10:59
richard_mawjjardon: if we need to expand to support anything else in the initramfs we would need to re-evaluate it.10:59
pedroalvarezfranred: hmm... good point, I guess11:00
pedroalvarezis anybody using cxmanage stratum/system nowadays?11:00
KinnisonI think ssam2's user might be the only one11:01
Kinnisonfranred: how's that test-build going?11:02
pedroalvarez90% stuck in systemd11:03
franredKinnison, 103/108 building linux-x86-64-generic11:03
pedroalvarezoh, nearly there11:03
jjardonrichard_maw: yeah, I know several tools exists (Arch have mkinitcpio), but seems that dracut is the one more distros use. Thanks for the info anyway11:03
*** edcragg has joined #baserock11:04
franredKinnison, pedroalvarez, do we have any reference about which systems we are going to maintain and how to propose system for maintenance and how to remove system which are not longer maintainable and where to track these systems if at some point we need to revive them?11:07
KinnisonI imagine that's another policy which needs writing11:08
franredpedroalvarez, Kinnison, something is very wrong with this machine... I can't login into it11:14
pedroalvarezfranred: :)11:15
Kinnisonfranred: :(11:15
franredpedroalvarez, Kinnison, "Module is unknown"11:15
radiofreeoooooooh11:28
radiofreei forgot to add INITRAMFS_PATH11:28
*** wschaller_ has quit IRC11:45
pedroalvarezradiofree: :)11:48
pedroalvarezis it working now?11:48
pedroalvarezin other news, adding the /etc/pam.d/ files provided by shadow, makes possible to log in as root, with the wrong password11:49
pedroalvareztiagogomes: I kind of don't want to merge 50811:53
richard_mawbut not no password?11:53
pedroalvarezrichard_maw: no password, (just pressing enter) also logs in11:55
pedroalvarezsame behavior for a non-root user11:56
tiagogomespedroalvarez 508, I am surprised. Repeating those variables can lead to errors. I don't know why we use YAML inheritance if we only partially use it12:01
pedroalvareztiagogomes: and I understand12:01
pedroalvarezI'd prefer if we can avoid the use of those variables where it isn't needed12:01
pedroalvarezfor example, in the network node12:01
pedroalvarezI think your patches allow that12:02
pedroalvarezyup12:02
pedroalvarezafter 513 and 514 we won't need to set those variables in the network/controller node12:03
pedroalvarezonly in the compute node12:03
pedroalvarezI think that woud be much neater than using NOVA_TYPE (for example) in the controller and in the network node, and removing it from the compute node because it is inherited12:04
tiagogomespedroalvarez I don't follow, why we don' need to set those variables after 513 51412:07
pedroalvarezoh12:08
pedroalvarezright12:09
pedroalvarezI am wrong12:09
pedroalvarezyou do need those variables even in the controller node12:09
pedroalvarezlet me chek one last thing12:09
tiagogomesAlso bear in mind that's because everything is set on the controller node, it doesn mean that all those settings will be on /etc/openstack/ on the networking node, as we choose which variables we want to pass in the configuration extension12:10
pedroalvareztiagogomes: yeah, but NOVA_VIRT_TYPE: kvm in the network node doesn't sound right to me12:11
tiagogomesI am not fussy about separating things, but I like consistency. Having everything in the networking node on the cluster is what is being done for most of the settings, I just added the remaining12:11
pedroalvarezmeh, I don't mind really12:13
pedroalvarezI think the cluster is being done that way because the configure extensions were a bit silly12:13
tiagogomespedroalvarez it wasn't be that put it there tough. But the compute node inherits from the networking and not the controller, so to remove it it needs to go to the compute12:13
pedroalvarezthe patch is ok as it is12:14
pedroalvarezI only think that it could be better after your patches12:15
*** edcragg has quit IRC12:15
tiagogomessame for cinder, as cinder runs in the controller and compute12:15
tiagogomessame for nova12:15
tiagogomessame for ceilometer12:15
pedroalvarezanyway, removing duplication sounds ok, to avoid confusion12:16
