== Meeting information == * #ubuntu-meeting Meeting, 09 Jul at 15:05 — 15:50 UTC * Full logs at [[http://ubottu.com/meetingology/logs/ubuntu-meeting/2015/ubuntu-meeting.2015-07-09-15.05.log.html]] == Meeting summary == === Lightning round === The discussion about "Lightning round" started at 15:05. === Team changes === The discussion about "Team changes" started at 15:27. === AOB === The discussion about "AOB" started at 15:33. * ''LINK:'' https://www.kernel.org/doc/Documentation/filesystems/tmpfs.txt == Vote results == == Done items == * (none) == People present (lines said) == * pitti (62) * slangasek (61) * infinity (36) * sil2100 (20) * cyphermox (18) * barry (13) * tyhicks (12) * caribou (10) * doko (8) * ubottu (8) * meetingology (3) * ogra_ (3) == Full Log == 15:05 #startmeeting 15:05 Meeting started Thu Jul 9 15:05:03 2015 UTC. The chair is slangasek. Information about MeetBot at http://wiki.ubuntu.com/meetingology. 15:05 15:05 Available commands: action commands idea info link nick 15:05 [TOPIC] Lightning round 15:05 $ echo $(shuf -e barry doko stgraber bdmurray slangasek caribou infinity sil2100 robru cyphermox pitti) 15:05 caribou stgraber bdmurray infinity robru cyphermox pitti slangasek barry sil2100 doko 15:06 oh, I get to start 15:06 caribou: yep :) 15:06 Bugfix : 15:06 - ifenslave ambigous message 15:06 * identified root cause, open Debian bug with patch 15:06 Crash Dump 15:06 - Worked on Kdump enablement cloud-init script 15:06 Charm dev 15:06 - Developed Kernel Crash Dump subordinate charm 15:06 (done) 15:08 ah, stgraber probably not around 15:08 bdrung_work: 15:08 eh 15:08 bdmurray: 15:08 stupid tabcomplete 15:08 We might not have a Brian either. 15:09 yes, we indeed might not since I approved his vacation request 15:09 infinity: 15:09 Short (3-day) week: 15:09 - General SRU/AA work 15:09 - Helped apw sort out ubuntu-fan for wily 15:09 - Worked with apw on ubuntu-fan SRU proposal 15:09 - Worked with rbasak on docker.io SRUs 15:09 - Found, debugged, and ushered in a fix for a kernel panic on the arm64 buildds 15:09 - Working on LP: #1414818 for IBM 15:09 Launchpad bug 1414818 in opensm (Ubuntu) "Add missing OFED packages for Ubuntu 14.04.2" [High,New] https://launchpad.net/bugs/1414818 15:09 - Identified casper backports needed for 14.04.3 and got them in with apw 15:09 - Kernel SRU wrangling for 14.04.3 15:09 - Some merges and bugfixes in wily 15:09 - Failed to sleep all week 15:09 (done) 15:10 ssh infinity sleep 28800 15:10 sleep: command not found 15:11 pitti: underlying kernel bug, never puts the CPU into anything below C0 15:11 robru? 15:11 robru: 15:12 slangasek: That's the most accurate nerd analogy for insomnia that I've ever seen. 15:12 robru might be in C3? 15:13 heh 15:13 should I? 15:14 cyphermox: go ahead 15:14 * fixed d-i overlay support (net-retriever, base-installer, apt-setup) 15:14 * updated sg3-utils (+ added udeb) for QEMU/IPR support for multipath 15:14 * finished multipath-tools merge 15:14 * updated partman-multipath, partman-base, hw-detect for naming change 15:14 * investigated/fixed tgt for proposed migration (blocked multipath-tools) 15:14 * investigated lava-dispatcher (blocked multipath-tools, skipped) 15:14 * fix fwupdate FTBFS 15:14 * SRUs: 15:14 - debian-installer-utils bug 1402042 (support --- in user-params) 15:14 bug 1402042 in debian-installer-utils (Ubuntu Utopic) "console= parameters need to be added before -- on kernel cmdline" [Undecided,In progress] https://launchpad.net/bugs/1402042 15:14 - udev (systemd) bug 1437375 for trusty (exception for ibmveth MAC) 15:14 bug 1437375 in systemd (Ubuntu Trusty) "[udev] Adding "Austin" adapter to Ubuntu partition take over system network interface" [Undecided,In progress] https://launchpad.net/bugs/1437375 15:14 - partman-auto bug 1461860 for trusty (ppc64el recipe) 15:14 bug 1461860 in partman-auto (Ubuntu Trusty) "Installation scheme based on Separate /home, /usr, /var, and /tmp partitions fail" [Medium,In progress] https://launchpad.net/bugs/1461860 15:14 * debugging multipath-tools: 15:14 - bug 1463046: local HDs found as multipath when they shouldn't be 15:14 bug 1463046 in multipath-tools (Ubuntu) "installation of multipath-tools-boot can break boot" [High,Triaged] https://launchpad.net/bugs/1463046 15:14 - bug 1468897: skip USB devices 15:14 bug 1468897 in multipath-tools (Ubuntu) "multipath creates binding for Removable(USB) drives" [Medium,In progress] https://launchpad.net/bugs/1468897 15:14 * investigating iso smoketest failures 15:14 (done) 15:15 autopkgtest cloud: 15:15 - britney: Add much simpler