Hi all, I'm trying to get acquainted with XBean so I can setup MINA with an XSD. Don't want to bother y'all too much here but wondering where it's best to look to find my answers. I don't see much documentation on the site so I was wondering if there's something out there that I'm not aware of. Thanks!

For example one of my questions is:

If I have an inheritance hierarchy say Foo implementing Bar and Bar interface has a destroy() method. How do I mark up the Javadocs so Foo.destroy() is called as XBean DestroyMethod?

akarasulu_: afraid that you won't find a lot of xbean doc. may be that dain could help you

ahh okie thanks kevan

maybe

:)

hiya dain

there isn't much doc beyoned what's on the site

FYI guys what documentation is online seems to be having some issue it seems: "An error occurred: Connection refused. The system administrator has been notified. "

just saw that

hmm. what link is that?

let me take a look

http://geronimo.apache.org/xbean/custom-xml.html

for example

hmm. i got it just fine

kevan, if you search for "error occurred" you'll see where there are some examples that should show no longer do.

they're replaced with that error message above

ah. i see

Is XBean development active?

Wondering if you guys switched to something else.

akarasulu_: it works in cwiki http://cwiki.apache.org/confluence/display/XB/Custom+XML

ahh okie thanks

dain: What do you recommend for a new comer trying to learn how to use XBean beyond these simple cases in the doco? Trying to figure out the best way to learn the ins and outs.

kevan: any idea how to properly format a snippit

heh. i was waiting for you to show me how... ;-)

I suggest the simple examples + hacking a jnuit test case

assuming you know spring the transalation is pretty simple

seems XBean has many more features than is shown in the simple examples

Like for example where do I learn about how inheritance is handled?

what do you mean about inheritance?

If I have some inheritance hierarchy say Foo implementing Bar and Bar interface has a destroy() method. How do I mark up the Javadocs so Foo.destroy() is called as XBean DestroyMethod?

Err, rather Bar is a class that is extended by Foo and it has the implementation for the destroy() method?

do I add XBean tags to the javadocs of Bar ?

and Foo just inherits it ?

I don't know :)

you'll have to try it

ok will do

dain, just a couple questions if you have a minute

I do... been trying to fix the site in another window :)

ahh okie

dain are you editing custom xml?

kevan: say what?

are you updating the custom xml page in the wiki

better? ;-)

Just curious, how active is the dev community around XBean?

kevan: no... I think it is fine... I'm trying a new export

akarasulu_: we fix bugs as people find them but not to much new has happened lately

okie

I need to spend some time to update the code to support an annotation processor, but I've been lazy

Well I'd be glad to help out with writing some additional documentation if someone would help me lessen this learning curve. I'm afraid if I have to just try every combination or open up the source its going to take me a long while to understand things.

kevan: a fresh export cleaned up the xml http://cwiki.apache.org/XB/custom-xml.html

akarasulu_: I'll help

thanks dain

it broke the page that i added geronimo/ prefixes onto the url snippets

there aren't that many options possible... it is a pretty simple framework

akarasulu_: checkout the source and load it into your ide

which i fixed. how did the export "fix the problem"? and how is the base url configured?

okie doing so now ...

you only need the xbean-spring package

the test directory contains tons of unit tests and examples

kevan: I think the page had been exported with an error and never reexported

the base url is configured in the snippit plugin

ahh okie I'll start looking there

xbean/ is mapped to http://svn.apache.org/repos/asf/geronimo/xbean/

kevan: do you know how the pages are synced from cwiki to the main site?

k. when i added geronim/ to the url that fixed the problem on the page i was editing

i don't really know how that works

that works because geronimo/ is mapped to http://svn.apache.org/repos/asf/geronimo/

if you are a cwiki admin, click administration|plugins|snippit|configure

it will be very obvious at that point

k

dain, got a sec for a question?

yep

I have this signature here: http://mina.apache.org/.../AbstractIoAcceptor.html#setDefaultLocalAddresses(java.util.List)

trying to figure out what XBean will require in terms of XML for it

I have a SocketAddress property editor that takes simple strings like :22 or 127.0.0.1:1000 etc

so I'm good there

can't seem to figure out what XBean is expecting in the XML

ignoreing the constructor args, because I don't know how they apply to abstract classes :)

here's what I have been trying ...

