Time Nick Message
12:50 pdurbin this is really cool. shows pep8 (python style guide) violoations: http://jenkins.saltstack.org/job/pep8/violations
12:51 pdurbin see also http://irclog.perlgeek.de/salt/2012-05-03#i_5532745
12:57 pdurbin westmaas: it's too bad https://jenkins.openstack.org requires login
12:57 SEJeff It requires a launchpad.net login, which is free fwiw
12:59 pdurbin SEJeff: sure, but it's nice that http://jenkins.saltstack.org is open
13:00 SEJeff indeed
13:01 pdurbin i love the rain clouds: http://jenkins.saltstack.org/static/71a03e0e/images/32x32/health-00to19.png
13:01 SEJeff This is one of my favs: http://cloud.github.com/
13:02 pdurbin heh. nice
13:02 SEJeff Move your mouse around. It uses the jquery parallax plugin. Super fancy stuff
13:06 pdurbin yep. very nice
15:59 mattdm Your ideas are intriguing to me and I wish to subscribe to your newsletter.
16:03 SEJeff Doh!
16:05 mattdm Right. :)
16:05 mattdm Anyway, I'm Matthew Miller from Academic Computing at SEAS. I may have met some of y'all before.
16:05 mattdm And this does look very cool.
16:07 pdurbin mattdm: welcome!
16:08 pdurbin now we've talked about seas on two days :) http://irclog.perlgeek.de/search.pl?channel=crimsonfu&nick=&q=seas
16:09 mattdm that seems to include today :)
16:09 pdurbin yep :)
16:09 pdurbin but we think about you often ;)
16:09 pdurbin oh oh
16:10 pdurbin and not everyone here knows what seas is
16:10 pdurbin crimsonfu is not a harvard thing
16:10 mattdm i thought we were world famous and stuff
16:10 pdurbin despite the name
16:10 mattdm SEAS is Harvard School of Engineering and Applied Sciences
16:11 pdurbin mattdm: thanks
16:11 mattdm the Academic Computing group that I work for is separate from the IT group; we do everything (?) related to instruction and research
16:11 mattdm with some infrastructure where we had to build it ourselves to support that.
16:11 pdurbin here, for E-Z clicking from the logs: http://seas.harvard.edu
16:12 mattdm Goal is to do less infrastructure, more cool stuff.
16:12 pdurbin less infrastructure!
16:13 pdurbin don't join ##infra-talk then ;)
16:13 pdurbin actually do. some great people there
16:14 mattdm Right I'm not opposed to infrastructure. Sommmmeone has to do it. :)
16:14 pdurbin that's what puppet is for ;)
16:15 mattdm yeah.
16:15 mattdm http://acops.seas.harvard.edu/puppet-presentation/slides.html#slide1
16:15 mattdm (first part of that is a generic puppet overview; second part is what we're doing)
16:15 pdurbin hmm! haven't seen that link before
16:16 pdurbin why is "Academic Computing" in quotes? :)
16:16 mattdm Oh! And in keeping with the radical transparency notions that this group is all about:
16:16 mattdm https://code.seas.harvard.edu/puppet
16:17 pdurbin reminds me of the "Blog" of "Unnecessary" Quotation Marks - http://www.unnecessaryquotes.com
16:17 shuff yay! very nice
16:17 mattdm https://code.seas.harvard.edu/puppet/puppet/trees/develop
16:18 mattdm I think it's in quotes because of some sort of markup language fail
16:18 * pdurbin runs git clone git://code.seas.harvard.edu/puppet/seas-puppet-core.git
16:18 pdurbin hmm. `ack -f -a | wc -l` is 15 but it's a start!
16:18 mattdm Presentation was written in restructured text. Lars has an unhealthy love for restructured text.
16:19 pdurbin i've open sourced my configs for my home stuff: http://git.greptilian.com/?p=salt.git
16:20 mattdm The seas-puppet-core package is not the interesting part.
16:20 pdurbin not to steal shuff's thunder but he linked to http://forge.puppetlabs.com/huit/pam_access yesterday
16:20 mattdm Take a look at the second URL up there:
16:21 mattdm git clone http://code.seas.harvard.edu/git/puppet/puppet.git
16:21 SEJeff Anything interesting in your fork?
