MEETING NOTES: jUDDI status meeting tomorrow Friday 14

[Kurt T Stam <ku...@gmail.com>]

KurtStam: k let's do a quick go around then; I think Jeremi last since I 
think we'll talk about the testing the longest
[10:02am] KurtStam: who wants to go first?
[10:03am] tcunning: i can go first
[10:03am] tcunning: right now i'm working on getting scout working with 
juddi v3
[10:03am] tcunning: just finished off removing getContent
[10:04am] tcunning: and i'll probably be working on the dist bundle next 
week
[10:04am] tcunning: that's about it
[10:05am] tcunning: np
[10:06am] KurtStam: Jeff?
[10:06am] KurtStam: Well I can go while Jeff is getting his coffee
[10:06am] jfaath: working on the JIRAs assigned to me
[10:06am] KurtStam: ah
[10:07am] jfaath: which isn't much right now
[10:07am] KurtStam: k
[10:07am] KurtStam: I've been spending some time with scout and jUDDIv2 
since that still ships with the AppServers (JBoss, Gerominimo etc) 
because of the TCK
[10:08am] KurtStam: so that side tracked me from jUDDIv3 work a bit.
[10:08am] KurtStam: but that is done now; so I've been working and on v3 
the last week or so.
[10:09am] KurtStam: It think things are looking pretty good. I have 
rearranged Jira a bit; I think we need to release 3.0 sooner rather then 
later
[10:10am] KurtStam: I'm happy with the results from the load tests so 
far. I think some of the issues found will need to address for 3.0 and 
some can wait till 3.0.1
[10:10am] KurtStam: Rene want to jump in to introduce the load testing 
stuff and then jeremi can take us through some details?
[10:11am] Natalie__ joined the chat room.
[10:11am] ReneXceptance: Yes.
[10:11am] KurtStam: Hi Natalie
[10:11am] Natalie__: Hi there.  Sorry to jump in late.
[10:11am] tcunning: Natalie - we're doing a go around of what everyone 
is working on, i can send you a log of what's gone on so far
[10:11am] ReneXceptance: So basically we are using our tool, that is 
capable of executing any kind of http/html reuqests.
[10:12am] ReneXceptance: It was extended to support web services and you 
program it directly on the Java-API.
[10:12am] Natalie__: Ok great, thanks
[10:12am] ReneXceptance: Therefore the tests are as close as possible to 
the later usage scenarios. No tricks with recorded urls and so on.
[10:13am] ReneXceptance: The juddi project has a full license and the 
scripts are yours... just have to find a nice place to store them. They 
are pure Java
[10:13am] ReneXceptance: and you can, at the end, integrate the stuff as 
regression test into the build.
[10:14am] ReneXceptance: If you have a way to get the server up during 
the build, so you can run performance regression.
[10:14am] KurtStam: that's great; we definetly need to hook them up like 
that
[10:14am] KurtStam: yup we do that now in the uddi-client module actually
[10:14am] ReneXceptance: And I guess Jeremi can kick in with some stuff 
we did
[10:14am] KurtStam: thx Rene
[10:15am] jeremi: So I first did some juddi-out-of the -box testing
[10:16am] jeremi: and got some benchmark values that can be seen in the 
last update I sent
[10:16am] jeremi: found here, 
file:///home/jeremi/.gvfs/ftp%20as%20ftp1058375-jthebeau%20on%20www.xceptance.de/download/results/juddi/JuddiReports/RE:%20jUDDI%20load%20testing%20update:%20performance%20comparaison.html
[10:17am] KurtStam: maybe you have an http link?
[10:17am] ReneXceptance: maybe not the local file copy...
[10:17am] jeremi: oops
[10:18am] jeremi: 
http://xlt.xceptance.de/download/results/juddi/JuddiReports/RE: jUDDI 
load testing update: performance comparaison.html
[10:19am] KurtStam: I think the link got cut off..
[10:19am] jeremi: then I compared that with juddi using MySQL as a DB 
instead of Derby
[10:19am] KurtStam: so this is both using hibernate right?
[10:20am] KurtStam: with OpenJPA we see a memory leak?
[10:20am] jeremi: yes
[10:20am] KurtStam: 
http://xlt.xceptance.de/download/results/juddi/20090812-124530-vs-20090812-134739/
[10:20am] KurtStam: is the comparison of dbs?