<tagName closeOnDeactivation="true" defaultLocalAddress="something"/>

http://rafb.net/p/ucYBit46.html

um... I don't think that's right

let me look at the examples :)

getting complaints on socketAddress as you probably expect

man the docs suck

:)

I'll try to add some if I can get up to speed - this is good shit it deserves the docs (many people would love to use it)

http://svn.apache.org/viewvc/geronimo/.../RestaurantService.java?view=markup

@org.apache.xbean.Property nestedType="org.apache.xbean.spring.example.PizzaService"

ahhhh

in my case the nested type is a SocketAddress

which is not annotated since it's in the JDK

I guess this is not an issue

let me give it a try

IIRC the nestedType is there because the code doesn't understand generics

seems I'm stuck since SocketAddress is not in the mina.apache.org namespace

How can I tell XBean to interpret <socketAddress> as the proper Spring presentation?

brb I'm going to need some beer for this

back

Do you think I have to hand create a separate XSD for those objects like SocketAddress that I don't have access to the source?

dain, if you're busy I can come back later no worries, just let me know

sorry, got distracted

um, I'm not sure

np seriously

I can come back later

This stuff is killing me anyways

I'm not sure that xbean can handle a list of primitives like that

ahh ok

:(

I bet you could do defaultLocalAddresses="192.168.0.1:10001 192.168.0.1:10002"

let me try that

oh boy, seems like the parser hangs

<nioDatagramAcceptor id="datagramAcceptor3" defaultLocalAddresses="192.168.0.1:110 192.168.0.2:110" />

that flips out the parsing of the XML - just hung here

maybe a comma

tried that

no go

:(

I got the code loaded now, so give me a minute

np

oh intellij decided to GC

going to be a another minute

on ya... xbean-spring doesn't use the super awesome conversion libraries in xbean-reflect

it uses spring's tame conversion libraries

oh

you mean the nice set of PropertyEditors y'all have in there?

yeah this looks nice - I could put these PropertyEditors into my custom registrar to register in place of all the Spring stuff

xbean-reflect has a ton of property editors, but we don't import

don't want to overwrite the spring stuff since people depend on that

aye

akarasulu_: ask on the xbean list

post the spring you have and the xbean you want it to look like

dan or guillume may have added something for this

otherwise I can most likely hack something for you

thanks dain will do that

appreciate your help tonight

akarasulu_: try * @org.apache.xbean.FlatCollection childElement="recipe"

then don't add the wrapper element

this on which method the setDefaultLocalAddresses() method ?

never mind that isn't going to work

ah ok

the problem is you have a primitive

SocketAddress is considered a string I guess ?

ya

is socket address your class?

nope JDK

that's half the problem

I have a PropertyEditor registered for SocketAddress and that works like a charm when the list does not appear

if you use <value>92.168.0.1:110</value><value>92.168.0.1:111</value> it would most likely work

for example the property defaultLocalAddress is set just fine

but that is lame

hehe

let me try that but I guess I have to surround that with <defaultSocketAddresses> right?

yes

how do you feel about adding a SocketAddressFactoryBean

ok tried that and it just hung

adding that to what ?

sounds like you don't mind :)

I'll try to hack something

sure don't mind but I have no idea what yer talkning about at this point :)

dain I whipped together an email you want me to still send that to the ML ?

sure

okie sent

akarasulu_: where did you send it

xbean-user

I don't think there is a user list

I sub'd no problem

confirmed too

http://geronimo.apache.org/xbean/lists.html

humm there wasn't supposed to be a user list

i received it

I'm not subscribed to it

i guess I am

:)

low traffic

my filters are setup wrong

akarasulu_: wrote a working example for you

http://svn.apache.org/viewvc?view=rev&revision=689356

uses a static factory to create the SocketAddress instances

ohh coolio man thanks

basically puts it in your namespacke

ahhhh

looking ...

Interesting

So basically this Factory takes the place of a PropertyEditor ?

in this case yes

it should just work, but this is a good hack

ok thanks a lot

let me work it

site is working again http://geronimo.apache.org/xbean/custom-xml.html

akarasulu_: hope you're not still up :-)

hiya djencks

yeah unfortunately hehe

hey... don't you ever sleep :-)

good to see you

yeah I try

:)

looks like dain had some suggestions.... did you see them?

djencks: hey wanted to ask you a question

about the destroy method -- I haven't played around to see what happens, but I'd expect that you'd need to "annotate" the method on the concrete class

did you write a distributed TransactionManager implementation

yeah saw em and was going to use em

AFAIK you can't annotate interfaces and have that make anything happen

djencks, oh cool

hi dain

hi

dain: long ago for jboss

djencks, so there is a sort of inheritance going on

but geronimo doesn't

dain right

dain I'd need some idea about transport.... after that it is fairly easy

I'm asking because I was updating the wadi

dain e.g we could do it over the openejb protocol

and it is relocating method calls to the stateful instance

which isn't going to work for anything that relies on thead state

dain: I saw your email about that.... when can that actually happen?

which part?

dain tx needs to move

for the data to be elsewhere.... there was at least one call to it there

yep, and anything else riding on the thread

lets just talk tx

how can you get into a situation where a tx has to be moved?

dTX shouldn't be that hard since we already have import working for RAs

I didn't

akarasulu_: it might not look that way but I'm still listening :-)

was reading wadi stateful container and was suprised to see what they are doing

it isn't going to work without dTX

djencks, no worries

dain: (a) hard parts are agreeing on tx transport and writing an XAResource for the dtx client

