uPortal IRC Logs-2007-08-28

[10:10:18 EDT(-0400)] * EricDalquist (n=dalquist@bohemia.doit.wisc.edu) has joined ##uportal
[10:47:03 EDT(-0400)] <EricDalquist> yay for very rare race conditions
[11:04:15 EDT(-0400)] <pberry> negative. boo for those
[11:05:03 EDT(-0400)] <apetro_work_desk> race conditions are fun
[11:05:16 EDT(-0400)] <apetro_work_desk> speaking of which, these layout cache keys
[11:05:25 EDT(-0400)] <pberry> is that like what the weather is like when you go do a half-marathon?
[11:05:45 EDT(-0400)] <apetro_work_desk> my careful synchronization of my new, simpler, non-GUID serial number generator, it's cute and all.
[11:05:59 EDT(-0400)] <apetro_work_desk> but since state changes on the layout manager itself aren't threadsafe, it's all for naught, right?
[11:09:26 EDT(-0400)] <apetro_work_desk> in other news, I wrote some code to detect repeated node ids in ALM layouts, the intent was to continue to go after infinite loops.
[11:10:42 EDT(-0400)] <apetro_work_desk> I always get a hit on node ID 1, but this isn't apparently actually a problem. So I've coded around it – exceptionally, 1 is allowed to multiply appear – but I feel pretty stupid coding in the exception. Anyone know why it would be okay for a node ID to multiply appear in the recursive appendBlah() ALM calls?
[11:11:55 EDT(-0400)] <apetro_work_desk> Despite my doubts, the 242 maint branch is looking like a success. In testing I can't get it to break fatally.
[11:29:08 EDT(-0400)] * apetro_work_desk (n=apetro@uni1.unicon.net) has joined ##uportal
[11:29:16 EDT(-0400)] * Topic is 'race conditions for fun and for profit' set by apetro_work_desk on 2007-08-28 11:29:16 EDT(-0400)
[11:30:20 EDT(-0400)] <pberry> we had an interesting day yesterday http://tinyurl.com/2ultza
[11:30:35 EDT(-0400)] <EricDalquist> very nice apetro_work_desk
[11:31:27 EDT(-0400)] <EricDalquist> nice graph
[11:31:49 EDT(-0400)] <pberry> thanks
[11:31:53 EDT(-0400)] <apetro_work_desk> looks musical
[11:32:02 EDT(-0400)] <pberry> I wish it didn't have that big blank spot in it
[11:32:10 EDT(-0400)] <apetro_work_desk>
[11:32:56 EDT(-0400)] <pberry> up2.6, Java 5, Tomcat 5.5, two Dell 2850s - handled 1500+ concurrent users like it was nothing
[11:33:21 EDT(-0400)] <pberry> and we had 1,500+ for most of the day
[11:33:28 EDT(-0400)] <pberry> er, most of the school day
[11:33:55 EDT(-0400)] <pberry> it's 8:30am and we already have 800+
[11:34:33 EDT(-0400)] <apetro_work_desk> mmm
[11:35:08 EDT(-0400)] <apetro_work_desk> not to rain on the parade, but you're sure it's not artificially generating sessions through silly redirects and session management, and/or retaining sessions it shouldn't?
[11:35:16 EDT(-0400)] <pberry> yes
[11:35:16 EDT(-0400)] <apetro_work_desk> that's legitimate usage you're talking about?
[11:35:34 EDT(-0400)] <pberry> we track guest sessions and non-guest sessions
[11:35:50 EDT(-0400)] <pberry> but first, a coffee run...
[11:36:01 EDT(-0400)] <apetro_work_desk> that's pretty impressive. 750 concurrent per box, nice.
[11:36:24 EDT(-0400)] <apetro_work_desk> I sense an enhanced case study coming on
[11:49:33 EDT(-0400)] <pberry> yeah
[11:49:41 EDT(-0400)] <pberry> it definitely needs some editing
[11:49:53 EDT(-0400)] <pberry> I really need a CAS cluster
[11:50:06 EDT(-0400)] <pberry> the wiki entry makes it sound pretty easy
[11:53:33 EDT(-0400)] <apetro_work_desk> do you really need a CAS cluster?