[10:20am] jeremi: this was with the Snapshot that Kurt sent me
[10:20am] KurtStam: cool; yeah that was a recent snapshot buil;d
[10:21am] ReneXceptance: if i remember correctly the first build had 
memory issues
[10:21am] KurtStam: the later builds not anymore?
[10:21am] jeremi: the comparison is of juddi with these two different DBs
[10:22am] ReneXceptance: Baseline Derby compared to MySQL
[10:22am] ReneXceptance: green is good. check the requests section
[10:22am] KurtStam: so mysql is quite a but faster
[10:22am] KurtStam: bit
[10:23am] jeremi: after that we ran some performance test to compare 
them under a load setting they could both take
[10:24am] jeremi: In which juddi + MySQL responded  faster on average
[10:25am] KurtStam: Cool; yeah the snapshots run hibernate; so we prob 
still have an issue using openjpa.
[10:25am] ReneXceptance: shall we run mysql and openJPA?
[10:25am] jeremi: Finding a load profile that both could take was a bit 
difficult since juddi + Derby kept freezing up a loads that juddi+ MySQL 
could easily take
[10:25am] KurtStam: when you crack up the load what is the failure mode?
[10:26am] KurtStam: right; that's what I was looking for.. yeah Derby is 
not the way to go in production..
[10:26am] jeremi: the system becomes unresponsive,
[10:26am] KurtStam: how does mysql start to fail?
[10:26am] tcunning: production would probably be 
oracle/postgres/mysql/sybase etc
[10:27am] KurtStam: does it lock; or just slow down..
[10:27am] tcunning: is the mysql database on the same machine as the 
juddi server?
[10:27am] jeremi: the same way but only at a much higher load or after 
operation lor longer
[10:27am] jeremi: yes MySQL is on the same machine as juddi
[10:28am] ReneXceptance: how is the load of the machine during test. Is 
the CPU tanked?
[10:28am] jeremi: The cpu is not tanked
[10:29am] jeremi: jstack shows that all threads are waiting on the data 
source
[10:29am] KurtStam: it sounds like we bound by the speed of the txns? 
You re contiunously inserting during the test right?
[10:29am] KurtStam: I mean not just simple reads...
[10:30am] jeremi: the system becomes locked up. If I stop the test and 
try to send another request It does not respond
[10:30am] KurtStam: I think in a real life scenario the load will be 
mostly reads/finders
[10:30am] jeremi: I have to shut it down and start it up again
[10:31am] jeremi: I am continuously registering businesses and services 
during this test
[10:33am] KurtStam: right.. so it should be able to handle it, but I 
think we should try to make a scenario which switches to more reads 
after a little while.
[10:33am] jeremi: and for a just read scenario (find_business) MySQL can 
take about 2.5 times more load
[10:34am] KurtStam: I'd like to fix the load issues in 3.0.1, and want 
to make sure things are stable enough for 3.0.0. We should surely fix 
the memory issues seen using openjpa for 3.0.0.
[10:34am] KurtStam: What would be agood place to check in the scripts so 
we can reproduce the issues on a dev box?
[10:35am] KurtStam: make a juddi-funcational test module?
[10:36am] KurtStam: jeremi can you open a jira and attach the latest 
scripts?
[10:36am] jeremi: sure
[10:37am] ReneXceptance: please add a little description how to use it 
to the file, maybe as readme. i know you already have some stuff ready 
to share.
[10:37am] KurtStam: The way apache works is that we start a vote for 
committer rights when we can show your contributions.. so hope at some 
time soon you can maintain them yourself?
[10:38am] KurtStam: I think the easiest way to share the test results 
maybe to check them into SVN?
[10:38am] jeremi: ok
[10:39am] KurtStam: cool
[10:39am] ReneXceptance: the results are about 2-3 Mb per archive or 
plain checkin without an archive... do you want to pollute SVN with all 
the images and unchanging data?
[10:40am] jeremi: Someone at xceptance will probably be taking over, 
since I will be going on paternity leave around the 21
[10:40am] KurtStam: k
[10:40am] KurtStam: not sure about where to host the results..
[10:41am] ReneXceptance: yes, jeremi will share all stuff and Hardy or I 
will take over
[10:41am] KurtStam: I think it is good to have a history of these tests
[10:41am] ReneXceptance: we can also ask Patrick to join the testing
[10:42am] KurtStam: I think I like SVN then we can tag the scripts with 
each result report
[10:42am] jeremi: During testing I also found that I could not switch 
data sources after the first time a fresh install was started
[10:43am] jeremi: Has anyone ever had this problem?
