*** pmannidi_2 has joined #openstack-dib | 00:10 | |
*** andreas_s has joined #openstack-dib | 00:29 | |
*** andreas_s has quit IRC | 00:33 | |
*** Sukhdev has joined #openstack-dib | 00:43 | |
*** Sukhdev has quit IRC | 01:38 | |
*** andreas_s has joined #openstack-dib | 02:09 | |
*** andreas_s has quit IRC | 02:14 | |
rm_work | ugh, this is weird during the build, i think it has to to with the FS type maybe? http://paste.openstack.org/show/626566/ | 03:04 |
---|---|---|
rm_work | saw something about issues with FS that don't support xattrs on either side | 03:04 |
*** andreas_s has joined #openstack-dib | 03:33 | |
*** jesusaur has quit IRC | 03:36 | |
*** Sukhdev has joined #openstack-dib | 03:36 | |
*** andreas_s has quit IRC | 03:38 | |
*** andreas_s has joined #openstack-dib | 03:49 | |
*** andreas_s has quit IRC | 03:53 | |
*** jesusaur has joined #openstack-dib | 04:23 | |
*** pmannidi has quit IRC | 04:48 | |
rm_work | yeah it's not handling symlinks | 04:50 |
*** andreas_s has joined #openstack-dib | 04:50 | |
rm_work | https://github.com/pypa/pip/pull/1311 was merged, and I see it's in the version of python I have, but ... still this is failing. I guess probably it is the xattrs thing | 04:54 |
*** andreas_s has quit IRC | 04:54 | |
*** jesusaurum has joined #openstack-dib | 05:00 | |
rm_work | http://paste.openstack.org/show/626571/ | 05:02 |
rm_work | yeah :( | 05:02 |
*** jesusaur has quit IRC | 05:02 | |
rm_work | maybe http://paste.openstack.org/show/626572/ explains it best | 05:03 |
rm_work | that's both on py34 and py35 | 05:04 |
*** Sukhdev has quit IRC | 05:26 | |
prometheanfire | ya, that's confusing | 05:40 |
prometheanfire | try without and see if manually installing works or not? | 05:40 |
*** Sukhdev has joined #openstack-dib | 05:41 | |
*** Sukhdev has quit IRC | 05:42 | |
rm_work | ah so | 06:07 |
rm_work | prometheanfire: i got further | 06:07 |
rm_work | prometheanfire: it's related to python being compiled with LCHMOD support | 06:08 |
rm_work | it's trying to chmod symlinks when it copies them | 06:08 |
rm_work | and exploding because it seems we don't ACTUALLY have support for that | 06:08 |
prometheanfire | /tmp may be noexec? | 06:08 |
rm_work | unrelated to /tmp | 06:09 |
prometheanfire | ok | 06:09 |
rm_work | http://paste.openstack.org/show/626574/ | 06:09 |
prometheanfire | so, gentoo python doesn't support that? | 06:09 |
rm_work | I boiled it down to this | 06:09 |
rm_work | gentoo python is compiled with support for it set to true, but *gentoo* doesn't support it | 06:09 |
rm_work | or at least it doesn't with musl libc | 06:09 |
prometheanfire | does the folder '1' exist? | 06:09 |
rm_work | yes | 06:09 |
prometheanfire | k | 06:10 |
rm_work | it actually *copies* the files | 06:10 |
rm_work | i have traced it down to the exact line in python | 06:10 |
rm_work | File "/usr/lib/python3.5/shutil.py", line 201, in copystat | 06:10 |
prometheanfire | (gdbm ipv6 ncurses readline sqlite ssl threads wide-unicode xml -berkdb -build -doc -examples -hardened -libressl -tk -wininst ELIBC="-uclibc") 3.5.4(3.5){xpak}(03:20:47 PM 10/29/2017)(gdbm ipv6 ncurses readline sqlite ssl threads xml -build -examples -hardened -libressl -tk -wininst ELIBC="-uclibc") | 06:10 |
rm_work | or... hold on that line number is different for you | 06:10 |