tiagogomespedroalvarez I think that the compute should inherit from the controller instead12:16
pedroalvareztiagogomes: we can figure out how to improve this later12:16
tiagogomesthat way we could move all the cinder,ceilometer,nova to the controller part12:17
pedroalvarezbut yes, I was thinking the same12:17
*** zoli__ has quit IRC12:25
pedroalvareztiagogomes: https://gerrit.baserock.org/#/c/516/2 biff12:25
tiagogomespedroalvarez good catch12:26
pedroalvarezI was about to merge it12:26
tiagogomesIt wouldn't break anything12:32
jjardonfranred: if you are still working on that; git 2.4 has been already released. It comes with a interesting feature: atomic pushes12:33
*** zoli__ has joined #baserock12:35
*** zoli__ has quit IRC12:51
franredjjardon, Im about testing 2.3.7... and it takes time to do it because we have to test a demo trove to update git... and there are other things broken when I try to test it which me and pedroalvarez are trying to fix12:54
* richard_maw wants git 2.412:58
richard_mawI wouldn't've had to do http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/morph.git/commit/?id=eaa06853a161d87b1623cf56f75d3ec404ba06f6 if I'd had atomic pushes way back12:58
* richard_maw is rather annoyed that atomic pushes aren't the default12:58
tiagogomesERROR: Cluster clusters/release.morph contains the following duplicate deployment names:13:03
tiagogomes    initramfs13:03
* richard_maw usually appends -${NAME_OF_PARENT_DEPLOYMENT}13:03
richard_mawstraycat: any opinions on 13:14 < tiagogomes> same for ceilometer13:04
richard_maw13:15 <+pedroalvarez> anyway, removing duplication sounds ok, to avoid confusion13:04
richard_mawbah, mis-pasted13:04
richard_mawstraycat: https://gerrit.baserock.org/#/c/492/13:04
* straycat will take a look13:07
pedroalvarezsigh..13:07
pedroalvareztiagogomes: are you fixing that?13:07
franredrichard_maw, jjardon, I won't be able to test 2.4 today neither next week, so feel free to take over the update git topic "https://gerrit.baserock.org/#/c/484/"13:11
*** ssam2 has quit IRC13:12
franredKinnison, removing selinux makes login to fail...granting access to any user13:15
Kinnisonfranred: er13:15
Kinnisonfranred: it didn't on our test system13:15
pedroalvarezKinnison: this is on our test system13:16
Kinnisonpedroalvarez: But we logged in before lunch13:16
pedroalvarezyes, the current problem is that you can put any password, it will let you in13:17
Kinnisonaah13:17
Kinnisoninteresting13:17
pedroalvarezwe were testing that nothing else was broken13:17
KinnisonI fear I need tea to cope with this13:17
pedroalvarezjjardon: do you mind if I resend 265 just to upgrade ppc64 kernel?13:20
tiagogomespedroalvarez not planning13:20
*** edcragg has joined #baserock13:20
* pedroalvarez sends https://gerrit.baserock.org/561 13:26
tiagogomespedroalvarez why systems/build-system-x86_32.morph needs a initramfs?13:29
pedroalvarezis not needed, is not needed either for the 64b sytems either13:31
pedroalvarezbut having initramfs on all makes more sense than having it only in some13:31
jjardonpedroalvarez: send a different patch and I will rebase that against yours13:36
jjardonSo we can keep track of what still needs to be tested13:37
*** zoli__ has joined #baserock13:46
tiagogomescan I checkout a a topic branch pushed to gerrit?13:57
radiofreei think so14:00
*** Albert__ has joined #baserock14:02
*** Albert_ has quit IRC14:02
*** Guest26523 has quit IRC14:02
richard_mawstraycat: I've updated to fix the whitespace, can I assume you have no other concerns with the patch?14:02
rdalei'm using the latest morph and latest definitions, and i'm getting this error: ERROR: Cluster clusters/release.morph contains the following duplicate deployment names: initramfs14:03
richard_mawrdale: 14:25  * pedroalvarez sends https://gerrit.baserock.org/56114:04
richard_mawrdale: we're aware and there's an outstanding patch that may fix it, if you'd like to help review it14:05
rdaleok, i can test the fix14:05
pedroalvarezmany interesting things to fix regarding pam, shadow, and systemd14:08
straycatrichard_maw, *nod*14:09
richard_mawthanks straycat14:11