[11:53:42 EDT(-0400)] <apetro_work_desk> Yale still doesn't cluster their CAS
[11:53:44 EDT(-0400)] <apetro_work_desk> tjey
[11:54:08 EDT(-0400)] <apetro_work_desk> they've got a warm spare, but no clustering or hot failover automation
[11:54:17 EDT(-0400)] <pberry> we need to do one of two things: off-load SSL from tomcat or cluster (or both of course)
[11:54:25 EDT(-0400)] <apetro_work_desk> every once in a very long while they take an outage for a couple minutes at some unholy early morning hour
[11:54:37 EDT(-0400)] <apetro_work_desk> oh, offloading SSL is key
[11:54:47 EDT(-0400)] <pberry> yeah, we don't have that right now...
[11:54:53 EDT(-0400)] <apetro_work_desk> hardware accelerator
[11:55:02 EDT(-0400)] <apetro_work_desk> definitely an excellent Yale CAS investment
[11:55:07 EDT(-0400)] <pberry> even Apache would be huge improvement
[11:55:11 EDT(-0400)] <apetro_work_desk> sure
[11:55:37 EDT(-0400)] <pberry> Tomcat 5.0.28 with JDK 1.4 can deal with about 11 connections/second doing the SSL itself
[11:56:08 EDT(-0400)] <apetro_work_desk> I dunno. Clustering's not that bad. And hey, certainly don't want to talk you out of hiring your friendly neighborhood vendor to help with a clustered CAS 3.1 install. winks. But
[11:56:15 EDT(-0400)] <apetro_work_desk> I'm still hung up on the simplicity thing
[11:56:33 EDT(-0400)] <apetro_work_desk> there's robustness to be had in making things as stupidly simple as possible
[11:56:51 EDT(-0400)] <apetro_work_desk> add a clustering technology, it's one more thing to misconfigure or break.
[11:57:07 EDT(-0400)] <EricDalquist> until your chillers die and you have to shut down a data center
[11:57:18 EDT(-0400)] <apetro_work_desk> indeed
[11:57:29 EDT(-0400)] <apetro_work_desk> but how much worse off are you with a warm spare in the other data center?
[11:57:38 EDT(-0400)] <EricDalquist> we're (UW as a whole) slowly splitting as many services as possible between two datacenters
[11:57:53 EDT(-0400)] <EricDalquist> and if we had CAS clustering would be nessedary
[11:57:55 EDT(-0400)] <apetro_work_desk> an excellent idea. I'm with you 100% on that one.
[11:58:03 EDT(-0400)] <EricDalquist> or all active SSO users would have to re=login
[11:58:08 EDT(-0400)] <pberry> here's the thing, clustering CAS isn't rocket science anymore. I think there is tremendous value in getting that last 9
[11:58:26 EDT(-0400)] <pberry> if only for the end of year service level numbers
[11:58:33 EDT(-0400)] <EricDalquist>
[11:58:34 EDT(-0400)] <apetro_work_desk>
[11:58:42 EDT(-0400)] <pberry> or not having to create a meeting of 20 people to figure out when we can upgrade the kernel
[11:58:49 EDT(-0400)] <apetro_work_desk> yeah
[11:58:58 EDT(-0400)] <apetro_work_desk> organizational considerations may drive that clustering technology
[11:59:09 EDT(-0400)] <apetro_work_desk> I'm not yet convinced the technology considerations drive it.
[11:59:15 EDT(-0400)] <pberry> yeah, it will still happen at 0-dark early, but we don't have to deal with the notifications.
[11:59:24 EDT(-0400)] <apetro_work_desk> but you're right, it's not rocket science anymore, so it's less risky to add the technology
[11:59:26 EDT(-0400)] <pberry> we just take down one part of the cluster, update it, put it back in
[11:59:41 EDT(-0400)] <pberry> lather, rinse, repeat
[12:00:00 EDT(-0400)] <apetro_work_desk> there's some appeal in that
[12:00:42 EDT(-0400)] <pberry> and once we do the work to off-load SSL, we're practically done with clustering anyway
[12:00:56 EDT(-0400)] <apetro_work_desk> more fun if you take the next step and have Amazon dynamically virtually host it, spinning up additional server images into the cluster as load requires, and spinning them down again when load diminishes. Dynamically scaled application clusters via virtualization. It's gonna be big.
[12:01:07 EDT(-0400)] <pberry> we'll probably also stick at least one part of the cluster on a VMWare instance
[12:01:13 EDT(-0400)] <apetro_work_desk> t'da!
[12:01:28 EDT(-0400)] <pberry> no that is something I wait for other people to test
[12:01:34 EDT(-0400)] <pberry> now
[12:23:38 EDT(-0400)] <pberry> apetro_work_desk: you mentioned we might be counting bogus sessions...