rm_work | python3 -c 'import sysconfig; print(sysconfig.get_config_var("HAVE_LCHMOD"))' | 06:10 |
rm_work | that should return 0 | 06:10 |
rm_work | but it returns 1 | 06:10 |
prometheanfire | returns 0 on glibc | 06:11 |
rm_work | ok well | 06:11 |
rm_work | i can tell you this gentoo image doesn't have lchmod support :P | 06:11 |
rm_work | so why python is compiled with it is beyond me | 06:11 |
rm_work | it thinks it can do it, goes out to the OS, and explodes | 06:12 |
rm_work | thus the error | 06:12 |
prometheanfire | ya | 06:12 |
rm_work | so this is a problem for me <_< | 06:12 |
prometheanfire | grep -E '(HAVE_FCHMODAT|HAVE_LCHMOD)' /usr/include/python3.5m/pyconfig.h | 06:13 |
rm_work | k yes i see that :P | 06:16 |
rm_work | it should not be 1 | 06:16 |
rm_work | somehow it got set to 1 during the compile process | 06:16 |
prometheanfire | huh | 06:16 |
rm_work | this was a binary package, wasn't it? | 06:16 |
rm_work | or did we compile python | 06:16 |
prometheanfire | ya, looks like a legit bug, python? | 06:16 |
prometheanfire | no | 06:16 |
prometheanfire | it's compiled | 06:16 |
rm_work | hmm | 06:16 |
prometheanfire | especially if you selected 3.5 | 06:16 |
rm_work | well | 06:16 |
rm_work | normally python doesn't do that flag | 06:16 |
prometheanfire | ya | 06:16 |
rm_work | something had to *set it* | 06:16 |
rm_work | where is the ebuild for this... | 06:17 |
* rm_work looks | 06:17 | |
prometheanfire | ya, looking | 06:18 |
* rm_work installs mlocate | 06:18 | |
rm_work | maybe /usr/portage/dev-lang/python/python-3.5.4.ebuild | 06:20 |
prometheanfire | ya, that's the ebuild | 06:21 |
prometheanfire | looks like a legit bug | 06:24 |
rm_work | hmmm | 06:24 |
rm_work | trying to figure out how python *gets* that | 06:24 |
rm_work | we don't pass it as an arg | 06:24 |
prometheanfire | look at python's configure | 06:25 |
rm_work | err, trying to find | 06:32 |
prometheanfire | https://github.com/docker-library/python/issues/155 | 06:33 |
prometheanfire | I just grep'd | 06:33 |
prometheanfire | didn't find anything though | 06:33 |
prometheanfire | think it's a linked lib | 06:33 |
prometheanfire | posix something something | 06:33 |
rm_work | hmmmm | 06:37 |
prometheanfire | rm_work: are you sure it's musl that you need for fast boot, have you tested a normal gentoo image? | 06:44 |
rm_work | i don't know really | 06:44 |
rm_work | i don't know if this is only a bug on musl either | 06:45 |
prometheanfire | it isn't | 06:46 |
prometheanfire | I did test it on my laptop at least | 06:46 |
rm_work | so, would a normal image help me at all? T_T | 06:46 |
prometheanfire | try seeing how long it takes to boot this http://distfiles.gentoo.org/experimental/amd64/openstack/gentoo-openstack-amd64-hardened-nomultilib-20170819.qcow2 | 06:46 |
*** pmannidi_2 has quit IRC | 06:47 | |
prometheanfire | http://distfiles.gentoo.org/experimental/amd64/openstack/gentoo-openstack-amd64-hardened-nomultilib-20171021.qcow2 | 06:47 |
prometheanfire | that's more updated | 06:47 |
rm_work | k | 06:47 |
prometheanfire | iirc the musl image took 2-3 min to boot | 06:48 |
prometheanfire | that's why we are doing all this right? fast boot on non-virt systems | 06:48 |
rm_work | yes | 06:48 |
prometheanfire | also, https://github.com/esmil/musl/blob/master/src/stat/lchmod.c | 06:48 |
