| mmedvede | different error this time! | 00:00 |
|---|---|---|
| *** cloudnull has quit IRC | 00:14 | |
| *** cloudnull has joined #openstack-dib | 00:14 | |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 00:16 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 00:16 |
| ianw | ok, that's trying to kpartx it just before we install grub | 00:17 |
| ianw | if grub installs to /dev/mapper devices but not /dev/loop devices, well that's a least a big clue | 00:17 |
| mmedvede | ianw: I think I see a bad copy paste, is it supposed to be TMP_IMAGE_PATH in 50-bootloader? | 00:18 |
| ianw | mmedvede: yes, i think so, that should be the .raw device | 00:19 |
| mmedvede | oh, it is global from disk-image-create | 00:21 |
| mmedvede | ianw: /tmp/in_target.d/finalise.d/50-bootloader: line 28: TMP_IMAGE_PATH: unbound variable | 00:25 |
| ianw | :( | 00:25 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 00:31 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 00:31 |
| mmedvede | ianw: also, is BOOT_DEVICE needs to be BOOT_DEV? | 00:34 |
| ianw | yeah, probably | 00:34 |
| ianw | let's see if this mounts | 00:34 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 00:37 |
| mmedvede | ianw: 50-bootloader: line 25: kpartx: command not found | 00:39 |
| ianw | ? | 00:41 |
| mmedvede | http://dal05.objectstorage.softlayer.net/v1/AUTH_3d8e6ecb-f597-448c-8ec2-164e9f710dd6/pkvmci/diskimage-builder/99/448099/13/check/dib-dsvm-functests-devstack-xenial/75f4985/ | 00:41 |
| mmedvede | ianw: it could not run kpartx from bootloader element | 00:41 |
| ianw | i mean how can that be | 00:41 |
| mmedvede | yeah, it is installed | 00:42 |
| mmedvede | ianw: finalize step is in chroot, maybe it is only installed on host | 00:43 |
| mmedvede | or I am getting it backwards | 00:44 |
| mmedvede | right, block-device is outside chroot, finalise is in chroot | 00:46 |
| mmedvede | and TMP_IMG_DIR not accessible there | 00:48 |
| ianw | yeah, sigh | 00:51 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 00:58 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 00:58 |
| ianw | starting to clutch at straws here ... | 00:58 |
| mmedvede | I think part of it is that there might be some confusion of when we are in chroot vs outside | 01:07 |
| ianw | hang on, now it's not failing in stable-vm | 01:12 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 01:15 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 01:15 |
| mmedvede | watching it live, got to stable-vm | 01:19 |
| ianw | mmedvede: it seemed to pass ... | 01:25 |
| mmedvede | ianw: so you clutching to straws did work :) | 01:25 |
| ianw | well it would be nice if it didn't raise more questions than it answered :) | 01:26 |
| mmedvede | ianw: there is the strange warning "warning: unknown device type dm-1" during grub-install | 01:26 |
| ianw | yeah, that's in the host system though | 01:26 |
| ianw | no in the image (i think) | 01:26 |
| mmedvede | ianw: I'll test a full-blown dsvm image build in my env | 01:26 |
| ianw | i guess this suggests that grub-install only works on /dev/mapper, and probably has something to do with the probe paths | 01:27 |
| ianw | i mean the way it probes paths | 01:28 |
| mmedvede | if that is the reason, that would mean it is prone to breaking again | 01:29 |
| mmedvede | ok, I am building a full image now. Would test if it is bootable too | 01:32 |
| ianw | mmedvede: great, let me know. grabbing some lunch, i'll think on it a bit ... | 01:48 |
| mmedvede | ianw: thanks. which tz are you? | 01:48 |
| mmedvede | it is 21:00 where I am | 01:49 |
| mmedvede | I am guessing somewhere in aus :) | 01:51 |
| mmedvede | ianw: the image dib built on ppc is good. VM boots | 02:05 |
| ianw | mmedvede: yes, sydney .au | 02:50 |
| ianw | ok, that's good! well, now to figure out what to do ... | 02:50 |
