| *** hwoarang has quit IRC | 00:05 | |
| *** hwoarang has joined #openstack-dib | 00:09 | |
| *** sm806 has quit IRC | 00:22 | |
| *** sm806 has joined #openstack-dib | 00:23 | |
| *** sm806 has quit IRC | 00:27 | |
| *** sm806 has joined #openstack-dib | 00:28 | |
| *** jesusaur has quit IRC | 00:30 | |
| *** jesusaur has joined #openstack-dib | 00:34 | |
| *** hwoarang has quit IRC | 02:14 | |
| *** hwoarang has joined #openstack-dib | 02:16 | |
| *** hwoarang has quit IRC | 03:12 | |
| *** hwoarang has joined #openstack-dib | 03:16 | |
| *** hwoarang has quit IRC | 04:12 | |
| *** hwoarang has joined #openstack-dib | 04:18 | |
| *** hwoarang has quit IRC | 06:26 | |
| *** hwoarang has joined #openstack-dib | 06:28 | |
| *** sm806 has quit IRC | 07:42 | |
| *** sm806 has joined #openstack-dib | 07:43 | |
| *** pVito has joined #openstack-dib | 08:08 | |
| *** sm806 has quit IRC | 08:48 | |
| *** sm806 has joined #openstack-dib | 08:49 | |
| *** sm806 has quit IRC | 09:37 | |
| *** sm806 has joined #openstack-dib | 09:38 | |
| *** sm806 has quit IRC | 10:49 | |
| *** sm806 has joined #openstack-dib | 10:50 | |
| *** sm806 has quit IRC | 11:08 | |
| *** sm806 has joined #openstack-dib | 11:09 | |
| *** pVito has quit IRC | 11:17 | |
| *** pVito has joined #openstack-dib | 11:17 | |
| jangutter | yolanda: I'm trying to pare down my config to the barest minimum and I found something "interesting" with the block-device-efi element + centos7. | 12:24 |
|---|---|---|
| jangutter | yolanda: with the partition layout as is, it seems two partitions pick up the cloudimg-rootfs label. | 12:24 |
| jangutter | yolanda: meaning that the root=LABEL=cloudimg-rootfs cmdline makes systemd sad. | 12:25 |
| jangutter | yolanda: specifically "BSP" and "root" get detected by blkid as LABEL="cloudimg-rootfs" | 12:26 |
| jangutter | ianw: you might also be interested | 12:27 |
| yolanda | mm that's interesting... see that i specifically add a LABEL here http://git.openstack.org/cgit/openstack/tripleo-image-elements/tree/elements/overcloud-secure-uefi/block-device-default.yaml#n57 | 12:41 |
| yolanda | jangutter, can you try modifying the yaml you use here http://git.openstack.org/cgit/openstack/diskimage-builder/tree/diskimage_builder/elements/block-device-efi/block-device-default.yaml and add a label as i do? | 12:41 |
| yolanda | it may be getting wrong defaults | 12:41 |
| jangutter | yolanda: I tried adding cloudimg-rootfs as label to the fs, and that didn't seem to work. | 12:42 |
| yolanda | can you compare with the element i pasted you? it's the one i use it and seems to behave properly | 12:43 |
| jangutter | yolanda: it did work if I used explicit --root-label in create-disk-image, I think... (will have to check my notes) | 12:43 |
| jangutter | yolanda: is the BSP partition required? It looks like it's the one screwing up. | 12:44 |
| jangutter | yolanda: I think the problem is likely that lsblk gets confused by an "empty" partition or something. Maybe the BSP partition isn't properly zeroed out or something. | 12:46 |
| yolanda | jangutter, i'd say is needed but i don't know the details | 12:49 |
| yolanda | looking at the type, it's the BIOS boot partition | 12:50 |
| jangutter | heh: turns out removing it also makes the system boot. | 12:51 |
| jangutter | BUT, I'd be hesitant to sell it as a "fix". as you say, it might be needed in cases I'm not testing. | 12:52 |
| jangutter | yolanda: is there an easy way to get dib to "zero it out", or is it zeroed out by default? | 12:54 |
| jangutter | yolanda in the block-device-default.yaml I mean. | 12:54 |
| jangutter | Oh, this is the reason why you've got img-rootfs and block-device-efi has cloudimg-rootfs: https://github.com/openstack/diskimage-builder/blob/master/diskimage_builder/block_device/blockdevice.py#L208-L219 | 12:59 |
| *** TheJulia_sick is now known as TheJulia | 13:05 | |
| *** sm806 has quit IRC | 13:33 | |
| *** sm806 has joined #openstack-dib | 13:33 | |