prometheanfire | musl supports it aparently | 06:49 |
rm_work | ermmmm | 06:49 |
rm_work | ok | 06:49 |
rm_work | well | 06:49 |
rm_work | it doesn't work?? | 06:49 |
rm_work | weird | 06:49 |
prometheanfire | maybe musl's fault, maybe not? | 06:49 |
rm_work | i mean, outside of python | 06:49 |
rm_work | i can't get it to work *not*using python either | 06:57 |
prometheanfire | oh, fun :( | 06:57 |
rm_work | maybe i just don't know how correctly... | 06:58 |
prometheanfire | https://github.com/libarchive/libarchive/commit/65fe103 | 06:59 |
rm_work | >_> | 06:59 |
prometheanfire | yep, sounds like musl is making others work around it so far | 07:01 |
rm_work | >_< | 07:01 |
prometheanfire | well, I bugged people in #musl | 07:03 |
rm_work | whatever | 07:03 |
rm_work | i have a solution | 07:03 |
rm_work | none of the symlinks matter. i'm going to update the install step to delete them all | 07:03 |
rm_work | lol | 07:03 |
prometheanfire | lol | 07:05 |
rm_work | it's all docs | 07:05 |
*** andreas_s has joined #openstack-dib | 07:29 | |
*** andreas_s has quit IRC | 07:29 | |
*** andreas_s has joined #openstack-dib | 07:29 | |
rm_work | whelp ok | 07:32 |
rm_work | image ... ummm | 07:32 |
rm_work | built | 07:32 |
rm_work | so \o/ | 07:33 |
rm_work | but now i need to fix the code for our agent to understand how to do the networking for the system | 07:34 |
prometheanfire | right | 07:36 |
prometheanfire | so... I still think you should test the normal glibc image | 07:36 |
prometheanfire | but as far as networking goes... I've done this for my openstack https://dev.gentoo.org/~prometheanfire/images/VM-host-network.svg | 07:37 |
prometheanfire | not sure exactly what you need though | 07:37 |
prometheanfire | https://wiki.gentoo.org/wiki/Handbook:AMD64/Full/Networking has some stuff | 07:38 |
rm_work | yeah i mean just | 07:39 |
rm_work | what the format of its interface definition files are | 07:39 |
rm_work | how to reload networking | 07:39 |
rm_work | how to handle namespaces | 07:39 |
rm_work | simple operational stuff | 07:39 |
rm_work | see: https://github.com/openstack/octavia/blob/master/octavia/amphorae/backends/agent/api_server/osutils.py | 07:40 |
rm_work | compare redhat/ubuntu style | 07:41 |
prometheanfire | it's all there :P | 07:41 |
rm_work | yeah it's THERE | 07:41 |
rm_work | but i have to tell it how to use it | 07:42 |
prometheanfire | lol | 07:42 |
rm_work | does it use debian style interface files? | 07:42 |
rm_work | or is it closer to redhat? or its own thing | 07:42 |
prometheanfire | I'm surprised amphora uses os config style at all | 07:42 |
prometheanfire | it's it's own thing | 07:42 |
prometheanfire | netifrc is the format | 07:42 |
prometheanfire | it can't do veths though (bugs me) | 07:42 |
rm_work | we have to define a bunch of custom interfaces and stuff | 07:43 |
rm_work | hmmmm | 07:43 |
rm_work | i don't ... THINK we need that... | 07:43 |
prometheanfire | it should do what neutron does imo | 07:43 |
rm_work | err | 07:43 |
prometheanfire | but that's an outsider looking in | 07:43 |
rm_work | what does neutron do? | 07:44 |
rm_work | does neutron set up multiple interfaces with namespacing inside VMs? | 07:44 |
rm_work | lol | 07:44 |