AMQP/cloud based autopkgtest triggering, to get rid of lp:auto-package-testing and much indirection+duplication (landed) 15:15 - britney: Add autopkgtest result retrieval from swift (MP pending review) 15:15 - create swift server mock for writing britney tests 15:15 - do britney run for wily, triggering 500 tests; various robustifications and fixes from these 15:15 - Rewrite autopkgtest-cloud's swift results retriever to be suitable for inclusion into debci 15:15 systemd: package/test 222 (landed now), various bug fixes (#1450009, 1471258) 15:15 ecryptfs-utils: Fix broken cryptswap configuration with LVM, clean up on upgrades, prepare SRUs (bug 1453738), fix broken libecryptfs0 shipping libecryptfs.so.1 15:15 bug 1453738 in ecryptfs-utils (Ubuntu Vivid) "installer in LVM mode sets up broken encrypted swap, using duplicate unencrypted swap" [High,In progress] https://launchpad.net/bugs/1453738 15:15 misc: 15:15 - review https://code.launchpad.net/~brian-murray/apport/support-ppa-packages/+merge/263437, almost ready now 15:15 - merges: openvpn, policykit-1, ruby-defaults 15:15 [END] 15:17 * short week: holiday last Friday, sick yesterday 15:17 * miscellaneous SRU processing 15:17 * helping set up ci train silo for the upcoming gcc 5 transition 15:17 * livecd-rootfs sponsorship to move a scope out of the phone rootfs 15:17 * TB discussions about ubuntu-fan and SRUability 15:17 * discussions around KVM support for ppc64el on 14.04 15:17 * discussions around UEFI capsule update support 15:17 * finished paperwork for new Java hire, who will be starting July 27 15:17 * continuing search for Foundations engineers 15:17 (done) 15:17 short week due to usa holiday and pto 15:17 more python 3.5 test rebuilds, with patches going upstream and into debian as appropriate 15:17 python issue #15014 fixes RFC 4952 AUTH initial response (py3.5 fix for regression breaking packages in the wild) 15:17 more git-dpm conversion work for dpmt 15:17 lazr.delegates 2.0.3 (upstream) and 2.0.3-1 (debian) for launchpad related fix 15:17 zope.interface 4.1.2-1 new upstream and cross build fixes 15:18 reported LP: #1473093 (checkbox-ng for py3.5) 15:18 Launchpad bug 1473093 in Next Generation Checkbox (CLI) "tests fail on python3.5" [Critical,Confirmed] https://launchpad.net/bugs/1473093 15:18 --done-- 15:18 o/ 15:18 - Landing team work, silo coordination, preparing landing e-mails 15:18 - Manually prepare languagepack packages for most important languages 15:18 - Prepare livecd-rootfs for youtube scope removal 15:18 - Coordinate device and custom tarball landings for OTA-5 15:18 - Coordinate landings for OTA-5, image preparations 15:18 - Documenting the new release schedule for OTA upgrades 15:18 - RTM status meetings and discussions 15:18 - Gathering feedback regarding derived vs. overlay 15:18 - Discussions regarding fixing the broken wily language-packs 15:18 - Selective work on triaging the appmenu-qt5 bug with hiding windows 15:18 - Cleanup in assigned bugs 15:18 - Various changes and improvements to commitlog generation infrastructure 15:18 - Discussions about the setup of new channels 15:18 (done) 15:19 doko: 15:19 - announced the GCC 5 transition 15:19 - reviewd and fixed 30 GCC 5 ftbfs in Debian, merged to Ubuntu 15:19 - populated https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/landing-016/+packages 15:19 - getting angry with unmerged packages since 2013 ... trafficserver 15:19 - python3.5 beta3 15:19 (done) 15:20 * infinity checks if trafficserver is his... 15:20 unmerged packages> amen 15:20 Oh look, it is. 15:20 *sigh* 15:21 doko: Hint taken. :P 15:21 ;-P 15:21 those are the ones which really drive me away from fixing -proposed stuff in some cases (becoming TIL on even more crappy packages) 15:21 :) 15:22 pitti: Yeah, it's a vicous cycle. Cause you end up TIL, then you don't have enough time for merges, then someone else gets angry with you for not merging, then they get annoyed enough to merge it themselves, then they end up being the one who lags, repeat. 15:22 alrighty - any questions over status? 15:22 infinity: heh, yes; we still suffer a lot from the old years where we tossed just about anything into ubuntu that people threw at us.. 15:22 fwiw I've asked bdmurray to work on getting better reporting out of MoM so that we can actually see "merges I'm responsible for, sorted by age" 15:23 bdmurray: can you add "merges i will never get to, let's just be honest about it"? 