| jangutter | yolanda: I can confirm that just setting label: cloudimg-rootfs and name: cloudimg-rootfs in block-device-efi/block-device-default.yaml still confuses the bootup. | 13:35 |
| jangutter | yolanda, but if I do cat /dev/zero > /dev/sda2 in the rescue prompt and reboot, everything works fine. | 13:35 |
| *** mjturek has joined #openstack-dib | 13:37 | |
| *** mjturek has quit IRC | 13:40 | |
| yolanda | well good to know that it worked. However i didn't get that problem, i wonder what's different | 13:56 |
| jangutter | yolanda: I think I traced it down. Somehow the cloudimg-rootfs label gets associated with the BSP partition. It's either a bug in blkid, or the BSP partition really should be zeroed out (or be missing). | 13:58 |
| jangutter | yolanda: if I submit a workaround to the element that zeroes out the partition, would it be acceptable? | 13:59 |
| *** sm806 has quit IRC | 14:16 | |
| *** sm806 has joined #openstack-dib | 14:17 | |
| *** mjturek has joined #openstack-dib | 14:30 | |
| *** sm806 has quit IRC | 14:43 | |
| *** sm806 has joined #openstack-dib | 14:44 | |
| *** sm806 has quit IRC | 14:55 | |
| *** sm806 has joined #openstack-dib | 14:55 | |
| yolanda | jangutter, i'd need to check and investigate more about it. I wonder why with a similar block device config, i don't hit that | 15:20 |
| yolanda | in which OS do you hit it? | 15:20 |
| jangutter | yolanda: centos7 (not centos7-minimal) | 15:21 |
| jangutter | yolanda: I built a quick element that fixes it from my side. Let me send up a review. | 15:23 |
| yolanda | jangutter, cool , i'll take a look asap | 15:39 |
| yolanda | i also use centos | 15:39 |
| *** sm806 has quit IRC | 15:46 | |
| *** sm806 has joined #openstack-dib | 15:46 | |
| *** openstackgerrit has joined #openstack-dib | 15:50 | |
| openstackgerrit | Jan Gutter proposed openstack/diskimage-builder master: Clean the BSP partition for the EFI layout https://review.openstack.org/641426 | 15:50 |
| jangutter | yolanda ^ | 15:50 |
| jangutter | yolanda: other workarounds include just removing the BSP partition or passing --root-label to disk-image-create | 15:51 |
| clarkb | jangutter: I left a question on ^ about tidying up that non zero exit code | 15:59 |
| *** sm806 has quit IRC | 16:00 | |
| *** sm806 has joined #openstack-dib | 16:00 | |
| *** pVito has quit IRC | 16:02 | |
| jangutter | clarkb: blockdev --getsize64 <---- you think this is a good idea maybe? | 16:09 |
| clarkb | jangutter: ya then divide by 512? | 16:11 |
| clarkb | something like that so that we can have a successful dd exit code on successful updates | 16:11 |
| jangutter | clarkb: facepalm moment: blockdev --getsz already divides by 512 | 16:11 |
| jangutter | clarkb - it's a sign that people have already been down this road before. | 16:12 |
| clarkb | indeed | 16:12 |
| *** sm806 has quit IRC | 16:27 | |
| *** sm806 has joined #openstack-dib | 16:28 | |
| jangutter | logan- is it possible to generate a dual-boot dib with EFI and BIOS boot? (is it a valid use case) | 16:31 |
| logan- | afaik it is a pretty common disk layout for efi installs, I think that's why it is the default layout in that element. I know I rely on that capability for bare metal images | 16:33 |
| jangutter | logan- hahaha, this particular usecase _is_ bare metal on my part! | 16:35 |
| logan- | if your metal is all EFI capable and configured to boot EFI, then you probably don't need a BSP or bios compat at all. if you use your images across a pretty wide range of hardware/configs, its more of a concern | 16:35 |
| jangutter | logan- yep... in fact, the _safest_ way to do this is to zero out the BSP partition just after it's created. | 16:37 |
| logan- | i'm curious if this finding applies to centos-minimal and other distros also | 16:37 |
| jangutter | logan- but, that would only work if nothing in the world actually _depends_ on this kind of deal. | 16:37 |
| jangutter | logan- I can check on centos-minimal. (I just want to finish the blockdev --getsz version.) | 16:38 |
| openstackgerrit | Jan Gutter proposed openstack/diskimage-builder master: [DNM] Clean the BSP partition for the EFI layout https://review.openstack.org/641426 | 16:43 |