paulsherwood6/win 2314:13
paulsherwood... i should never multitask14:14
rdalei've given a +1 from me for 561 as it fixed the problem14:14
*** Guest26523 has joined #baserock14:16
jonathanmawwell, that's going to bite GENIVI in about a month. genivi-demo-platform-hmi depends on dbus-c++, which is a project that hasn't been updated since 2011/2012, and is hosted on gitorious14:17
rdalegitorious is staying, but read only14:18
*** zoli__ has quit IRC14:26
Kinnisonpedroalvarez: Just in case you're comparing Debian (or derived) systems and Redhat (or derived) systems, be aware that Debian has been using common-* since lenny (2009ish) and RedHat system-* for a similar amount of time14:37
Kinnisonpedroalvarez: effectively just two different ways for the same things14:37
pedroalvarezKinnison: thanks, and thanks for your help debugging it :)14:39
KinnisonThere's also some potential for cleverness in jumping/skipping around if you want14:40
KinnisonFor non pam_unix.so resolution14:40
KinnisonBut I don't think we care about that for now14:40
Kinnisonpedroalvarez: It has been fun and I've learned more and understand more about pam now :)14:43
franredpedroalvarez, http://linux.die.net/man/5/pam.d between requisite or required (and I would say requisite)14:54
Kinnisonfor pam_deny -- requisite14:56
Kinnisonrequired=[success=oknew_authtok_reqd=okignore=ignoredefault=bad]14:56
Kinnisonrequisite=[success=oknew_authtok_reqd=okignore=ignoredefault=die]14:56
Kinnisonthe die is the critical difference14:56
Kinnisongah, I apologise for the tabs14:56
pedroalvarezi missed that bit14:57
DavePagepam is one of those bits of *nix which is potentially cool but rarely used interestingly14:57
KinnisonDavePage: and one of those bits which when subtly wrong results in very *VERY* confusing errors14:58
radiofreepedroalvarez: does the installer system have fdisk/parted?15:02
radiofreei was thinking the "partition your hard drive" section could just call that15:03
pedroalvarezradiofree: I don't think it has those tools15:04
radiofreepedroalvarez: i've not tested the installer, but looking at the code it looks very automatic15:04
franredtiagogomes, https://gerrit.baserock.org/#/c/517/3 biff :)15:04
pedroalvarezradiofree: yes, it has to be automatic15:05
tiagogomesfranred biff for yourself as well15:05
radiofreewhy?15:05
radiofreei don't like the idea of creating a "insert this usb stick to totally obliterate /dev/sda"15:05
jjardonrichard_maw: can system/initramfs-x86_64 be used for x86_64 and x86_32 systems?15:06
franredtiagogomes, :) https://gerrit.baserock.org/#/c/517/315:06
pedroalvarezradiofree: then, don't use the installer system :)15:06
pedroalvarezradiofree: it was created to automate the deployment of baserock on hardware15:06
radiofreeah15:06
radiofreewell i might give up with this then, there's probably no real use-case if we can easily boot off usb sticks15:07
radiofreethat defconfig is generic enough to work on x220's at least15:07
pedroalvarezIt has some interactive things, and more can be added, but it should continue allowing automatic installation15:07
richard_mawjjardon: no, it needs to build busybox etc. for the appropriate target. Theoretically you could use the x86_32 build for both if you build your kernel to support the 32-bit ABI15:07
pedroalvarezjjardon: oh, good catch15:09
*** zoli___ has joined #baserock15:12
pedroalvarezjjardon: fixed15:13
Zaranewest thing that has confused me: http://paste.baserock.org/dimujixudi . the system in question includes build-essential, which includes glib-stage2. Does this refer to something else? Am I missing a flag somewhere?15:15
jjardonpedroalvarez: thanks! (and sorry)15:15
*** zoli___ has quit IRC15:16
pedroalvarezedcragg: I can't see why https://gerrit.baserock.org/#/c/562/ solves syntax errors15:17
pedroalvarezwe are doing the same in other systems15:18
richard_mawit's not a syntax error15:19
richard_mawI think15:19
richard_mawAIUI it just folds the lines15:19
*** ssam2 has joined #baserock15:22
*** ChanServ sets mode: +v ssam215:22
tiagogomesrichard_maw, do you remember why did are you turn /var/run to be a symbolic link to /run?15:28
