| prometheanfire | ianw: not here https://gist.github.com/prometheanfire/a5d677906993f7c46e0bd812bebc7253 | 00:18 |
|---|---|---|
| ianw | prometheanfire: hrm, you crossing file-system boundaries, though? | 00:20 |
| ianw | are you disabling building with tmpfs? | 00:21 |
| prometheanfire | not for this | 00:22 |
| prometheanfire | not building with tmpfs | 00:22 |
| prometheanfire | The bind mount call attaches only (part of) a single filesystem, not possible submounts. The entire file hierarchy including submounts is attached a second place by using: | 00:24 |
| prometheanfire | mount --rbind olddir newdir | 00:24 |
| prometheanfire | so I don't think I can get around it with symlinks as it'd try and target something within the chroot I think | 00:24 |
| ianw | if we recursively mounted the hooks in run_in_target_d, would that work? | 00:34 |
| prometheanfire | ya, should, is that acceptable to you? | 00:35 |
| prometheanfire | testing changing 'sudo mount --bind ${TMP_HOOKS_PATH} $TMP_MOUNT_PATH/tmp/in_target.d' from function run_d_in_target in diskimage_builder/lib/img-functions to --rbind | 00:38 |
| ianw | hrm, the other thing is that the hooks dir is mounted in R/O, which is not what we want for a cache | 00:39 |
| prometheanfire | ya, I don't know if rbind will keep the stuff ro or not | 00:40 |
| prometheanfire | since it is a separate fs | 00:40 |
| ianw | a pre-finalise.d phase, that runs outside the chroot, but has access to the final image, would solve this anyway, right? you could bind the cache in there, and then remove it at the end of cleanup.d? | 00:42 |
| ianw | pre-finalise.d is probabaly == something running at 99 level in block-device.d; but i could see the argument for adding it as a new phase | 00:43 |
| prometheanfire | ya, that'd work too | 00:43 |
| ianw | there is certainly precedent that we have pre-install.d (outside) install.d (inside) | 00:45 |
| prometheanfire | if block-device.d has access it could go there, but that's poluting it possibly | 00:45 |
| ianw | yeah, i feel that it can get tricky enough there when you start with lvm etc etc | 00:48 |
| prometheanfire | going to need a 99 in install.d tooyep | 00:49 |
| prometheanfire | going to need a 99 in install.d too (if doing rbind) | 00:49 |
| prometheanfire | because all that is rbound needs to be unmounted | 00:49 |
| prometheanfire | 2018-10-12 00:48:13.713 | umount: /tmp/dib_build.VjFipiAq: target is busy. | 00:49 |
| ianw | having explored the options pretty thoroughly, i'm feeling like a pre-finalise.d that runs outside the chroot is going to be the best option | 00:49 |
| prometheanfire | yep | 00:49 |
| ianw | getting into recursive binds is just asking for trouble i think | 00:50 |
| prometheanfire | agreed | 00:50 |
| prometheanfire | hmm, --make-rslave can help, but ya, better in a new phase | 00:51 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Minor documentation updates https://review.openstack.org/609636 | 00:52 |
| prometheanfire | umount can be recursive too | 00:52 |
| prometheanfire | so, what's the next step, I'm not sure how to add a phase | 00:54 |
| ianw | let's see if it slots in ... | 00:59 |
| openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Add a pre-finalise.d phase https://review.openstack.org/609863 | 01:06 |
| prometheanfire | nice | 01:08 |
| prometheanfire | testing | 01:17 |
| prometheanfire | ianw: looks like the bind mount isn't hapening so it doesn't see the stage dirs | 01:34 |
| prometheanfire | ianw: https://gist.githubusercontent.com/prometheanfire/811ad40c64a46bb9a65c23721b7e4ac6/raw/c09e59eee0983361a5b16d78bbfa16021bdd5fd0/gistfile1.txt | 01:35 |
| prometheanfire | ize vs ise, because 'merica | 01:49 |
| *** zigo has quit IRC | 02:29 | |
| prometheanfire | ianw: it works :D | 05:25 |
| *** olivierb has joined #openstack-dib | 05:31 | |
| *** jesusaur has quit IRC | 05:41 | |
| *** jesusaur has joined #openstack-dib | 06:11 | |
| ianw | i think it was written by lifeless, who is .nz I believe where it would be finaliSe | 07:08 |
| *** jesusaur has quit IRC | 07:08 | |
| *** jesusaur has joined #openstack-dib | 07:13 | |
| *** zigo has joined #openstack-dib | 08:35 | |
| *** davido has joined #openstack-dib | 09:36 | |
| davido | Hi. I'm trying to establish if and when "diskimage-builder" should ever be init-capped. | 09:37 |
| davido | Looking at https://docs.openstack.org/diskimage-builder/latest/index.html | 09:37 |
| davido | the first heading uses "Diskimage-builder" but the first paragraph starts with "diskimage-builder" | 09:37 |
| davido | I'm editing training material for OpenStack and would like to get it right | 09:38 |
| *** davido has quit IRC | 16:06 | |
| *** davido has joined #openstack-dib | 16:06 | |
| *** mjturek has joined #openstack-dib | 17:39 | |
| *** olivierb has quit IRC | 18:21 | |
| *** dhill_ has quit IRC | 19:37 | |
| *** dhill_ has joined #openstack-dib | 19:41 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!