20:01:41 <smooge> #startmeeting Infrastructure Meeting [2010-06-17]
20:01:41 <zodbot> Meeting started Thu Jun 17 20:01:41 2010 UTC.  The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:01:41 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
20:02:11 <smooge> #topic Roll Call
20:02:13 * abadger1999 is here
20:02:13 <smooge> who is here?
20:02:22 * sijis is here
20:02:23 * a-k is
20:02:27 * ianweller 
20:03:11 * skvidal is here
20:03:20 <smooge> #topic Old Business
20:03:43 <smooge> Last week we did our updates and reboots of all the servers. [or was that 2 weeks ago?]
20:04:00 * a-k thinks 2 weeks
20:04:08 <smooge> You are right.. time flies
20:04:30 <smooge> Turbogears2 has been downgraded to TG2.0 versus 2.1
20:04:31 <sijis> you are still thinking about it, eh?
20:04:54 <smooge> and a bunch of other items to make fedoracommunity sort of work
20:05:11 <smooge> Any other old business from last meeting or so?
20:05:59 <smooge> #topic Meeting Tickets
20:06:05 <smooge> https://fedorahosted.org/fedora-infrastructure/ticket/2234
20:06:20 <smooge> ianweller, abadger1999 I think this is yours
20:06:41 <ianweller> yup.
20:06:47 <ianweller> .ticket 2234
20:06:53 <zodbot> ianweller: #2234 (request for resources: datanommer) - Fedora Infrastructure - Trac - https://fedorahosted.org/fedora-infrastructure/ticket/2234
20:07:14 <ianweller> ok so i just wanted to bring up the fact that i'm working on this, and it'll be Yet Another TG2 app on infrastructure, as well as another database engine.
20:07:24 <ianweller> (it can be regular TG if you all want.)
20:07:45 <smooge> I would prefer TG2 so we could get the rest to it.
20:07:58 <smooge> Not that it would seem TG2 (2.0 -> 2.1) is seemless
20:08:08 <ianweller> ok.
20:08:14 <ianweller> it seemed like abadger1999 and others wanted TG2 too
20:08:15 <ianweller> which is fine.
20:08:26 <ianweller> the database backend is most likely going to be mongoDB
20:08:38 <ianweller> which, for those who don't know, is a document-based, no-schema backend
20:08:52 <ianweller> this has some interesting situations for us:
20:08:59 <ianweller> 1) mongoDB is actually licensed AGPLv3
20:09:07 <smooge> oh god
20:09:14 <ianweller> it scared me at first too
20:09:19 <ianweller> but.
20:09:35 * skvidal waits for prince_humperdink to get cranky
20:09:37 <smooge> s/god/${DIETY_OF_CHOICE||NONE}
20:09:54 <ianweller> a) the python driver for the database doesn't build against anything in mongoDB. so it's actually licensed apache
20:10:20 <ianweller> b) since it's all python, we *should* be able to write the TG2 app under the GPLv2 and still be fine
20:10:27 <ianweller> (an email has already been sent to spot asking about this all for sure)
20:10:33 <ianweller> c) i don't see us hotfixing a database any time soon
20:11:32 <smooge> ok I really really really HATE to ask this.. but why mongodb over say berkely or something nosql that isnt agpl?
20:12:13 <ianweller> i haven't looked specifically at berkeley
20:13:04 <ianweller> http://www.mongodb.org/display/DOCS/Licensing <-- "If you are using a vanilla MongoDB server from either source or binary packages you have NO obligations. You can ignore the rest of this page."
20:13:17 <abadger1999> I think berkeley would be a mistake.  But there are some new key-value databases that could take berkeley's place.
20:13:26 * skvidal +1's abadger1999
20:13:32 <abadger1999> (tokyo cabinet w/tokyo tyrant)
20:13:36 <skvidal> let us NEVER speak of bezerkley again
20:13:41 <ianweller> heh
20:13:52 <smooge> abadger1999,  I pulled the name out my arse.
20:14:03 <abadger1999> I haven't looked into that specfically; I've heard that mongo is faster and "better".
20:14:29 <smooge> ok I just really hate dealing with license madness.
20:14:44 <ianweller> mongo is very very very fast.
20:14:44 <smooge> apl vs gpl vs agpl vs lmnop
20:14:52 <abadger1999> The other high profile no-sql dbs (couchdb, cassandra, reik, etc) all seem to be either erlang or java.... not sure we want to suck infrastructure down those roads.
20:15:09 * ianweller doesn't want to be That Guy bringing java or erlang in...