[12:24:16 EDT(-0400)] <pberry> we're using some code from Adam R. with 'NamingEnumeration ne = context.listBindings("sessions");'
[12:25:39 EDT(-0400)] <apetro_work_desk> interesting
[12:26:11 EDT(-0400)] <apetro_work_desk> I'd like to see CRuntimeData, RDBMStats recorder, and this associated stuff drawn into the core uPortal distro
[12:26:48 EDT(-0400)] <apetro_work_desk> I'd also like to see someone, possibly me, find the time to make the CUserManager UI less appalling since that has been drawn into the core distro
[12:28:13 EDT(-0400)] <pberry> Scott has done a lot of work on cleaning up DLM Preferences. I'll dig around and see who is actively in that area...or maybe I'll just fire the xsl off to the dev list
[12:30:20 EDT(-0400)] <pberry> uh-oh. wiki is having problem with uploading attachments
[12:30:37 EDT(-0400)] <EricDalquist> ?
[12:30:38 EDT(-0400)] * pberry tries firefox
[12:30:53 EDT(-0400)] <pberry> I'm updating the Chico portal profile
[12:30:56 EDT(-0400)] <pberry> updating tab images
[12:31:01 EDT(-0400)] <pberry> uploads just kinda hang
[12:31:09 EDT(-0400)] <EricDalquist> weird
[12:31:14 EDT(-0400)] <pberry> yeah, could be safari
[12:31:34 EDT(-0400)] <pberry> using the 3beta
[12:32:04 EDT(-0400)] <pberry> and there it went
[12:32:09 EDT(-0400)] <pberry> no problem with FF.
[12:32:14 EDT(-0400)] <pberry> Seems like a safari bug
[12:32:17 EDT(-0400)] <pberry> panic attack over
[12:50:34 EDT(-0400)] <pberry> http://www.ja-sig.org/wiki/x/lz0 updated
[12:50:44 EDT(-0400)] <pberry> Are there other things you guys think should be included?
[13:29:54 EDT(-0400)] <pberry> not a single macports mirror has maven2
[13:29:58 EDT(-0400)] <pberry> what the frick?
[13:30:32 EDT(-0400)] <apetro_work_desk> Ubuntu will run on that mac hardware, you know.
[13:30:45 EDT(-0400)] <apetro_work_desk> I like the profile / case study writeup.
[13:30:53 EDT(-0400)] <pberry> yes, but I like to actually get work done
[13:30:55 EDT(-0400)] <pberry>
[13:31:21 EDT(-0400)] <apetro_work_desk> needs more hyperlinks to unicon.net, but other than that, it's a good writeup
[13:31:55 EDT(-0400)] <pberry> Linux is my server platform of choice. I just can't handle the "desktop" apps
[13:38:30 EDT(-0400)] * pberry plays with the stable CAS 3.1 release
[13:40:54 EDT(-0400)] <pberry> note to self, update macports, then try to install software
[13:42:48 EDT(-0400)] <lescour> pberry, didn't you tell me you are on cas 1?
[13:44:35 EDT(-0400)] <pberry> cas 2, using protocol v1
[13:46:07 EDT(-0400)] <pberry> pop quiz, how do you make maven2 give you a list of valid "goals"?
[13:48:57 EDT(-0400)] <EricDalquist> you don't
[13:49:11 EDT(-0400)] <EricDalquist> well I guess there is a way
[13:49:18 EDT(-0400)] <EricDalquist> but generally you just use the standard set
[13:49:27 EDT(-0400)] <EricDalquist> and hope the project documented any special cases for you
[13:49:48 EDT(-0400)] <EricDalquist> the standard set is: compile test package clean install
[13:49:56 EDT(-0400)] <EricDalquist> at least that is all I really end up using
[13:52:45 EDT(-0400)] <pberry> gotcha
[13:52:55 EDT(-0400)] <pberry> I was hoping there would be something in the README
[13:53:03 EDT(-0400)] <pberry> like a section for "What is this pom.xml for?"
[13:53:08 EDT(-0400)] <EricDalquist>
[13:53:24 EDT(-0400)] <EricDalquist> there is on the maven site
[13:55:49 EDT(-0400)] <pberry> that is accurate, and yet, not useful
[13:56:06 EDT(-0400)] <EricDalquist> sounds about right
[14:15:50 EDT(-0400)] <pberry> maven, not for the faint of heart...