| *** aparnav has joined #openstack-dib | 03:32 | |
| *** Administrator_ has joined #openstack-dib | 04:38 | |
| *** zhugaoxiao has quit IRC | 04:41 | |
| *** chhavi has joined #openstack-dib | 05:02 | |
| *** chhavi has quit IRC | 05:15 | |
| *** chhavi has joined #openstack-dib | 05:16 | |
| *** chhavi has quit IRC | 05:21 | |
| ianw | oohh, i have another idea ... | 05:35 |
| ianw | i'll update the bug | 05:35 |
| *** venki has quit IRC | 07:42 | |
| *** venki has joined #openstack-dib | 07:42 | |
| *** andreas-f has quit IRC | 07:48 | |
| *** hashar has joined #openstack-dib | 08:19 | |
| *** venki has quit IRC | 08:23 | |
| *** chhavi has joined #openstack-dib | 09:28 | |
| openstackgerrit | yolanda.robla proposed openstack/diskimage-builder master: Add lvm management to diskimage-builder https://review.openstack.org/444403 | 09:49 |
| *** venki has joined #openstack-dib | 10:42 | |
| openstackgerrit | yolanda.robla proposed openstack/diskimage-builder master: Fix typo in CMDLINE env var for bootloader https://review.openstack.org/448508 | 11:02 |
| *** hashar is now known as hasharLunch | 11:08 | |
| *** chhavi has quit IRC | 11:13 | |
| *** venki has quit IRC | 11:18 | |
| *** venki has joined #openstack-dib | 11:19 | |
| *** aparnav has quit IRC | 12:02 | |
| *** venki has quit IRC | 12:19 | |
| *** venki has joined #openstack-dib | 12:19 | |
| *** hasharLunch is now known as hashar | 13:05 | |
| *** mgould has joined #openstack-dib | 13:17 | |
| mgould | hi everyone! could someone please review https://review.openstack.org/#/c/344214/ ? It has one +2 and green CI | 13:17 |
| *** venki has quit IRC | 13:25 | |
| *** venki has joined #openstack-dib | 13:25 | |
| *** chhavi has joined #openstack-dib | 13:31 | |
| *** venki has quit IRC | 14:05 | |
| *** venki has joined #openstack-dib | 14:05 | |
| *** chhavi has quit IRC | 14:25 | |
| openstackgerrit | Dan Radez proposed openstack/diskimage-builder master: Adding aarch64 support for CentOS7 https://review.openstack.org/422232 | 14:47 |
| *** qwebirc12229 has joined #openstack-dib | 15:19 | |
| qwebirc12229 | diskimage-builder centos7 | 15:19 |
| qwebirc12229 | how to do | 15:20 |
| mgould | qwebirc12229: disk-image-create -o [output image name] centos [other elements to include] | 15:29 |
| *** qwebirc12229 has quit IRC | 15:51 | |
| *** chhavi has joined #openstack-dib | 16:16 | |
| *** hashar has quit IRC | 17:29 | |
| *** mgould is now known as mgould|afk | 18:18 | |
| *** chhavi has quit IRC | 18:49 | |
| *** andreas-f has joined #openstack-dib | 19:26 | |
| *** hashar has joined #openstack-dib | 19:40 | |
| ianw | andreas-f: when you get a minute, could you look at https://bugs.launchpad.net/diskimage-builder/+bug/1674402 ... i have a question in there about using /dev/mapper | 20:20 |
| openstack | Launchpad bug 1674402 in diskimage-builder "bootloader element fails to install grub on ppc64 platform" [Undecided,In progress] - Assigned to Ian Wienand (iwienand) | 20:20 |
| *** hashar has quit IRC | 20:23 | |
| *** hashar has joined #openstack-dib | 20:23 | |
| andreas-f | iamw: I already read this. | 20:24 |
| andreas-f | My current idea is to unify this. | 20:24 |
| andreas-f | e.g. use kpartx *for all* | 20:24 |
| andreas-f | (and drop partprobe) | 20:25 |
| andreas-f | Do you know why there were (are) two ways of handling partitions? | 20:25 |
| openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Use stevedore for module config of block device https://review.openstack.org/447090 | 20:27 |
| ianw | andreas-f: no, i don't really know why | 20:29 |
| ianw | it seems it has just always been this way | 20:29 |
| andreas-f | I run some tests on Ubuntu: both ways are working there. | 20:29 |
| andreas-f | Can we get more PowerKVM machines: the CI finished in 6:30! ;-) | 20:36 |
