Camelia, the Perl 6 bug

IRC log for #crimsonfu, 2012-12-04

crimsonfu - sysadmins who code

| Channels | #crimsonfu index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:20 \etc\bin joined #crimsonfu
00:20 \etc\bin joined #crimsonfu
00:21 jgtimmer joined #crimsonfu
01:12 jgtimmer joined #crimsonfu
06:28 comptona joined #crimsonfu
08:13 Itkovian joined #crimsonfu
08:30 Itkovian joined #crimsonfu
08:56 boegel joined #crimsonfu
10:25 pdurbin joined #crimsonfu
10:28 Itkovian joined #crimsonfu
10:36 jgtimmer joined #crimsonfu
12:40 chasmo joined #crimsonfu
14:03 Itkovian joined #crimsonfu
14:13 pdurbin boegel: you around?
14:13 jgtimmer pdurbin: he's here
14:14 boegel boegel: I am
14:14 pdurbin i bet whorka will be un-idle in an hour or so. maybe you can explain a bit how easybuild might be helpful in deploying commercial apps
14:14 boegel pdurbin: I am :)
14:14 boegel pdurbin: sure
14:14 pdurbin (i mentioned easybuild at lunch yesterday)
14:15 boegel pdurbin: and how did they react? disbelief? :P
14:16 pdurbin heh. curious, i'd say
14:16 boegel sounds good
14:16 pdurbin do you even have generic build instructions for common commercial packages?
14:16 boegel pdurbin: I have a conf call scheduled in 45m, and I'll be leaving work in just over 2 hours, but let him ask his questions, I'll try and respond ASAP
14:17 boegel generic? well, once we support a certain package it's all very generic, yes
14:17 pdurbin oh, i don't think he's in a hurry. could wait until next week, next month :)
14:18 pdurbin huh. you have a page for this already, it seems: List of supported software packages · hpcugent/easybuild Wiki - https://github.com/hpcugent/easybuild/w​iki/List-of-supported-software-packages
14:19 jgtimmer pdurbin: we have generic binary and tarred(packed binary) support, which will basically just unpack the binary and move it to the install dir, and then check for /bin /lib dirs and make the modulefile
14:20 jgtimmer we also compile some commercial software from source
14:20 pdurbin i don't see Mathematica, MATLAB or Stata on your list
14:20 jgtimmer we have an internal one for matlab, haven't open sourced it yet since it currently contains our license keys hardcoded
14:20 jgtimmer will have to fix this
14:21 pdurbin do you call it a matlab build script?
14:22 jgtimmer mathematica and stata are not on our systems no, so we're not planing to add it any time soon, but we'll glady accept pull request, and I can sponsor people into adding support
14:22 pdurbin hmm, easyconfig. easyblock maybe
14:22 pdurbin matlab sounds like a good place to start, if you can open source your easyconfig thing or whatever you call it
14:23 jgtimmer easyblock == build script/recipe, easyconfig == configuration
14:24 pdurbin thanks :)
14:29 pdurbin here's a list of candidates for easybuild: http://support.hmdc.harvard.edu/kb-111​2/using_statistical_programs_overview
14:30 pdurbin but they have a good solution for open source apps
14:30 pdurbin i think easybuild might be a win for commercial apps though
14:30 jgtimmer pdurbin: https://github.com/hpcugent/easybuild/wi​ki/List-of-supported-software-packages/e​de46976d7367a86fe76ae79adba7b8e9fd9f118
14:30 jgtimmer this is what we have internally
14:31 pdurbin cool
14:31 pdurbin i'm glad to see a * next to MATLAB :)
14:32 jgtimmer the ones that aren't in the current easybuild need to be checked out, they are written for our old internal easybuild version, before we did the cleanup/partial rewrite
14:33 pdurbin ok, i guess i'd say... please let us know when your matlab recipe is open source
14:33 jgtimmer http://pastebin.com/qx7fDV8c
14:33 jgtimmer here you go
14:33 pdurbin nice
14:34 jgtimmer but this is an old version, it doesn't use some of the new features of the framework yet (modulegenerator etc) and has old style function naming...
