IRC Logs for #crux-devel Monday, 2015-03-30

prologicjaeger: thanks I figured it out in the end00:11
prologicturns out that if you have a zfs rootfs/boot00:11
prologicthen loading the zfs module auotmatically brings the pool online00:11
prologichowever if you don't - as in my case00:11
prologicthen you have to manually import the pool via zppol import00:12
prologicI'm just doing zpool import -a in /etc/rc.local00:12
jaegeryou'd need an initrd for the former00:19
prologiccorrect :)00:26
prologicdo you have a zfs root/boot fs?00:26
jaegernot anymore, I did once in the past to see if it could be done easily00:28
jaegerit worked fine00:28
prologictbqh I don't see the point myself00:29
prologicso I'm keeping my os/rootfs separate00:29
jaegerI did it more as proof of concept than out of any need00:33
jaegerIf you want a redundant rootfs, though, it's no worse than raid, lvm, btrfs, etc.00:34
*** Romster has quit IRC00:36
*** sudobaal has quit IRC00:37
prologicI just think that a rootfs can be replaced00:38
prologicdata probably can't :)00:38
jaegerI don't understand what you're saying, I guess00:42
jaegerBackups are important but redundant rootfs isn't a bad thing, either00:42
*** Feksclaus has quit IRC01:07
*** Romster has joined #crux-devel01:46
Romsterif we do remove python from mesa3d then i'll have to go around a ton of ports and check what needs python added to it's dependencies.01:59
Romsteror maybe not xorg-libevdev uses python as well.01:59
Romsterxorg-libevdev doesn't link with python is that build time dep?02:00
Romstertalloc uses python, libxml2-python yep python, xorg-xcb-proto ... python, there is no point in removing python form mesa3d at this rate.02:02
Romsterafk work02:02
Romsteroh and stance on python3 do we introduce it or do i jsut add 2 new packages to back port python 3 features that i need now for a port in contrib?02:02
jaegerI assume when you say "introduce" you mean "replace with"02:07
jaegerbecause python3 could go into contrib at any time, right? with altinstall02:08
Worksterwell adding python3 to opt02:30
Worksteralong with setuptools302:30
*** mavrick61 has quit IRC02:38
*** mavrick61 has joined #crux-devel02:39
*** nilp has joined #crux-devel05:41
*** heroux has quit IRC06:09
*** heroux has joined #crux-devel06:11
*** Feksclaus has joined #crux-devel06:15
*** Workster has quit IRC06:45
*** heroux has quit IRC07:00
*** heroux has joined #crux-devel07:02
*** Feksclaus has quit IRC07:08
*** heroux has quit IRC07:08
*** heroux has joined #crux-devel07:09
*** heroux has quit IRC07:26
*** heroux has joined #crux-devel07:28
*** sudobaal has joined #crux-devel11:08
*** nilp has quit IRC11:10
*** nilp has joined #crux-devel11:10
*** nilp has quit IRC11:11
*** heroux has quit IRC12:47
*** heroux has joined #crux-devel12:49
jaegerI/O error : Attempt to load network entity
jaeger/dev/shm/ports-work-compton/src/compton-0.1_beta2/man/compton-trans.1.xml:2: warning: failed to load external entity ""12:52
jaegerD DocBook XML V4.5//EN" ""12:52
jaegerWhy does this still fucking happen? god damn, docbook12:52
jaegerI think I see at least part of the problem:
jaegerseems like xmlcatalog calls in the Pkgfile are overwriting the public entry every time13:46
jaegerinstead of adding13:46
frinnstheh yeah. I have a debian vm for generating the btrfs manpages13:47
frinnstfucking docbook13:47
frinnsthave you managed to get xmllint to consume all your ram yet?13:49
jaegernot yet but I do have 32G in this machine. We'll see13:49
frinnstits awesome. easily eats 16gig ram and ~20gb swap before OOM'ing13:49
jaegerI think xmlcatalog is trying to do some short-circuiting here because after all the add public commands and the rewriteSystem and rewriteURI commands there are no public entries left13:50
jaeger  <rewriteURI uriStartString="" rewritePrefix="file:///usr/share/xml/docbook/xml-dtd-4.1.2"/>13:51
jaegerthat's the only line that's left from the 4.1.2 section13:51
frinnstyour patience is longer than mine, thats for sure13:51
frinnstI gave up pretty quick13:51
jaegerthen if I run the 4.5 section manually it removes the 4.1.2 section13:52
jaegerthis explains why it never fucking works because 4.1.2 is run last13:52
jaegerthus we basically only have 4.1.2 entities13:52
jaegerso xmllint or xmlto or whatever that uses "--nonet" will fail when it tries to find 4.513:53
jaegerit could find 4.1.2 but it's looking for 4.5 instead and --nonet says "you can't go to the internet to get it"13:53
jaegergrrr... I'm going to do my best to fix this shit, I'm so tired of docbook not working properly13:54
jaegerfrinnst: would you mind doing me a favor with your debian VM?13:56
jaegerwould you run the commands in that paste and see if debian's xmlcatalog does the same thing? It creates /tmp/docbook-xml so doesn't mess with your system catalog or need root privs13:57
frinnst      ?13:57
jaegerI expect the xml file will be somewhat different, I'm just really curious if each --add overwrites the previous one13:58
