Time  Nick         Message
12:51 pdurbin      "it is possible, using koan, to specify which libvirt pool a given vm should end up in? or should I just point it at the directory directly?" -- [cobbler] using koan with libvirt pools... - https://fedorahosted.org/pipermail/cobbler/2012-June/007785.html
12:51 pdurbin      inquiring minds want to know, including me and sjoeboo
12:58 westmaas     pdurbin: http://docs.openstack.org/essex/openstack-compute/install/yum/content/
12:59 westmaas     also, I can't find anyone in rackspace land with the info you guys want about iscsi backed volume storage, but there is a thread ongoing on the mailing list about it
12:59 pdurbin      yeah? link?
13:00 westmaas     https://lists.launchpad.net/openstack/msg13009.html
13:00 pdurbin      westmaas: this is awesome. "You have a collection of compute nodes, each installed with a supported Linux distribution: Fedora 17, Ubuntu Server 12.04, RHEL 6.2, Scientific Linux 6.1 or CentOS 6 + CR distributions (continuous release ( CR ) repository)." -- http://docs.openstack.org/essex/openstack-compute/install/yum/content/assumptions.html thanks!
13:00 westmaas     np :)
13:05 pdurbin      agoddard: it sure sounds like openstack supports iSCSI... "* create an LVM pool  * create a volume in this pool (with nova command or horizon UI)  * share it with an iscsi target (the compute/volume node)  * attach it with an iscsi initiator (the VM)"
13:06 SEJeff       pdurbin, I suspect no wrt koan
13:06 SEJeff       Ask on #cobbler though
13:06 SEJeff       jimi is normally pretty decent about answering
13:06 pdurbin      i believe sjoeboo asked in #cobbler last night
13:08 pdurbin      "iSCSI support for Compute/Nova. Nova currently supports persistent volumes using ATA-over-Ethernet (AoE). This is to discuss adding iSCSI support as an alternative." http://wiki.openstack.org/IscsiBlueprint
13:10 westmaas     yikes, that was updated 2010
13:11 pdurbin      we have a bunch of iSCSI devices. Dell MD3000i's. wondering if OpenStack would work well with them
13:12 pdurbin      heh. first hit in google for 'openstack md3000i" is #crimsonfu chat logs :)
13:15 pdurbin      "Volumes must be pre-allocated on the iSCSI server, and cannot be created via the libvirt APIs." http://libvirt.org/storage.html
13:15 pdurbin      i think agoddard has been telling me this over and over :)
13:20 agoddard     ohai
13:21 pdurbin      agoddard: so xenserver manages your iscsi for you? it creates luns for your guest VMs and such?
13:21 agoddard     so, my understanding is that nova-volume takes a volume and then shares it out over iSCSI/LVM to nodes - but what I don't know is whether you can just bypass the nova-volume node and tell openstack to build LV's on a big iSCSI LUN itself
13:22 agoddard     pdurbin: yep, we pass big LUNs from the SAN to XenServer, it then throws LVM on them and when we ask for a disk, it carves us a new LV
13:22 pdurbin      that's pretty sweet
13:23 agoddard     the difficult part of that, I believe, is that with LVM, only one node can be making the changes, so XenServer manages this by doing all the storage operations with one node (the master) I *assume*
13:23 pdurbin      hmm. right
13:23 agoddard     otherwise if you accidentally the LVM by modifying VG's or LV's on different nodes at the same time, you could fjuk it up
13:23 pdurbin      i want openstack or whatever to manage my iscsi devices for me. create luns and whatnot
13:24 agoddard     the other way is to use CLVM, which keeps LVM metadata synced across the nodes, but then you lose stuff like snapshotting
13:24 pdurbin      hmm. ok
13:24 pdurbin      i mean, we may use this iSCSI hardware for something else anyway. i can probably let openstack just manage some other hardware the way it wants to
13:25 agoddard     I think openstack will manage some iSCSI devices, if you have a SAN that supports it, this would mean the SAN would need an API which openstack is a client of, it would tell the SAN to create a new LUN and then attach it to a machine - that would be totally killer, but I don't think the PowerVault's do that
13:26 agoddard     one way would be provisioning a LUN to a nova-volume box and having the nova-volume box re-share it as LVM over its own iSCSI connection to the compute nodes, but this seems a little backasswards
13:26 * agoddard   has never actually used OpenStack btw, so this is all just assumptions
13:26 pdurbin      cool, cool. i'll keep thinking about it
13:27 agoddard     I wonder whether you could set it up with nova-volume and then just tell the nodes that their iSCSI host is the SAN and not the nova-volume host.. but I dunno
13:27 agoddard     I'm also installing openstack in prod today, so I
13:27 pdurbin      sjoeboo: look at that, i linked to the patch from the ticket: Issue #79: koan --virt-image-type ? · cobbler/cobbler - https://github.com/cobbler/cobbler/issues/79#issuecomment-5873608
13:27 agoddard     * will have more answers
13:27 pdurbin      agoddard: cool. please keep us posted
13:27 agoddard     pdurbin: definitely
14:48 pdurbin      https://github.com/fasrc/cobbler/commits/qcow2_fix
14:55 agoddard     in puppet...
14:55 agoddard     if you have two nodes, let's say you want their mysql innodb memory to be relative to their total memory or something.. how would you go about it?
14:55 * agoddard   has never used puppet, is a chef fanboi
15:07 agoddard     westmaas: I totally missed this message the other day: http://irclog.perlgeek.de/crimsonfu/2012-06-07#i_5694456 might have some questions for you after we stand up some boxes today - imma use the rackspace cloud builders cookbooks
15:16 westmaas     cool
15:36 westmaas     agoddard: you guys a chef shop only?  One of the guys that used to work for me is building puppet manifests for fedora now that he works at redhat, if that would be useful
15:38 agoddard     westmaas: yep just chef
16:06 pdurbin      if i want to write a little web app that has an api is django a good choice?