| logan- | jangutter: there shouldn't be a filesystem on the BSP at all, much less one labelled 'cloudimg-rootfs'. something is definitely off with that image build. | 16:46 |
| jangutter | logan- yeah, share your sentiment. hence my hamfisted fix (hamfix). If 'twere up to me I'd just whack the BSP element. | 16:48 |
| jangutter | *partition | 16:48 |
| logan- | yeah I wonder if it is something related to how the cloud image element works, since it is not building a partition layout from scratch. if you could share a full DIB log with debug mode enabled of your image build that might be interesting to dig thru. | 16:49 |
| jangutter | logan- I could do that.... just normal '-x' is enough? disclaimer: this is running on software-factory so it's not master. | 16:50 |
| logan- | yeah I think -x | 16:51 |
| jangutter | logan-: it's really not super high priority from my side due to the plethora of workarounds.... | 16:53 |
| jangutter | logan-: but it's such an interesting issue to trace! | 16:54 |
| jangutter | logan-: centos-minimal does not have the same issue! | 17:12 |
| jangutter | logan-: let me just verify that I didn't accidentally add any workaround. | 17:12 |
| logan- | interesting | 17:12 |
| jangutter | logan-: ah, facepalm. it's still in place. | 17:13 |
| jangutter | logan-: this is really weeird. | 17:34 |
| jangutter | /dev/sda2: LABEL="cloudimg-rootfs" UUID="553934fe-094d-47be-8fbf-7b9e4751e02c" TYPE="ext4" PARTLABEL="BSP" PARTUUID="a624dc82-997a-4ab4-b902-25740c7a1e70" | 17:34 |
| jangutter | /dev/sda3: LABEL="cloudimg-rootfs" UUID="2cd0ae33-1ac6-4f74-bd8a-92679bfee76c" TYPE="ext4" PARTLABEL="root" PARTUUID="b5556d99-28a2-44e5-9987-b64911228d3c" | 17:34 |
| jangutter | and yet, the centos-minimal based image boots. | 17:34 |
| jangutter | logan-: summary centos7 and centos-minimal exhibit the same partition layout and labelling. centos-minimal seems to be able to tolerate it to some extent (systemd-fsck-root.service complains though) | 17:38 |
| logan- | is there actually an ext4 filesystem present in that partition? ie can you mount it or run 'tune2fs -l /dev/<bsp part>' | 17:41 |
| jangutter | logan-: full log's at - http://paste.openstack.org/show/747369/ | 17:43 |
| jangutter | logan-: checking.... | 17:44 |
| jangutter | logan-: to my astonishment, it appears tune2fs -l recognises _something_?? | 17:45 |
| jangutter | logan-: EXT4-fs (sda2): bad geometry: block count 556544 exceeds size of device (2048 blocks) so it's definitely a superblock or something on it in busted form. | 17:46 |
| jangutter | logan-: argh, hit the line limit at paste.openstack.org | 17:48 |
| jangutter | logan-: full log at https://file.io/aSoi1h | 17:51 |
| jangutter | logan-: sorry I'm just full of fail. PM-ing you the ephemeral link | 17:53 |
| *** irclogbot_0 has joined #openstack-dib | 18:12 | |
| *** irclogbot_0 has quit IRC | 18:36 | |
| *** mjturek has quit IRC | 18:51 | |
| *** mjturek has joined #openstack-dib | 19:13 | |
| *** mjturek has quit IRC | 20:06 | |
| *** mjturek has joined #openstack-dib | 20:07 | |
| *** pVito has joined #openstack-dib | 20:08 | |
| *** mjturek has quit IRC | 20:11 | |
| *** mjturek has joined #openstack-dib | 20:12 | |
| *** irclogbot_0 has joined #openstack-dib | 21:09 | |
| *** mjturek has quit IRC | 21:16 | |
| *** pVito has quit IRC | 21:24 | |
| *** pVito has joined #openstack-dib | 21:26 | |
| *** irclogbot_0 has quit IRC | 21:28 | |
| *** pVito has quit IRC | 21:34 | |
| *** mjturek has joined #openstack-dib | 21:39 | |
| *** pVito has joined #openstack-dib | 21:45 | |
| *** pVito has quit IRC | 21:52 | |
| *** sm806 has quit IRC | 22:07 | |
| *** sm806 has joined #openstack-dib | 22:07 | |
| *** irclogbot_0 has joined #openstack-dib | 22:20 | |
| *** irclogbot_0 has quit IRC | 22:34 | |
| *** irclogbot_0 has joined #openstack-dib | 22:36 | |
| *** irclogbot_0 has quit IRC | 22:44 | |
| *** mjturek has quit IRC | 22:51 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!