| mmedvede | andreas-f: it is not running a full set of tests :) | 20:38 |
| mmedvede | but I hope I can expand the coverage | 20:38 |
| andreas-f | It does not build any 'vm' image? (Or do I miss something?) | 20:41 |
| andreas-f | (Would be nice to have, because this really triggers the partitioning code...) | 20:42 |
| ianw | andreas-f: we've been testing with https://review.openstack.org/#/c/448099/ | 20:42 |
| mmedvede | PowerKVM CI only runs partial functional test, and functional testing does not use vm element. That's the reason for ^ | 20:42 |
| ianw | andreas-f: i don't see where the code takes the results of kpartx? _notify_os_of_partition_changes() calls ... won't we need to parse the result to find out where the partitions are? | 20:43 |
| andreas-f | Is the test 'just' a build test - or is there also a test where the image is really started? | 20:43 |
| andreas-f | The current code is mostly a port of an 'old' bash script. | 20:44 |
| ianw | mmedvede did report that a built image booted | 20:44 |
| mmedvede | andreas-f: at the moment yes. I'd need to add nodepool job to test the VM booting | 20:44 |
| andreas-f | And it looks that nobody really knows why it is written as it is. | 20:45 |
| ianw | that's about right it seems | 20:46 |
| andreas-f | ianw: do you mind when I remove the 'partprobe' and 'udevadm' in https://review.openstack.org/#/c/447739/17/diskimage_builder/block_device/level1/partitioning.py? | 20:49 |
| andreas-f | Want to see the CI result for this. | 20:49 |
| openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 21:00 |
| bkero | ianw: ever figure out the prep partitoin thing? | 21:03 |
| ianw | bkero: sort of https://bugs.launchpad.net/diskimage-builder/+bug/1674402/comments/6 | 21:11 |
| openstack | Launchpad bug 1674402 in diskimage-builder "bootloader element fails to install grub on ppc64 platform" [Undecided,In progress] - Assigned to Andreas Florath (ansreas) | 21:11 |
| ianw | andreas-f: sure. will pull that apart into some separate changes in a bit | 21:11 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 21:12 |
| andreas-f | One technical question for the PPC change: Is this correct that the partition 0 will not be mounted at all? | 21:17 |
| andreas-f | (So no data?) | 21:18 |
| andreas-f | Where does it get the kernel / initramfs from? | 21:18 |
| andreas-f | Must this be a special (e.g. second) partition? | 21:19 |
| ianw | andreas-f: aiui this is for openfirmware code | 21:20 |
| ianw | grub copies it in | 21:20 |
| ianw | https://www.ibm.com/developerworks/community/wikis/home?lang=en#!/wiki/W51a7ffcf4dfd_4b40_9d82_446ebc23c550/page/PowerLinux%20Boot%20howto does explain it a bit | 21:21 |
| ianw | openfirmware looks for a prep typed partition that has stage 1 code | 21:22 |
| ianw | that page also does say to grub-install to /dev/sda1 ... however when i tried installing it to /dev/loop0p1 (the prep partition) it failed | 21:24 |
| andreas-f | Maybe I'm missing something, but this does not explain where the kernel and ramdisk have to be. | 21:24 |
| andreas-f | I'm asking because currently it is not possible to have a separate /boot and / partition for the other architectures (because there is the need to change the bootloader element for this) - but this should be implemented (as Yolanda requested this). | 21:24 |
| andreas-f | Found it: looks the /boot partition is not needed (but possible). | 21:26 |
| ianw | yeah, i guess only if your root is something sufficiently weird grub can't read it | 21:30 |
| openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 22:08 |
| andreas-f | CI is green: removed the partprobe and udevsettle - but I'm not sure if this was really tested.... | 22:09 |
| andreas-f | for 448099 it's not green: | 22:10 |