16:07 * pdurbin    googles
16:08 SEJeff       pdurbin, What do you want to do with said api?
16:08 * SEJeff     has done this 3-4 times with django
16:08 SEJeff       and once with flask
16:08 pdurbin      "it looked like Django coupled with Piston was the quickly becoming the defacto standard for developing new API's in Python" Creating a very basic API using Python, Django and Piston | RobertShady.com - http://www.robertshady.com/content/creating-very-basic-api-using-python-django-and-piston
16:08 SEJeff       pdurbin, FYI piston was abandoned for the better part of a year or two
16:09 pdurbin      oh. bummer
16:09 SEJeff       just recently was picked up with a new maintainer, who has merged a few patches and not done a ton
16:09 SEJeff       However, piston is very stable and good software
16:09 pdurbin      in perl i'd probably use dancer
16:09 SEJeff       If you've never done piston before, I would point you towards tastypie
16:09 SEJeff       much better
16:09 SEJeff       And sing Tiny Dancer the entire time
16:09 SEJeff       http://tastypieapi.org
16:10 SEJeff       pdurbin, I can help you do a piston api as I maintain snowy
16:10 SEJeff       http://git.gnome.org/browse/snowy which happens to be a very big api service ontop of oauth + openid + piston + rest
16:10 pdurbin      SEJeff: so you agree with this answer for tastypie: http://stackoverflow.com/questions/656979/django-and-restful-apis/9519849#9519849
16:10 SEJeff       Yes I do
16:11 SEJeff       I've heard very good things about django-rest-framework as well, but tastypie has the most traction and it has the guy who wrote django-haystack behind it
16:11 SEJeff       He is a legend in the django community
16:11 SEJeff       His tag is funny too: toast driven development
16:11 SEJeff       making fun of TDD
16:12 pdurbin      http://django-tastypie.readthedocs.org/en/latest/tutorial.html#installation Python 2.4+ Django 1.0+ . reasonable requirements
16:13 SEJeff       Yup
16:13 SEJeff       The guy that wrote bitbucket.org (jespern is his online handle) wrote piston
16:13 SEJeff       It is fantastic
16:13 SEJeff       very complete
16:14 SEJeff       But when github started spanking bitbucket more or less, he was pulled to other things and piston lagged for years
16:14 SEJeff       Thats why both tastypie and django-rest-framework were created
16:14 SEJeff       pdurbin, If you need any help with any django stuff I'm game
16:17 pdurbin      SEJeff: this is pretty much the problem space. what the webapp would address: How do large organizations with large turnover and many systems perform account provisioning? - Server Fault - http://serverfault.com/questions/198389/how-do-large-organizations-with-large-turnover-and-many-systems-perform-account-provisioning
16:18 pdurbin      i just want to make sure it has an api
16:18 pdurbin      "Do not implement any product that does not provide an API." -- http://blog.websages.com/2010/12/10/jameswhite-manifesto/
16:19 SEJeff       Very very wise advice ^^
16:19 SEJeff       Being in crimsonfu, I can assume you're a systems engineer
16:19 SEJeff       but often system admins (more jr ones) are basically just lego builders
16:19 SEJeff       unable to make custom blocks if the ones they have don't fit
16:20 SEJeff       I see python as one of the best glue languages just like I see perl as one of the best text processing langauges
16:20 SEJeff       Which makes python exceedingly good for cobbling together things in an api
16:21 pdurbin      yeah, and i don't want to force people to write perl. it's too hard ;)
16:21 SEJeff       Ha
16:21 SEJeff       perl isn't that bad
16:22 SEJeff       just reading other people's "optimized" perl is
16:22 SEJeff       I was able to go from 0 perl to a script that did xmlrpc in < 1 week and have it work
16:22 * pdurbin    *coughs* SHUUUUFFFF #optimized
16:23 SEJeff       Yeah every perlmonger knows what I'm talking about
16:23 SEJeff       guido made a conscious choice (right or wrong doesn't matter here) to not allow that sort of thing in the language's syntax
17:33 agoddard     pdurbin: SEJeff I'm late to the party here, but I'm a big fan of Sinatra (but s/python/ruby) ;)
17:34 SEJeff       agoddard, Yeah there is def some really good ruby chops here in #crimsonfu
17:34 SEJeff       sinatra is good stuff
17:34 * SEJeff     is a huge fan of gdash for graphite metrics
17:34 SEJeff       And it is sinatra
17:35 agoddard     +1 for gdash
17:42 pdurbin      crimsonfubot: lucky gdash
17:42 crimsonfubot pdurbin: http://www.devco.net/archives/2011/10/08/gdash-graphite-dashboard.php
17:43 pdurbin      oooo, pretty
17:44 SEJeff       Yeah we configured some stock stuff and handed that to management
17:44 SEJeff       for app level metrics. They <3 it
17:51 pdurbin      SEJeff: on what level do you want to help with the python stuff? are you suggesting an open source app?
17:51 SEJeff       pdurbin, If you want to write an api in python + django
17:51 SEJeff       I've done that several times over
17:52 SEJeff       and will help you along if you have questions
17:52 SEJeff       s/api/REST ^/
17:52 SEJeff       gah
17:52 SEJeff       s/api/REST &/
17:56 pdurbin      heh
17:59 pdurbin      SEJeff: i'm sure to have questions
19:06 pdurbin      Team Based » Archive » Top Ten Project Success Factors - http://teambasedapproach.com/02/top-ten-project-success-factors/