prometheanfire | has 'border' interfaces it adds to a bridge, from that bridge it creates taps, which can go into a namespace if needed and do whatever else it wants | 07:44 |
rm_work | err | 07:47 |
rm_work | and neutron does that inside the VMs how? | 07:47 |
rm_work | I'm not aware that they even have an agent | 07:47 |
prometheanfire | not inside VMs | 07:48 |
prometheanfire | just in general | 07:48 |
prometheanfire | make it stateless more or less | 07:48 |
rm_work | err... ok | 07:48 |
rm_work | but inside the VMs we have to run HAProxy isolated from the other interfaces | 07:49 |
rm_work | how do we do that | 07:49 |
prometheanfire | same way neutron runs dhcp-agents isolated from the other interfaces | 07:50 |
prometheanfire | or the dhcp-servers | 07:50 |
prometheanfire | network namespaces | 07:50 |
rm_work | right | 07:50 |
rm_work | but those are inside the VM | 07:50 |
prometheanfire | so | 07:50 |
prometheanfire | ocravia tells the agent on the vm 'use this config' | 07:51 |
rm_work | yes | 07:51 |
rm_work | and then the agent lays down the config | 07:51 |
prometheanfire | agent caches the config then applies it statelessly using iputils commands | 07:51 |
prometheanfire | ip link add, etc | 07:51 |
prometheanfire | ip netns exec | 07:51 |
prometheanfire | that way it's os independent | 07:52 |
prometheanfire | when server reboots (if it does) the agent does it all again | 07:52 |
prometheanfire | as the network config isn't saved | 07:52 |
prometheanfire | but the octavia delivered config is cached | 07:53 |
rm_work | yeah, reboots *shouldn't* happen (if they do, good chance the VM gets failed before it comes back up) anyway | 07:53 |
rm_work | but can we really do 100% of what we need to do using that method, i wonder | 07:53 |
prometheanfire | I'd expect so | 07:53 |
prometheanfire | I can't think of a reason why not at least | 07:54 |
*** hashar has joined #openstack-dib | 09:34 | |
*** jaianshu has joined #openstack-dib | 11:17 | |
*** dhill_ has quit IRC | 13:15 | |
*** jaianshu has quit IRC | 13:28 | |
*** andreas_s has quit IRC | 13:33 | |
*** andreas_s has joined #openstack-dib | 13:33 | |
*** andreas_s has quit IRC | 13:38 | |
*** andreas_s has joined #openstack-dib | 15:09 | |
*** andreas_s has quit IRC | 15:13 | |
*** hashar has quit IRC | 16:14 | |
*** andreas_s has joined #openstack-dib | 17:30 | |
*** andreas_s has quit IRC | 17:35 | |
*** brault has quit IRC | 17:39 | |
*** jesusaurum has quit IRC | 17:54 | |
*** jesusaurum has joined #openstack-dib | 18:13 | |
*** jesusaurum has quit IRC | 18:22 | |
*** jesusaurum has joined #openstack-dib | 18:24 | |
*** dhill_ has joined #openstack-dib | 18:56 | |
*** Sukhdev has joined #openstack-dib | 19:38 | |
*** jesusaurum has quit IRC | 20:15 | |
*** jesusaurum has joined #openstack-dib | 20:17 | |
*** jesusaurum is now known as jesusaur | 20:17 | |
*** jesusaur has quit IRC | 20:20 | |
*** jesusaur has joined #openstack-dib | 20:22 | |
prometheanfire | rm_work: the shutil thing works in py2 | 20:23 |
prometheanfire | but not py3 | 20:23 |
*** Sukhdev has quit IRC | 20:51 | |
*** ianw has quit IRC | 21:02 | |
*** ianw has joined #openstack-dib | 21:09 | |
*** andreas_s has joined #openstack-dib | 21:43 | |
*** andreas_s has quit IRC | 21:47 | |
*** Sukhdev has joined #openstack-dib | 21:53 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!