| andreas-f | 2017-03-22 21:30:01.704209 | [debian-minimal/stable-vm] /usr/sbin/grub-install: error: cannot find a GRUB drive for /dev/mapper/loop0. Check your device.map. | 22:10 |
| openstackgerrit | Mikhail S Medvedev proposed openstack/diskimage-builder master: [DNM] Test boot partition code https://review.openstack.org/448099 | 22:17 |
| mmedvede | the 448099 tests the partitioning. If it is not green, something is broken | 22:18 |
| andreas-f | Yes - it looks that for other systems grub does not like /dev/mapper | 22:19 |
| ianw | urgh | 22:24 |
| andreas-f | Was a problem in the code: the /dev/loopX is still the same - only the partitions are mapped to /dev/mapper/loopXpY. | 22:28 |
| andreas-f | Had to 'adapt' disk-image-build... | 22:28 |
| *** pmannidi has quit IRC | 22:29 | |
| andreas-f | Local test succeeds: | 22:30 |
| andreas-f | + /usr/sbin/grub-install '--modules=biosdisk part_msdos' --force /dev/loop3 | 22:30 |
| andreas-f | Installing for i386-pc platform. | 22:30 |
| andreas-f | Installation finished. No error reported. | 22:30 |
| openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Create PReP boot partition for PPC https://review.openstack.org/447739 | 22:31 |
| andreas-f | This should eliminate the special case handling in disk-image-create. | 22:31 |
| andreas-f | The last 'special' handling for PPC remains in 5ß-bootloader. | 22:33 |
| andreas-f | And I have currently no idea how to clean this up... | 22:33 |
| andreas-f | One possibility to leave this now hackish (as it was) and always assume that this is partition 0 of the IMAGE_BLOCK_DEVICE (a sed should do). | 22:39 |
| andreas-f | The next steps would be to merge https://review.openstack.org/#/c/444586/ (Creates the base for disk-image base queries). | 22:41 |
| andreas-f | Add another patch that adds a get-boot-dev function to dib-block-device... | 22:42 |
| andreas-f | ... and use this in bootloader. | 22:42 |
| andreas-f | The device where the [boot] flag is set should be returned. | 22:42 |
| ianw | yeah, or just the one named "boot" :) | 22:43 |
| andreas-f | And a last step: change the bootloader element to use this. | 22:43 |
| andreas-f | This should then enable also the possibility to have a separate /boot partition on all systems. | 22:44 |
| ianw | which helps yolanda with the lvm stuff too, right? | 22:44 |
| andreas-f | No - lvm is separated from this. | 22:45 |
| andreas-f | But what I understood is that she needs a separate /boot partition because of some security policy | 22:45 |
| andreas-f | (maybe I misunderstood.) | 22:45 |
| ianw | oh, ok | 22:45 |
| andreas-f | Would this be a way to go? | 22:46 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Capture output in _exec_sudo https://review.openstack.org/448810 | 22:46 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Add default PPC block-device layout https://review.openstack.org/448811 | 22:46 |
| ianw | i thought you'd probably want a plain old /boot if you've got lvm for root, but may also be wrong | 22:46 |
| andreas-f | Grub *can* boot from LVM ;-) | 22:47 |
| ianw | i think ^^ can stand-alone at this point, trying to reduce the change load | 22:47 |
| ianw | yeah, i guess it can; although maybe when you start adding things like luks encryption that's where it fails | 22:48 |
| andreas-f | Here it is nearly midnight and I have to go to sleep... Will just look through your patches. | 22:48 |
| ianw | andreas-f: np, thanks. i think we're moving in the right direction, anyway | 22:49 |
| *** pmannidi has joined #openstack-dib | 22:51 | |
| *** pmannidi has quit IRC | 22:51 | |
| *** pmannidi has joined #openstack-dib | 22:52 | |
| *** pmannidi has quit IRC | 22:52 | |
| *** pmannidi has joined #openstack-dib | 22:52 | |
| andreas-f | good night. | 22:54 |
| *** hashar has quit IRC | 23:01 | |
| *** andreas-f has quit IRC | 23:11 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!