16:22 mattdm is not a fork. is collection of puppet modules we use to build our stuff.
16:22 pdurbin mattdm: 576 files! that's more like it. thanks!
16:22 pdurbin mmmm kvmcluster
16:23 mattdm yeah; speaking of infrastructure :)
16:24 pdurbin mmmm gmond
16:24 mattdm We're working on moving everything possible, including HPC compute, over to running under kvm
16:24 pdurbin me too!
16:24 pdurbin well, not really
16:24 pdurbin but sunsetting vmware anyway
16:25 SEJeff mattdm, Really? HPC compute under KVM? Do you use paravirt clocks and io?
16:25 mattdm For almost all of our use the flexibility is more important than io performance, and for situations where that's not the case the answer is usually heyyyyy, we know this other thing at harvard.....
16:25 SEJeff And you don't have problems with too many cs impacting overall system perf?
16:25 SEJeff Ah gotcha
16:26 * SEJeff works in a HPC-like low latency environment where perf is _everything_
16:26 SEJeff we take latency over throughput, power efficiency, or cost always
16:27 mattdm Absolutely. And there's definitely stuff that needs that (including here), but for the majority other considerations are bigger.
16:27 SEJeff Gotcha
16:27 mattdm And for CPU-bound workloads, the performance hit is only a few percent, which is pretty clearly a win overall.
16:27 SEJeff Academia would be interesting. /me works in finance
16:28 SEJeff Yes absolutely
16:28 mattdm So how do I get myself on http://crimsonfu.github.com/members/ ? Is there a secret initiation?
16:28 SEJeff What is the flexiblity win, do you use sheepdog or anything fancy like that for replication? lxc is starting to really come of age. The first serious user namespace patches *finally* hit lkml a few months ago
16:28 pdurbin jeez, we just met
16:28 pdurbin send us a pull request ;)
16:29 mattdm pdurbin: heh ok.
16:29 pdurbin you too, SEJeff
16:29 pdurbin though i'm warming up to you
16:29 mattdm SEJeff: ability to deploy different compute environments for different users.
16:30 SEJeff pdurbin, Nah I'm an acquired taste. Typical Sysadmin really who now codes more than does ops work.
16:30 mattdm We have a lot of peculiar, "interestingly-written" code which people want to run.
16:30 SEJeff mattdm, And containers aren't an option? Just curious. Those types of environments are fun to build
16:31 mattdm Containers might be an option. What we really want to do, though, is build a generic environment which can run on our KVM _or_ our new OpenStack deployment _or_ whatever wider Harvard infrstructure _or_ be pushed into EC2.
16:32 pdurbin westmaas: ^^ openstack
16:32 SEJeff kvm makes a lot of sense then. Good stuff
16:33 SEJeff Kind of like what richard jones (the febootstrap and virt-* redhat guy) was doing with the virt-sandbox stuff
16:33 SEJeff run 1 binary in a thin provisioned kvm vm or lxc container: http://fedoraproject.org/wiki/Features/VirtSandbox
16:34 pdurbin speaking of kvm, whorka seemed a bit surprised that live migration "just works" on stock centos: virsh migrate --live --verbose $DOMAIN_TO_MIGRATE qemu+ssh://$TARGET_KVM_HOST/system
16:34 pdurbin no downtime. works great
16:35 pdurbin i'm sure rhev is more slick though
16:35 pdurbin and oVirt, etc.
16:35 SEJeff :)
16:35 SEJeff oVirt is good stuff
16:35 SEJeff Did anyone here (other than me) go to the oVirt workshop in San Jose?
16:36 pdurbin SEJeff: we're on the wrong coast, unlike you
16:36 SEJeff sadness
16:37 pdurbin was it awesome? i'll take your answer offline. going to lunch
16:37 mattdm I just submitted a request to go to LinuxCon in San Diego in August, though....
16:37 SEJeff The cisco campus is really impressive actually. Much better than the MIT media lab for the GNOME boston summit imo.
16:37 SEJeff Yeah it was a great mini-con
16:37 mattdm SEJeff: hey I know you from LWN comments :)
16:38 SEJeff indeed. It is a unique name. I'm an ass on lwn at times as well :)
16:39 mattdm I always think you must be related to SELinux in some way
16:39 SEJeff You are a wise one
16:39 mattdm so it is?