jaegerit's like 6 commands, most of that was output13:58
frinnstjust a sec13:58
jaegerno rush at all, thank you13:59
frinnstno ssh, so have to type a lot :)14:00
jaegerdoh, sorry14:01
frinnstthats the output from all 3 commands14:02
frinnstlooks nothing alike..14:03
jaegeryeah, though it seems to have the same behavior14:03
jaegernotice the last line, the <public> kley14:04
jaeger-//OASIS//DTD DocBook XML v4.1.2//EN got replaced with -//OASIS//DTD DocBook XML CALS Table Model v4.1.2//EN14:04
jaegerwhich may be the intended behavior but I'll be damned if I understand it14:04
frinnstnobody understands xml14:04
jaegerthanks for testing, appreciate it14:06
jaeger <-- this /etc/xml/catalog seems to behave in a non-stupid fashion14:54
jaegerfrinnst: want to try my new version of docbook-xml with the btrfs-progs manpages in crux?15:04
jaegerit seems like xmlcatalog can't be trusted to properly add individual entries but it seems to be ok with rewrite entries15:06
frinnst    [ASCII]  btrfs-property.xml15:35
frinnst    [XMLTO]  btrfs-property.815:35
frinnst    [GZ]     btrfs-property.8.gz15:35
frinnstglory glory15:35
jaegerseems promising :D15:36
jaegerI *think* it should work for older docbook versions as well but I'm not sure how to test it15:36
jaeger(older meaning the included versions that aren't 4.5)15:36
jaegerhrmm.... not sure how/why we create /etc/xml/docbook-xml but not specifically manipulate /etc/xml/catalog... while not using nextCatalog to include one in the other16:14
jaegermaybe something I could track down with strace but if things work I'm less inclined to dig16:15
*** Feksclaus has joined #crux-devel18:45
*** heroux has quit IRC18:48
*** heroux has joined #crux-devel18:50
prologicjaeger, ping?21:59
jaegerI'm hereish. what's up?22:06
*** Feksclaus has quit IRC22:30
*** sudobaal has quit IRC22:55
prologicoh hey just wanted to hear your opinins and experiences again23:24
prologicyou've had a zfs system for a while23:24
prologicno doubt you've incurred and suffered fragmentation23:24
prologicunless you've rebuilt your pool recently23:24
prologicsome folk in teh zfs community have gotten me worried that I may run into performance issues and forcing me to reconsider how I design the pool23:24
prologicI was going for a straight 10 disk raidz223:24
prologic2x raidz2 (groups of 5) would give me 2x the performance at the sacrifice of 5TB of storage23:25
prologicso I'm paranoid and seeking "experience based opinions"23:25
prologicI've tried all sorts of benchmarking on my system so far iozone, bonniie++, fio, dd, dd with frandom; turns out benchmarking is "hard" at best and not really universal23:26
jaegerI would say that it likely doesn't make much difference for a home user. If you're not pushing a heavy IO workload then it'd probably be fine either way. I scrub my pool once a month but even that is probably overkill23:30
jaegernote that I have no hard data to back that up, I just doubt that it matters for non-enterprise workloads23:30
prologicyeah ofc23:41
prologica colleage here at works has said not to worry about it so much either23:41
prologicso in terms of heavy IO workloads however23:41
prologicis there a rough idea of what that maens in practice?23:41
prologicthe kinds of workloads I'm looking at are23:41
prologicmythtv backend serving and capture23:41
prologicover the network obviously23:41
prologicmusic serving via mpd23:42
prologicand general file storage for /home over nfs23:42
prologicI honestly *can't* imagine the performance is going to degrade as badly as some might suggest23:42
prologicI *would expect* it to degrade quite badly if you were using it for VM(s) with high load (e.g: high traffic sites, high database I/O, etc)23:43
jaegerI don't even consider my ZFS appliance at work heavy IO23:48
jaegerhome workloads are a joke23:48
jaegerI use a few Kodi frontends and VMs on mine, it's never under any real load23:49
prologicand hasn't degraded significantly?23:55
prologicso are these folk just a bit paranoid23:55
prologicor talking about much much bigger work laods23:55
prologicmany of the folk I've talked to run BIG ZFS based sytesms23:55
prologic40+ disk arrays23:56
jaegernot in the least, heh23:56
prologicand some many thousands of disks23:56
jaegerI've got a couple with 96 disks each at work, 4 disks at home, heh23:56
prologicbut then if you have that many disks and bays and money23:56
*** greduan has joined #crux-devel23:56
prologicyou can afford to stripe and sacrifice storage23:56
prologicwhat's the 96 disk setup like?23:56
prologicstriped raidz, mirrored pairs/triples?23:57
jaeger <- this sorta thing. it's a big pool composed of 7ish disks in raidz2s23:57
prologicahh wow23:58
jaegeryou can configure it for less redundancy than that, they were a bit paranoid when setting it up initially23:59
jaegerwe could lose an entire expander or ~22 disks depending on which they were23:59

Generated by 2.11.0 by Marius Gedminas - find it at!