20:15:48 <smooge> ok next are we talking about a single mongo or mongo's on every app/proxy/george/ringo/paul server
20:16:14 <ianweller> i am just talking about a single mongo
20:17:04 <smooge> np I just wanted to get an idea about where and how.. you guys can work on locking/concurrency/etc
20:17:11 <ianweller> yeah
20:17:23 * rbergeron has a bf writing a nosql but unfortunately it's not ready today :)
20:17:33 <ianweller> hah.
20:17:48 <rbergeron> problematic, i suppose... minor detail
20:18:10 <smooge> minor
20:18:22 <rbergeron> minor.
20:18:27 <smooge> ianweller, abadger1999 ok so what resources are you looking at and when
20:18:50 * ianweller will need abadger1999 to help fill in the holes here pretty quick :)
20:19:28 <ianweller> so my *hope* is that i have the TG2 app done by the end of the month
20:19:36 <ianweller> and to have it either on a publictest server, or on staging.
20:19:43 <abadger1999> I haven't looked at it -- we'll need a mongodb server -- we could treat it like our other db servers (db01,2,3)
20:20:01 <abadger1999> Not sure of the resources -- whether we'll Memory/CPU/or IO bound.
20:20:09 <ianweller> it'll be CPU
20:20:13 <ianweller> based on my testing here...
20:20:28 <abadger1999> <nod>
20:20:39 <abadger1999> The web application can live on the app servers.
20:20:45 <ianweller> yeah.
20:20:50 <ianweller> and that shouldn't take up too much resources
20:21:05 <ianweller> since all of the querying/sorting is going to be on the db side
20:21:33 <abadger1999> The piece that pulls in logs and pushes them into the mongodb will probably want its own machine.  skvidal thought maybe dual purposing log01 for that could work... I'm not sure if we want that or its own machine.
20:21:53 <skvidal> nod
20:22:01 <skvidal> I didn't actually say we had to dual purpose log01
20:22:06 <smooge> ianweller, one thing to remember is that CPU/Memory/IO bounding usually changes drastically when you start dealing with many on one issues.
20:22:06 <skvidal> just that it is doing similar things
20:22:16 <abadger1999> I think that decision is  more about security considerations than resources.
20:22:18 <abadger1999> <nod>
20:22:53 <ianweller> well, we'll be using a script to push web log data into the database.
20:23:03 <ianweller> so we may as well use log01, i guess
20:23:11 <smooge> it will need to be a seperate machine I think (long term). The collectd items can get log01 at times
20:23:54 <ianweller> <nod>
20:24:08 <ianweller> and, the log files can be removed after they're put in the database
20:24:28 <smooge> plus for security reasons you usually want a 'clean' system to deal with data so that things like "F*ckFEdora13" tried to log in 300 times with password ianweller doesn't show up
20:24:49 <ianweller> hey how'd you guess my password
20:24:50 <ianweller> ;)
20:25:12 <smooge> keylogger much faster
20:25:14 <jokajak> woops, sorry i'm late
20:25:38 <ianweller> smooge: what do you mean by 'clean' system, anywho
20:26:23 <a-k> No extra packages or config entropy ... ?
20:26:59 <ianweller> just that? i was wondering if he was extending 'clean' to "let's not put data previous to the setup date in datanommer"
20:28:15 <ianweller> so other than bringing the details and all to your attention...
20:29:28 <smooge> clean without secure log files and such that would have data you "no WANT!"
20:29:38 <smooge> jokajak, hi hope you are doing well
20:29:53 <ianweller> ok.
20:30:41 <ianweller> abadger1999: unless you have anything else...
20:30:54 <abadger1999> Nothing here
20:31:02 * ianweller is done
20:31:07 <smooge> ok any other items for today?
20:31:16 <smooge> #topic Open Floor
20:31:46 <smooge> I am dealing with trying to get fedora community working again. I thought I had it but it seems I missed something yesterday that will need to be redone
20:32:14 <smooge> after that my work is mostly going to be getting zikula onto staging so we can get Fedora Insight out the door.
20:32:18 <smooge> others?
20:33:36 <abadger1999> I've had to work on a lot of things not directly related to infra lately -- at least two weeks before htat changes.
20:36:03 <smooge> ah. skvidal anything for open floor?
20:36:10 <skvidal> not that I recall
20:37:50 <smooge> anyone else? If not I will close in 1
20:38:19 <smooge> #closemeeting
20:38:34 <smooge> #endmeeting