richard_mawbecause `ip netns` breaks if the files in /var/run are left over from a previous boot15:29
tiagogomesrichard_maw makes sense, ta15:31
* pedroalvarez wonders if we can put that symlink using install-essential-files15:38
pedroalvarezor fhs15:39
pedroalvarezto many alternatives nowadays :)15:39
KinnisonNominally /run is only available on systemd systems15:39
pedroalvarezI didn't know that15:39
pedroalvarezrichard_maw: any reason to not merge morph anchor?15:40
pedroalvarezssam2: nice bug https://gerrit.baserock.org/#/c/502/15:41
pedroalvarezoh no, the nice bug was alrady fixed15:42
pedroalvarez(using the location of sysroot deployment to / overwrites your system)15:43
franredZara you need pcre in your system15:44
franredor glib with g_regex15:44
franredsupport15:45
franredZara, maybe add strata/pcre-utils.morph as a build dependency of your stratum ?15:46
ZaraOkay. I'd assumed the glib in build-essential supported g_regex but I'm guessing not, then15:46
franredZara, just check if glib has g_regex enabled in the configure by default or if we do it in its chunk morphology15:47
pedroalvarezwow, CI is testing now 4 systems!!15:48
pedroalvarezgenivi, weston, build, and openstack!15:48
pedroalvarez"testing"15:48
Kinnisons/testing/test-building/ presumably15:48
pedroalvarezindeed15:48
KinnisonStill - that's good15:48
richard_mawpedroalvarez: no reason not to (I thought I had already submitted it)15:49
* pedroalvarez presses the button15:50
Zarafranred: thanks, in the end I did just add pcre-utils as a build dependency to the stratum, which fixed that error :)15:54
pedroalvarezZara: bear in mind that you might need it at run time15:58
franredwhich means that maybe you need to add the stratum also to the system15:59
*** CTtpollard has quit IRC15:59
*** Guest26523 has quit IRC16:00
*** fay_ has quit IRC16:00
ssam2pedroalvarez: yeah, the fact that configuration extensions can overwrite your system is a wider issue really16:01
Zaraokay, thanks for the warning :)16:01
*** fay_ has joined #baserock16:01
richard_mawyeah, I've wanted to be able to specify a system to deploy *FROM* for a while16:11
richard_mawthe difficulty is in working out how much sandboxing is appropriate16:15
ssam2that seems like a good solution, actually16:20
ssam2it's a bit crap that deployment can use anything from the host, it makes it all a bit harder to track what went on16:21
ssam2but we need extracting a system to not take 2 minutes for that to be practical16:21
richard_mawit would also make it easier to use tools specialist tools in write extensions16:21
richard_mawbut yeah, we definitely need fast sysroot creation16:22
richard_mawthough the existing hardlink staging area construction may be sufficient for the root16:22
*** straycat has quit IRC16:23
richard_mawactually, no, you don't want it to be able to invalidate the cache by remounting its staging as writable16:27
*** ssam2 has quit IRC16:40
*** jonathanmaw has quit IRC16:41
*** nowster has quit IRC16:41
*** franred has quit IRC16:50
*** Albert__ has quit IRC16:50
*** bashrc has quit IRC16:52
*** tiagogomes has quit IRC17:10
paulsherwoodrjek, DavePage - looking at http://wiki.baserock.org/vet-security/ and the statement 'The Git "Solution" - what about git merge-base --is-ancestor <commit> <commit>17:23
*** gary_perkins has quit IRC17:40
*** edcragg has quit IRC17:42
*** zoli__ has joined #baserock18:17
*** lachlanmackenzie has quit IRC18:20
*** zoli__ has quit IRC18:23
*** edcragg has joined #baserock19:34
*** edcragg has quit IRC20:14
*** zoli__ has joined #baserock20:58
*** zoli__ has quit IRC20:59
*** zoli__ has joined #baserock21:27
*** zoli__ has quit IRC21:31
*** zoli__ has joined #baserock21:37
*** zoli__ has quit IRC21:42
*** zoli__ has joined #baserock21:55
*** zoli__ has quit IRC21:58
*** zoli__ has joined #baserock22:03
*** zoli__ has quit IRC22:17
*** zoli__ has joined #baserock22:33
*** zoli__ has quit IRC22:38
*** zoli__ has joined #baserock22:42
*** edcragg has joined #baserock22:47

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