15:24 infinity, pitti: would be nice to track autoremovals in unstable, and then have a separate section for these in update_excuses 15:24 slangasek: If I could get age out of "grep-merges", that would do it. 15:25 infinity: heh, never heard of 'grep-merges'. But if the main report would list the age, presumably that would give you this? 15:25 doko: Yeah, looking at trafficserver, it probably wants a removal from wily and then a sync after the gcc-5 transition, so it can FTBFS. :P 15:26 slangasek: Yeah, not sure where grep-merges pulls from, but I assume it's the same report you're talking about. 15:26 merges.ubuntu.com/main.html etc 15:27 [TOPIC] Team changes 15:28 I mentioned earlier in the meeting that stgraber was probably not around... there's a specific reason for this :) 15:28 noooo! 15:28 as many of you have seen, stgraber has been working full time on lxd together with folks in CDO for quite a while now 15:29 this situation is now being formalized - Stéphane is now the LXD technical lead, reporting on the server team! 15:29 Oh no! 15:29 And now I'm the shortest member of the team. This is unacceptable. 15:29 so congratulations to stgraber on the new role, and we wish him well - while making no promises not to continue harrassing him with Foundations questions ;) 15:30 The foundations are cracking! 15:30 this means that, once the paperwork goes through, we will be hiring for not one, but two Foundations engineer positions 15:31 infinity: ^ you have a way to rectify that then :) 15:31 so pester all your brilliant and short friends to apply 15:31 hmmm ;) 15:31 slangasek: is one of those the java dude? 15:31 barry: this would be two, in addition to javadude (whose name is Tiago) 15:31 oh! we have a winner! cool 15:32 Woohoo 15:32 Damn, I'd like to have more time for some system-image server work, grrr 15:33 [TOPIC] AOB 15:33 anything else this week? 15:34 after spending the fourth or so day on cleaning up after "we messed up swap space config in the installer years ago" 15:34 I wondered if we should just stop doing it completely -- it's 2015.. 15:34 pitti: yeah, i still see that on some of my machines ;) 15:34 kirkland mentioned "swapspace", which dynamically creates swap files on demand on your root partition (or perhaps /var) instead of eternally blocking an entire partition 15:35 does anyone have experience with that? 15:35 I don't think "dynamically" creating swap partitions sounds like a great idea to me 15:35 pitti: I have no experience with userspace swapfile magic like that, but I might be inclined to agree that switching from partitions to files for desktop installs would be reasonable. 15:35 Maybe. 15:35 I mean, we need to fix ecryptfs-setup-swap either way 15:36 there was discussion long ago about switching from swap partition to swap file; I don't know why it was never done 15:36 but as a matter of fact, everyone who installed with "encrypt my home dir" between perhaps precise and vivid had no swap at all 15:36 pitti: I'm missing context - what's the problem with ecryptfs-setup-swap? 15:36 slangasek: there were various bugs in that and ubiquity which prevented having swap space at all, or in the case of LVM having unencryted plus encrypted where the unencrypted one won 15:37 erm? I thought the desktop LVM+crypt recipe gave you a single encrypted VG and nothing else 15:37 i. e. we already seeem to have a non-negligible user base with no swap at all, and not seem to have too many complaints 15:37 slangasek: no, not LVM crypt; unencrypted LVM plus "encrypt my home dir" (ecryptfs) 15:38 ah 15:38 or just plain simple partitioning plus ecryptfs 15:38 pitti, just keep in mind that no swap at all makes the kernel behave differently ... 15:38 so the decision of whether to enable swap at install time should in no way be based on whether the average user noticed they didn't have swap 15:38 (in case you consider dropping swap altogether) 15:39 the average user shouldn't need to know what swap is, and the decision of whether to enable it for the user should be based on whether it gives the best user experience 15:39 ogra_: well, it wasn't an active decision -- those were years old bugs in ecryptfs/ubiquity which only surfaced with the systemd transition 15:39 is it even possible to go back and fix all those installs that effectively have no swap, say in some kind of update? 