14:34 pdurbin and do we need to install atlas to use it?
14:34 pdurbin was it atlas? some big package and didn't compile with gcc or something
14:35 pdurbin that didn't compile, i mean
14:35 jgtimmer no, it just takes the install script and runs this.
14:35 jgtimmer hmm, wel, it seems to depend on java...
14:36 jgtimmer and we don't have java public yet either, I think we still used the system java at the time that was written :-s
14:36 pdurbin yeah, it was atlas: http://irclog.perlgeek.de/cr​imsonfu/2012-10-05#i_6039387
14:36 jgtimmer ah
14:36 raprasad joined #crimsonfu
14:36 jgtimmer no, it doesn't use any compiler at al
14:37 jgtimmer so you don't need the goalf toolkit
14:37 pdurbin great
14:38 jgtimmer actually we don't use it on our systems eiher
14:38 jgtimmer we use the intel compilers and mkl libraries
14:38 boegel pdurbin: getting support for what we have internally out there is usually a matter of cleaning up our easyblocks
14:39 jgtimmer we added the gcc/openmpi/atlas/fftw toolchainf to have a fully open source toolchain
14:39 boegel pdurbin: and only time is an issue there, time which we don't have :)
14:39 pdurbin :)
14:41 boegel pdurbin: if someone is interested in installing software with EasyBuild for which we have legagy support, we can supply them the old easyblock, and answer questions on how to make it work
14:45 pdurbin ok
14:47 boegel pdurbin: the best way is just to open a new issue in https://github.com/hpcugent/easyb​uild-easyblocks/issues?state=open, then we can let the person know whether we have legacy support for it or not
14:49 jgtimmer pdurbin: ATLAS builds with gcc, it just fails in a vm because it tries to do smart things with the cpu
14:50 pdurbin oh
14:50 boegel pdurbin: yeah, that was the issue in the log you sent me
14:50 boegel pdurbin: don't blame EasyBuild for that
14:50 pdurbin short memory. sorry. that's why i log this channel :)
14:51 boegel pdurbin: :)
14:53 jgtimmer I can look into adding a less heavy demo, e.g., just a plain python
14:54 pdurbin jgtimmer: that'd be great. put in in vagrant even :)
14:54 pdurbin it
14:56 boegel jgtimmer: the only reason why our current quick demo is heavy is the fact that it includes building a compiler toolchain (which is a one-time pain)
14:57 pdurbin i demand instant gratification
14:58 boegel pdurbin: https://github.com/hpcugent/e​asybuild/wiki/Getting-started
14:59 boegel pdurbin: the gzip dummy example easyconfig just builds gzip with your system compiler
14:59 boegel pdurbin: if you can make that fail, them you might as well change jobs :P
15:01 pdurbin i am changing jobs: https://twitter.com/philipdur​bin/status/273229031082692608
15:02 boegel pdurbin: ah, well, ok, then you're allowed to screw up a gzip dummy build :P
15:03 pdurbin huh. this might be new. this example
15:05 jgtimmer yes, if the quick demo fails, you should read the getting-started and do step by step :-)
15:07 boegel_ joined #crimsonfu
15:07 boegel_ pdurbin: I made some changes to that wiki page, but basically this gzip example has been on there for a while
15:07 boegel_ pdurbin: I'm gradually cleaning up documentation, I'm not fully there yet
15:09 jgtimmer pdurbin: are you aware of http://numfocus.org/
15:09 pdurbin nope
15:09 jgtimmer they are a comunity looking into creating reproducible computing
15:10 jgtimmer data is a big part of that
15:10 jgtimmer the other part is recreating the software environment
15:11 pdurbin huh. sounds great
15:15 jgtimmer And I spoke to some people at continuum, and they were working on wakari, they plan to make a product for researchers to easily let other people redo their calculations
15:15 jgtimmer http://continuum.io/wakari.html
15:15 jgtimmer so might be a bit inline with what thedata.org wants to do
15:19 pdurbin jgtimmer: yes! sounds like it. thanks
15:19 raprasad_ joined #crimsonfu
15:22 jgtimmer congrats on your job btw
15:24 pdurbin thanks. one sec. you guys are gonna love this
15:29 pdurbin https://github.com/pdurbin/easybuild-vagrant
15:29 pdurbin boegel: gzip example works fine
15:32 jgtimmer pdurbin:
15:32 jgtimmer I'm not familiar with vagrant
15:33 jgtimmer what is that supposed to do? Install a vm whit environment modules in it
15:33 jgtimmer ?