(b) I still don't get how its possible to get into the situation.... I'd like a realistic example

ya recovery becomes a problem

you can't with standard openejb

I've been hacking on Terracotta lately, and have been thinking about a distributed TxLog

any node could recover for a failed node

with terracotta you can effectively create single cluster wide tx manager

dain are you thinking persistence through having the log on a bunch of machines rather than on disk?

you can configer terracotta to be persistent or not

if you have high confidence in your uptime you could run non-persistent

but it is risky

Not sure exactly how you see this working.... like a normal tx manager but any node can recover for any other,

so if machine A goes down any work done in un-prepared txs is lost?

absolutely rollback on a dead machine

any un-preprepared work

djencks, mind if I ping you tomorrow online about a few questions regarding the MINA related XBean stuff that you dealt with before ?

basically, you write a log that keeps everything in menory on a single node, and let terracotta distribute it to all the other nodes in the cache

akarasulu_: I'll be in and out, but sure

thanks djencks

dain: ok, how do you detect that its time to do recovery?

akarasulu_: np

akarasulu_: hope I can remember some of it :-)

:)

there are a few tricks, but the key is when a node crashes, all of the distributed locks of the node are released

not sure what that means, but it sounds like terracotta will tell you if a node crashes?

in a word, yes :)

ok

are you thinking of one log for the cluster or one log for each machine?

not sure... haven't gotten that far

k

I also don't really understand how recover actually works :)

but its so simple :-D

want me to explain it?

yes please

I assume you mean in geronimo, not xa in general?

in geronimo

ok

so we start up after a crash and read in the log

each tx has a bunch of entries

in the TxLog.recover(XidFactory) method

one entry is the name of the xaresource and the branchXid sent to the xaresource for this tx

(I don't remember the method names, but that sounds right)

this tx was prepared... that's what being in the log means

so we know we should commit any branch we possibly can

ok, so we've read all these in and sorted them by XAResource name

now we start waiting for xaresources to register with the tm

backing up

as soon as one registers, we use the name to find the branch xids we're supposed to commit

after reading the log we effectively have Map<XaResourceName, Collection<BranchXid>>

so we ask the XAResource for the list of uncommitted xids

yes

anything in that list and in the Collection<BranchXid> we commit and remove from the collection.

anything in the collection<branchXid> that the xaresource doesn't say is uncommitted must have been committed before the crash.

So we can now delete the Map.Entry for that XAResource name

after all the xaresources have re-registered we're done

so if something is in the log it should be committed?

the reason I did this is so you don't have to get all the resource managers up before you can do recovery

yes

as soon as a rm comes up, we can do the part of recovery relevant to that rm

that sounds very simple :)

in particular if one rm is really broken we can recover everything else, releasing the locks on that rm, and hopefully allowing some new work to proceed

that's what I thought :-)

sounds good

having a distributed log would be cool as the other nodes can immediately recover without waiting for the system to reboot

yes

probably faster too :-)

if the cache is still writing to disk it may not be... howl is pretty fast

also since we have this nice clustering system, we can effectively have one coordinator for the system no matter how many nodes participate in the transaction

I'd expect you'd use no-disk logs this way

I think mark little told me that corba transaction service only needed one coordinator, but I didn't understand how it works

similarly I don't see how it would work here

the coordinator does rpc to the XAResouces on the nodes in the tx

suppose you have 2 nodes talking to one db... I'm not sure what happens if both nodes try to use the same branchXid

I suspect it depends on the DB

I suspect most will object

I'd guess you'd need to branches

which means you're likely to step on your own feet

I mean if you have multiple coordinators, you'd be multiple branchXids

well, that's tightly coupled or loosely coupled.... whether the 2 branches can see each other's changes

I'm thinking that even with one global coordinator, if you try to access the same db in the same tx from 2 nodes, you may have problems

unless you have different branch ids for each node

does the branchXid contain the globalXid?

well, we're talking loosely here

an Xid has 2 parts

the global xid is the same for every branch in the tx

the branch xid is different

so the db can detect that two branches are from the same globalTx

yes

that's cool

so, I'd say each node has its own branchXid

and you get the isolation the db gives you :)

as usual :)

I gotta go to sleep

me too :-).... I think what you are talking about is very similar to the corba tx service

you might want to read up on it.... its a bit different from xa

k

anyway... good talking with you.... time for some zzz s

are there plans to add support for OSGi?

djencks: ping

jgawor: pong

hsn_: what kind of osgi support are you interested in?

djencks: i am interested in running OSGi modules inside Geronimo

jdk1.6 is supported by geronimo? WAS CE 2.1 refuses to install on jdk1.6

djencks, you around bro?

yes but give me a few minutes.... will ping you

np thanks d