15:39 pitti, right, it looked to me you are discussing a future default though 15:39 right, so there's various alternatives: only configure swap partition with < 4 GB of RAM 15:39 or do away with having to decide about static swap at installer time and do dynamic swap 15:40 pitti: where does this 4GB number come from? Sounds arbitrary to me :) 15:40 ogra_: yes, it made me wonder whether "swap partition" is still a good answer 15:40 slangasek: exactly 15:40 pitti: I use far more than 4G normally. :P 15:40 OTOH, creating 32 GB of swap space on a 16 GB RAM system is just an utter waste 15:40 (by the old "twice your RAM" rule of thumb) 15:40 I have 8GB of RAM, and I have a swap partition, and I want a good reason if you're going to take it away ;) 15:40 (not sure what partman does, though) 15:40 partman-swap is a little smarter than that. 15:40 pitti: that's clearly not the right rule of thumb for high-mem systems, but that means we should fix the rule of thumb, not remove swap 15:41 I think it's "twice your RAM up to X, then a set value" or something. 15:41 yeah, I'm sure it does, but I wanted to get a feeling what you think about replacing it with dynamic swap files 15:41 pitti: take out the "dynamic" and yes 15:41 any swap file is "dynamic" in the sense that the user can "easily" resize it 15:41 I think files are probably the right answer (except maybe for servers, but serious server people tend to write their own recipes anyway). 15:41 but we don't need to support it being dynamic initially 15:41 well, "swap file" gets rid of this ever-breaking ecryptfs-setup-swap; but "dynamic" is really half the point here 15:42 (FWIW, I don't know what swapfile does exactly, I just got told an hour or so ago by kirkland that it's on-demand in some fashion) 15:42 pitti: How does dynamic work? When we're hitting the wall, does it just add another swapfile, and delete it when we don't need it anymore? 15:42 err, "swapspace" is the package, not swapfile 15:42 pitti: I think it'd be a good idea if you chatted with cjwatson to understand what the blockers were for us never having switched to swap files (aside from ENOTIME) 15:43 Small, stable system add-on that continuously and automatically adapts 15:43 slangasek: I'd guess ENOTIME, and also fragmentation/speed concerns on rotary disks. 15:43 available virtual memory space to your actual memory needs. Claims disk space 15:43 for use as swap space when needed; frees it up for use by the filesystem when 15:43 not needed. 15:43 well, the other thing is I don't want to "own" this project -- I'm oversubscribed already 15:44 but since this doesn't seem to be unanimous, I guess I just crawl back into my corner :) 15:44 (it really seemed to me like a case of "OMG nobody threw this out yet?") 15:44 but if people are still using it, so be it 15:44 it would be nice to remove swap partition setup goop from ecryptfs-utils 15:45 (thanks for fixing ecryptfs-setup-swap, pitti :) 15:45 I don't think we can do away with swap completely 15:45 tyhicks: fingers crossed that this is the last bug :) 15:45 pitti: I think there's a consensus for moving to a swap partition. I think having a *daemon* trying to manage the size of your swap at runtime is crazypants 15:45 sorry, moving to a swap /file/ 15:45 slangasek: file? ah yes 15:45 pitti: where would the swapfile be stored? does swapfile handle encryption? 15:46 tyhicks: I don't know anything about swapspace, so I don't know 15:46 kirkland looked into it some time ago and felt that it was worth taking a look 15:47 but if it's anything like a swap file it would just use the underlying fs' encryption 15:47 but encrypted home is per-user 15:47 well, not *that* encryption (ecryptfs) 15:47 we can't put the swapfile in /home/foo/ 15:47 e-s-swapspace is cryptsetup 15:47 ah, ok 15:47 no, should probably be in /var/cache or so 15:48 or maybe swapspace is even just using unallocated blocks 15:48 Obviously, you should just put your swapfile in a tmpfs, so it's zapped on power loss and you don't care about encryption. Duh. 15:48 infinity: that's a *great* idea! 15:48 pitti: I KNOW. 15:48 what was the name of that? memdouble or so (from DOS times) 15:48 I'm not sure tmpfs has the right semantics for this. Maybe we should use overlayfs for it 15:48 We have that too, it's called zram! 15:48 And works quite well. 15:49 and then you recursively use it 8 times, problem solved 15:49 that way we only have to store the delta of the memory 15:49 ok, I think we're through with this discussion :) 15:49 "tmpfs puts everything into the kernel internal caches and grows and 15:49 shrinks to accommodate the files it contains and is able to swap 15:49 unneeded pages out to swap space." 15:49 https://www.kernel.org/doc/Documentation/filesystems/tmpfs.txt 15:49 infinity: I know you were joking, but that would make for a fun test :) 15:50 tyhicks: Yeahp, and swapping to yourself is awesome. 15:50 #endmeeting Generated by MeetBot 0.1.5 (http://wiki.ubuntu.com/meetingology)