15:33 pdurbin jgtimmer: http://wiki.greptilian.com/vagrant
15:34 pdurbin yes, demo of `eb /easyconfigs/gzip.eb` ultimately, in a VM
15:35 jgtimmer I noticed you skipped the step of installing environment modules, so they are already in your vm?
15:35 pdurbin no i didn't :)
15:35 jgtimmer ah, puppet
15:35 pdurbin https://github.com/pdurbin/easybuild-​vagrant/blob/master/manifests/init.pp
15:35 jgtimmer I skipped that step
15:35 jgtimmer it went to fast
15:35 jgtimmer :-p
15:36 pdurbin oh? you got it to work already?
15:36 jgtimmer nono, when reading the readme
15:36 pdurbin ah :)
15:36 jgtimmer but it was just one line
15:37 boegel pdurbin: cool, thx
15:37 pdurbin well, please let me know if you get it to work
15:37 pdurbin boegel: sure. please give it a try
15:37 jgtimmer it's an easy way of testing easybuild under different OS'es
15:38 jgtimmer but what would be any other benefit? :-p
15:38 pdurbin marketing
15:38 pdurbin you could have it default to ubuntu and use centos instead by setting an enviroment variable: https://github.com/puppetlabs/puppetlabs-​openstack_dev_env/blob/master/Vagrantfile
15:41 pdurbin it would be great if you guys would throw your matlab.eb in https://github.com/pdurbin/easybui​ld-vagrant/tree/master/easyconfigs test it, and let me know if `eb /easyconfigs/matalb.eb` works
15:42 pdurbin typo but you know what i mean, i hope
15:43 rruma joined #crimsonfu
15:52 boegel pdurbin: that would require our matlab.py easyblock to be shipped in EasyBUild
15:53 pdurbin boegel: do you understand the value of that easybuild-vagrant repo? the idea is that it's an easy way for people to try out easybuild in a completely clean environment
15:53 pdurbin they can blow away the vm with `vagrant destroy`
15:54 pdurbin and run `vagrant up` to get it back
15:56 rruma joined #crimsonfu
16:18 boegel joined #crimsonfu
16:19 boegel pdurbin: https://github.com/hpcugent/e​asybuild-framework/issues/372
16:19 boegel pdurbin: I'll have to go, I may be online later tonight
16:20 boegel pdurbin: w.r.t. the guy(s) who had some questions, let them join #easybuild here on FreeNode, we might have someone there to answer the questions
16:20 boegel pdurbin: if not, tell them to use the bot and the !tell feature, that's always nice :)
16:20 boegel pdurbin: ttyl!
16:45 JoeJulian I read that wrong at first and was about to go to #easybuild to have the bot "!tell future"
16:47 pdurbin wouldn't that be nice
17:05 sjoeboo joined #crimsonfu
18:10 comptona joined #crimsonfu
18:48 raprasad joined #crimsonfu
18:50 * pdurbin heads to a meeting with the gluster guys
18:51 semiosis cool
18:52 Itkovian joined #crimsonfu
19:38 pdurbin gluster was inspired by gnu hurd!