[14:16:07 EDT(-0400)] <EricDalquist> yeah
[14:16:12 EDT(-0400)] <EricDalquist> it really has a learning curve to it
[14:16:21 EDT(-0400)] <EricDalquist> now that I've got it I don't ever want to go back though
[14:33:16 EDT(-0400)] <pberry> sure, and I totally get that
[14:33:53 EDT(-0400)] <pberry> right now I'm dealing with customizing CAS. So of course you need to configure and rebuild stuff
[14:34:45 EDT(-0400)] <EricDalquist>
[14:38:25 EDT(-0400)] <EricDalquist> hey apetro_work_desk ... do the scripts in etc work for cutting a release?
[15:07:26 EDT(-0400)] <apetro_work_desk> Probably not
[15:07:34 EDT(-0400)] <apetro_work_desk> Did I commit my updated SVN-based one?
[15:08:10 EDT(-0400)] <apetro_work_desk> pberry , last I looked the CAS build model was that
[15:08:20 EDT(-0400)] <apetro_work_desk> Maven was the platform for CAS developers developing CAS
[15:08:25 EDT(-0400)] <apetro_work_desk> the release deliverable, however
[15:08:34 EDT(-0400)] <apetro_work_desk> is customizable and "buildable" using a much simpler Ant build
[15:08:57 EDT(-0400)] <apetro_work_desk> so your workflow is: edit config files, drop in any optional module .jars you want and their configuration, hit "go" on the ant build, out comes a shiny .war
[15:09:12 EDT(-0400)] <apetro_work_desk> Maven wasn't required by deployers of CAS last I looked
[15:09:13 EDT(-0400)] <apetro_work_desk> may have changed
[15:12:00 EDT(-0400)] <apetro_work_desk> EricDalquist , I emailed you the .sh I most recently used. I don't know offhand how it relates to what's in etc.
[15:15:29 EDT(-0400)] <pberry> apetro_work_desk: nothing but pom.xml in the 3.1-stable download
[15:15:40 EDT(-0400)] <pberry> I don't mind needing to use maven2
[15:15:52 EDT(-0400)] <pberry> I just need very. explicit. instructions.
[15:16:04 EDT(-0400)] <apetro_work_desk> indeed.
[15:16:34 EDT(-0400)] <apetro_work_desk> I don't know enough about the maven2 build to have an opinion about whether an ant local deployment alternative is worth having
[15:16:57 EDT(-0400)] <apetro_work_desk> but I'm a little surprised the ant build approach isn't there, since there was a good deal of discussion and work around that feature in CAS 3.0
[15:19:28 EDT(-0400)] <lescour> well, i think the first thing would be to sync the docs
[15:19:40 EDT(-0400)] <lescour> the cas website install lists apache ant as neccessary
[15:20:00 EDT(-0400)] <lescour> and the wiki indicates you deploy with mvn
[15:25:46 EDT(-0400)] <apetro_work_desk> yeah
[15:26:05 EDT(-0400)] <apetro_work_desk> I'm pretty convinced that the thing to do is to rescue the documentation out of the website and get it into the wiki where it's easier to edit
[15:26:13 EDT(-0400)] <apetro_work_desk> and comment upon.
[15:26:25 EDT(-0400)] <apetro_work_desk> I get the idea everyone else is convinced too and it's just a matter of doing it
[15:26:49 EDT(-0400)] <apetro_work_desk> leave the website for the unspecific marketing stuff that makes CAS look beautiful on the web as a project but doesn't need edited much
[15:28:01 EDT(-0400)] <pberry> w00t, got the conference eval scores back
[15:33:07 EDT(-0400)] <apetro_work_desk> neat
[15:34:07 EDT(-0400)] * lescour sees the email preview portlet has java5 syntax
[15:37:00 EDT(-0400)] <pberry> for my first solo talk, I think the scores were pretty good (even though I know people go way easy on those)
[16:17:21 EDT(-0400)] <apetro_work_desk> it's a very positive, understanding, supportive audience
[16:17:26 EDT(-0400)] <apetro_work_desk> makes it a relatively fun conference to speak at
[22:28:38 EDT(-0400)] * jayshao (n=jayshao@pool-72-79-113-152.nwrknj.east.verizon.net) has joined ##uportal
[23:40:42 EDT(-0400)] * jayshao (n=jayshao@pool-72-79-113-152.nwrknj.east.verizon.net) has joined ##uportal