[10:43am] KurtStam: right; I think tomcat is caching somehting somewhere
[10:43am] jeremi: ok
[10:43am] KurtStam: I don't seem to have it.
[10:44am] KurtStam: but we should prob document it
[10:44am] KurtStam: anyhting else before we go over jira; to see what's 
needed to finish up 3.0.0? Thx so much for you testing efforts !
[10:45am] ReneXceptance: your welcome. just let us know, if you need 
help or a certain question answered
[10:45am] KurtStam: thx
[10:46am] KurtStam: I will be setting it up on my machine; so I will 
prob take you up on that
[10:46am] ReneXceptance: sure
[10:46am] KurtStam: So we currently have 18 jiras open for 3.0
[10:46am] KurtStam: half of them around the console
[10:47am] KurtStam: we need to add another one for openjpa leak.
[10:47am] KurtStam: but I think over all it's pretty manageable
[10:48am] jfaath: 267 is the openjpa leak
[10:48am] KurtStam: Ah right thx
[10:48am] KurtStam: We have a JProfiler license now; so with hooking up 
xlt I will take a look at that
[10:49am] tcunning: hey does anyone know why we are spewing the xsd's in 
tomcat after a request?
[10:49am] KurtStam: do you agree that limitations of jUDDI under load 
can wait till 3.0.1?
[10:50am] KurtStam: we are?
[10:50am] tcunning: yah
[10:50am] KurtStam: in the logs?
[10:50am] tcunning: console too
[10:51am] tcunning: i'll log something on it
[10:51am] KurtStam: k
[10:51am] KurtStam: is everyone happy with the jira list for 3.0?
[10:52am] KurtStam: 89 can prob go to 3.0.1
[10:52am] KurtStam: this is the v2 version of the inclauses issues we 
still have in v3 found by the XLT tests
[10:53am] KurtStam: but I think it's good enough for 3.0.0?
[10:54am] KurtStam: ok ..
[10:54am] KurtStam: I take that as a yes then
[10:55am] KurtStam: let's see if we can get to 3.0.0 soon then.
[10:55am] KurtStam: anyhting else anyone wants to bring up. I know Tom 
has been working on Scout to make it speak UDDIv3 which would open the 
way for jUDDIv3 to be included with AppServers..
[10:56am] Natalie__: I don't really have much to bring to the plate, but 
I'm trying to deploy juddi v3 to JBoss EAP 5 beta
[10:57am] KurtStam: cool ; you prob have to uninstall juddi v2 for that 
to work right?
[10:57am] Natalie__: Or at least will be actively next week, so it's 
probably not what you guys are looking for, but I can send in any notes 
or issues as they arise.
[10:57am] Natalie__: Yes
[10:58am] KurtStam: yeah that'd be great. we started a juddi.war module 
which should be what you deploy to an appserver but it still has 
hibernate in it I believe..
[10:58am] KurtStam: so what ever you find will end up in that module.
[10:58am] Natalie__: Sounds good.
[10:58am] KurtStam: are you planning on deploying the console also?
[10:59am] Natalie__: Yes - as soon as I get that working (and don't get 
side-tracked with other stuff).
[10:59am] KurtStam: great
[10:59am] KurtStam: ok well I think if there is nothing else.. our hours 
is up.
[10:59am] KurtStam: thx guys!
[11:00am] Natalie__: Thanks for letting me sit in - trying to learn a 
lot more about this.
[11:00am] KurtStam: anytime
[11:00am] jeremi: same here, thx
[11:01am] Natalie__: Have a nice weekend.  Bye, everyone.
[11:01am] Natalie__ left the chat room. ("Ex-Chat")
[11:02am] ReneXceptance: bye
[11:02am] ReneXceptance: and take care
[11:02am] KurtStam: cheers :)!
[11:02am] jeremi: bye


Kurt T Stam wrote:
> Hi guys,
>
> Let's meet up on IRC tomorrow 9am Chicago time/10 am Boston time / 4 
> pm Germany time (Jeremy does that work for you?).
>
> irc.freenode.net #juddi.
>
> Suggested Agenda:
>
> * Quick go around to see what everyone has been up to.
> * Check Jira to see what should go in the 3.0.0 release.
> * Introduction to load testing work by Jeremy
>    - Link to test results, or publish on the apache site?
>    - Check test scripts into the source tree?
> * Anything else that comes to mind.
>
> Hope you can make it!
>
> --Kurt
>
>