16:39 SEJeff Security Enhanced Jeff. Comp Sec / penetration testing is how I learned 'nix
16:39 SEJeff Yes
16:39 mattdm very nice
16:40 mattdm anyway. i have a meeting. lunch meeting, alledgly, but it's running out of lunchtime. see you all later.
16:40 * SEJeff waves
17:47 * pdurbin looks for skype in pidgin. slides over to a mac
18:01 SEJeff A gem from mdehaan, the cobbler and ansible author: http://slipsum.com
18:41 pdurbin woof. too long for me to really absorb, but cool that openstack has their meetings in public: http://eavesdrop.openstack.org/meetings/openstack-meeting/2011/openstack-meeting.2011-06-14-20.03.log.html
19:23 teancom I'm having a problem moving from puppet 2.6 -> 2.7. I've been finding plenty of places in my configs where I was not wrapping variables in curly braces ( foo-$var-bar vs foo-${var}-bar ), but I don't think this is the same problem. My configs and error output at https://gist.github.com/2587908
19:25 pdurbin teancom: do you have a "template" directory?
19:25 teancom Yes
19:25 teancom templates (with an S), that is
19:25 teancom And I should say, this works just fine under 2.6.9
19:25 teancom 2.7.14, not so much
19:26 pdurbin in our code we have template with no s
19:26 pdurbin but i'm not our puppet guru. . .
19:26 teancom Weird, I thought that was a built-in, like 'manifests' or 'files'.
19:29 pdurbin hmm, i see templates at https://code.seas.harvard.edu/puppet/puppet/trees/develop/templates but template at https://code.seas.harvard.edu/puppet/puppet/trees/develop/modules/template
19:30 pdurbin teancom: you could clone their repo if you want to poke around: http://irclog.perlgeek.de/crimsonfu/2012-05-03#i_5534116
19:30 pdurbin (this channel is logged, by the way)
19:31 teancom Well, I don't *think* this is a problem of a mis-named directory. The error shows that it is parsing the erb template, and then having an error. It isn't failing to find the template at all (which is what I would expect in that case).
19:34 pdurbin so the thing you changed was upgrading puppet from 2.6.9 to 2.7.14
19:35 pdurbin i went to a talk by a puppet dev and he recommended sticking with whatever version of puppet is in their "enterprise" product
19:36 pdurbin which i guess is 2.7.12, based on what i'm seeing at http://yum.puppetlabs.com/enterprise/sources/2.5.1/source/el/6/SRPMS/ . so you're not *too* far ahead. . .
19:39 pdurbin teancom: can you parse the erb template outside of puppet? with just ruby?
19:40 teancom Yeah. We have pre-commit hooks for all .erb files that run them through erb -T.
19:41 pdurbin i guess i would try downgrading puppet
19:42 pdurbin maybe back it down to 2.7.12 first, since that's what's in enterprise
19:42 pdurbin for what it's worth, we're on 2.7.13
19:43 teancom I'll give that a shot, if what LinuxJedi suggested in #salt doesn't work out.
19:47 pdurbin http://irclog.perlgeek.de/salt/2012-05-03#i_5535308 . right. http://stackoverflow.com/questions/6765576/puppet-best-way-to-collect-data-from-custom-types-to-be-used-in-templates . ok, well, please report back here too
19:57 shuff teancom: i've just taken a quick look at your gist, and nothing is immediately jumping out at me
19:57 shuff but stepping back a moment
19:57 shuff if you're managing RedHatish systems, let me suggest http://forge.puppetlabs.com/razorsedge/network rather than managing your static routing by yourself
19:58 shuff and since i see /etc/sysconfig/network-scripts i suspect you are :)
19:59 teancom Hrrm.
19:59 teancom Interesting!
20:05 shuff i've been using this module for a little while now; it's solid
20:13 SEJeff Not too too the salt horn _too_ much, but we just had a really impressive network module for rh-ish systems contributed
20:14 SEJeff Example states, it does some really intense stuff: https://github.com/saltstack/salt/blob/develop/salt/states/network.py#L12
20:14 SEJeff ethtool, routes, interfaces, bonding, etc
20:29 teancom SEJeff: you know if I could use salt, I would be...