19:38 pdurbin crimsonfubot: lucky gluster gnu hurd
19:38 crimsonfubot pdurbin: http://download.gluster.org/pub/gl​uster/glusterfs/LATEST/CentOS/epel​-7/x86_64/repodata/primary.xml.gz
19:38 pdurbin bah
19:40 pdurbin "negative lookup cache"
19:40 pdurbin remember when we looked for something and didn't find it
19:42 jgtimmer joined #crimsonfu
19:42 pdurbin semiosis: they're singing your praises
19:42 pdurbin JoeJulian: you too
19:43 pdurbin huh. epel-7 ^^
19:45 semiosis pdurbin: POIDH
19:45 pdurbin semiosis: huh?
19:45 semiosis hehe never mind
19:45 pdurbin :)
19:45 semiosis @lucky POIDH
19:45 crimsonfubot semiosis: http://www.urbandictionary​.com/define.php?term=POIDH
19:51 jgtimmer joined #crimsonfu
19:59 pdurbin interesting that they plan to merge their tweaks to swift code back in
19:59 pdurbin westmaas: ^^
20:02 Itkovian joined #crimsonfu
20:03 pdurbin "An efficient WAN replication method with GlusterFS is to use the marker framework / queue using extended attributes to feed rsync a list of changed files, scales better than inotify" -- http://www.jebriggs.com/blog/20​12/02/svlug-story-of-glusterfs/
20:03 pdurbin i don't think i'd heard of the gluster "marker framework"
20:08 simmel What was the name of that distributed filesystem that recently was in the news?
20:09 simmel Ah, http://www.xtreemfs.org/
20:09 pdurbin simmel: uh. gluster? :)
20:10 simmel Xtreemfs looks interesting. It's like AFS without the quirks.
20:10 pdurbin "XtreemFS is an open source distributed and replicated file system for the cloud"
20:10 pdurbin cool
20:11 simmel Would be awesome with Kerberos support for Xtreemfs but I guess it's only a matter of time before that is a reality.
20:30 pdurbin yeah
20:48 larsks Grrr, it's java based.
20:51 JoeJulian pdurbin: Nice of them to add to my already overly inflated ego. :D
20:54 JoeJulian wrt epel7 which leads me to rhel7: "On the desktop, RHEL 7 will also feature the GNOME 3 user experience, which will be new to RHEL users." I do not envy Red Hat's support staff.
20:55 boegel joined #crimsonfu
21:31 pdurbin JoeJulian: huh. i guess you're quoting from this but it's from july: The Future of Red Hat Enterprise Linux 7 - http://www.serverwatch.com/server-trends/th​e-future-of-red-hat-enterprise-linux-7.html
21:31 pdurbin larsks: ask me in 6 months how i feel about java. gonna be staring at a lot of java soon
21:35 larsks pdurbin: My biggest complaint about Java is when I have something that will run under java 1.6.21 but not 1.6.20 or 1.6.22.  Or something like that.  Which I run into with alarming frequency.
21:36 larsks "WRITE ONCE RUN IN VERY SPECIFIC PLACES"
21:36 pdurbin heh
21:51 simmel My biggest issue with java is that all libs/products seldom have any or even remotely sane defaults.
22:01 semiosis +1
22:01 semiosis i call it 'insane by default'
22:02 semiosis larsks: that sounds awful.  what could possibly change between those minor releases?
22:03 miah who cares about defaults. our job is to trash defaults and make shit awesome.
22:04 semiosis miah: spoken like a true java dev
22:04 miah i mean. how many people actually run a server with out of the box default configurations.
22:15 e1mer joined #crimsonfu
22:15 e1mer joined #crimsonfu
23:27 jgtimmer joined #crimsonfu

| Channels | #crimsonfu index | Today | | Search | Google Search | Plain-Text | summary

crimsonfu - sysadmins who code