20:54 Pax shuff: Question about the raxorsedge network module… So I'm using cobbler as my pseudo cmdb and I'm wondering if it might not be preferable to just write some calls to cobbler to present interface info as facts?
20:56 shuff i don't think that's an either/or proposition
20:56 shuff i think they'd work together really well
20:56 shuff for example, here's an example of a node config for our (non-cobblerized) environment:
20:57 SEJeff Pax, Ironically, I did something very similar to that for $real_job, but can't open source it. I can however point you in the right direction for the cobbler api if you have any questions
20:57 shuff node "stomp.unix.fas.harvard.edu" {
20:57 shuff class {
20:57 shuff "stomp_server":
20:57 shuff net_static => {
20:57 shuff "eth0" => {
20:57 shuff "macaddress" => $macaddress_eth0,
20:57 shuff "ipaddress" => "10.255.19.245",
20:57 shuff "netmask" => "255.255.255.0",
20:57 shuff "gateway" => "10.255.19.1",
20:57 shuff }
20:57 shuff },
20:57 shuff environment => "shuff";
20:57 shuff }
20:57 shuff }
20:58 SEJeff Can you get that data out programatically from puppet, or do you just apply that manifest and then use facter?
20:58 shuff there's no reason why those ipaddress, netmask, gateway params shouldn't instead be, i dunno, $ipaddress_eth0
20:58 SEJeff gotcha
20:59 Pax yeah exactly what I was thinking actually! Calling the XMLRPC and creating $IP_bond0 or whatever
20:59 shuff yup
20:59 SEJeff Pax, Are you ok with python?
20:59 Pax cause re-defining it in another place seems cumbersome, and I'm lazy
20:59 Pax SEJeff: I love me the python
21:00 Pax oh…god… I actually said that *shame*
21:08 SEJeff You could have said ADA, clojure, or something more obscure and easier to ridicule you for
21:09 shuff who doesn't love erlang? can i get an amen?
21:10 SEJeff Actually you can
21:10 shuff i should mention that i am thrilled with rabbitmq and couchdb
21:11 Pax I broke up with erlang a while back
21:11 shuff but much like C, i am glad that other people like to write erlang so that i don't have to
21:12 SEJeff Riak <3
22:09 pdurbin_m riak sounded cool from the floss weekly episode. couchdb too. I've at least touched couchdb
22:09 SEJeff couchdb is awesome for small stuff
22:10 SEJeff but it was never designed for massive zomg scale. Just look at the reasoning behind why Canonical is dropping it for Ubuntu One
22:10 SEJeff Riak was designed for that size stuff from the get go
22:10 SEJeff But as a shooting buddy with a former basho employee, I might be biased.
22:21 pdurbin_m right. reliability too, from what I understand. clustered out of the box
22:22 SEJeff pdurbin_m, Thats probably one of the coolest things about it. It is similar to elasticsearch
22:22 pdurbin_m I don't really have use case for either one... yet
22:22 SEJeff You configure a new node to point to the existing cluster, it joins and everything rebalances automatically
22:22 SEJeff Sure
22:23 pdurbin_m maybe I could store my tweets in couchdb. it's already JSON from the twitter API
22:25 pdurbin_m I was gonna just throw them in git: http://git.greptilian.com/?p=data-liberation.git
22:25 SEJeff http://www.h-online.com/open/news/item/Canonical-dropping-CouchDB-from-Ubuntu-One-1382809.html if you're interested
22:25 SEJeff Or why not just mirror them to a statusnet install or ident.ca?
22:27 pdurbin_m will statusnet pull in old tweets?
22:30 pdurbin_m of course, I already have this for old tweets: http://thinkup.greptilian.com/?v=tweets&u=philipdurbin&n=twitter
22:38 pdurbin_m hmm, via that article: U1DB, a synchronized database https://blueprints.launchpad.net/ubuntu/%2Bspec/desktop-p-u1db
22:39 SEJeff tweets are extra entrophy for a rng
22:39 SEJeff not much else
23:55 pdurbin the problem with twitter is all the tweets