pk n declara o DTD !?!?!

declara... se n declarasse eh q funcionava bem :p

barreiro: http://code.google.com/p/mobicents/issues/detail?id=279

pra reveres

ok, queres que corra o TCK e que veja no lab quanto é que isto marca ... certo ???

eh performance q se note? ;-)

n, n ? performance q se note muito

pois.. mas qq coisa eh boa :)

bem.. usando o bixaroco como singleton funcionou bem :)

n consigo faze um teste a s?rio, n sei se ? do stack de sip, se ? do meu pc devido a updates, se ? de n conseguir compilar com debug a off

enfim

ah barreiro

n sei se sabes mas pra configurares o log tens de configurar o do jboss

por la a categoria do org.mobicents a WARN

ou meteres um <level value="WARN"/> dentro do <root>

geralmente faço uma ronda de teste ... e depois removo tudo do root :D

se n o isDebugEnabled vai vir a ture

remover tudo do root n te serve de muito

pq por default tem level debug

e quando ele faz o isDebugEnabled e n acha a categoria (ele n ve as categorias do log do mobicents) vai usar essa

i.e., isDebugEnabled()= true :-\

mas se estiver a categoria do org.mobicents a WARN e o root sem level, ie a DEBUG, o isDebugEnabled() ja n da true?

ja n, mas tem de ser na config do jboss

e n funciona pra outros loggers q n estejam na config

sim sim... pensei q o isDebugEnabled() fosse mais global :P secalhar por isto estar assim fikei c essa ideia

ele vai do logger ate ao primeiro pai q tenha o LEVEL definifo

se tu n declaras na config n tem o level definido log vai para o pai

se n achar nenhum (desempacotando) ? o root

vou fazer reboot, brb

ja n fa?o ha n sei quanto tempo

hasta ma?ana, companheiros!

porteivos.

barreiro: o sipp da-me esta info estranha

em 119 165 chamadas

50935 out-of-call msg (discarded)

isto s?o byes

podem ser. Acontece por várias razões ...

experimenta por um <pause> depois de enviar o 200 OK do bye, para em caso de retransmissão a chamada ainda estar a decorrer.

de quanto

1s ou 2

assim tenho bue retransmissoes do ultimo 200 ok

entao e q treta ? esta nos logs do sipp

2008-07-16 03:17:48: Aborting call with Call-Id '9990-793@127.0.0.1'.

e n diz mais nada

?s paletes

espera lá ... quem é suposto enviar o BYE

o AS

manda BYEs a mais q se farta

e o sipp sempre a retransmitir 200 oks

é porque algo não está a corresponder ...

corre o sipp com -trace_msg ... uma chamada só ...

cria um ficheiro de log com as mensagens ... dá lá uma vista de olhos.

o teu patch passou o TCK, mas só à segunda ... por causa da treta dos profiles.

entao

tens probs com os profiles?

olha o trace do bye e 200 ok

----------------------------------------------- 2008-07-16 03:26:44

UDP message received [294] bytes :

BYE sip:sipp@127.0.0.1:5050 SIP/2.0

Via: SIP/2.0/UDP 127.0.0.1:5060;branch=z9hG4bK64e8debdf2823fafa8019dcf750bed54

CSeq: 1 BYE

From: "sut" <sip:service@127.0.0.1:5060>;tag=8294fc7a

To: "sipp" <sip:sipp@127.0.0.1:5050>;tag=2

Call-ID: 2-879@127.0.0.1

Max-Forwards: 70

Content-Length: 0

----------------------------------------------- 2008-07-16 03:26:44

UDP message sent:

SIP/2.0 200 OK

Via: SIP/2.0/UDP 127.0.0.1:5060;branch=z9hG4bK64e8debdf2823fafa8019dcf750bed54

From: "sut" <sip:service@127.0.0.1:5060>;tag=8294fc7a

To: "sipp" <sip:sipp@127.0.0.1:5050>;tag=2

Call-ID: 2-879@127.0.0.1

CSeq: 1 BYE

Max-Forwards: 70

Content-Length: 0

não ... mas houve um teste que falhou queixando-se da falta de um profile ... tive que correr o TCK outra vez.

o header contact n esta no 200 ok

essa linha branca a seguir ao CSeq é mesmo assim ou foi problema de copy-paste ??

deveria ser o contact

mas como o bye n o traz ...

se clahar ? melhor adicionar o contact ao bye n?

n me admirava se o jain-sip enviasse sem contact mas exigisse. É melhor adicionar sim, porque é um header obrigatório.

continua a n gostar do raio do 200 ok

RA novo !?!?!?

agora o sipp faz sempre pelo menos uma retransmissao do ultimo 200 ok

----------------------------------------------- 2008-07-16 03:49:18

UDP message received [345] bytes :

BYE sip:sipp@127.0.0.1:5050 SIP/2.0

Via: SIP/2.0/UDP 127.0.0.1:5060;branch=z9hG4bK6036c682c9d4f72890c5ea5e357d522b

CSeq: 1 BYE

From: "sut" <sip:service@127.0.0.1:5060>;tag=69b1c1bc

To: "sipp" <sip:sipp@127.0.0.1:5050>;tag=2

Call-ID: 2-1074@127.0.0.1

Max-Forwards: 70

Contact: "Mobicents SIP AS" <sip:127.0.0.1:5060>

Content-Length: 0

----------------------------------------------- 2008-07-16 03:49:18

UDP message sent:

SIP/2.0 200 OK

Via: SIP/2.0/UDP 127.0.0.1:5060;branch=z9hG4bK6036c682c9d4f72890c5ea5e357d522b

From: "sut" <sip:service@127.0.0.1:5060>;tag=69b1c1bc

To: "sipp" <sip:sipp@127.0.0.1:5050>;tag=2

Call-ID: 2-1074@127.0.0.1

CSeq: 1 BYE

Contact: sip:sipp@127.0.0.1:5050

Max-Forwards: 70

Content-Length: 0

o via n esta mal?

n, deve ser o mesmo do bye axo

acho que n, mas vou confirmar.

bem, as mensagem parecem-me correctas ...

o retrans="750" no 200 ok n faz o mesmo q o pause a seguir?

penso que não, porque assim que ele envia o 200 OK do bye a chamada é descartada ... e com o pause fica aberta.

vou experimentar com o ra antigo

a mesma coisa

vou experimentar o stack 1.2.1

de qq modo o stack n devia papar o 200ok do sipp e deixar de mandar mais byes? pra q ? q o sipp vai mandar mais 200 oks, n atrapalha mais?

o 1.2.1 esta na mesma

por algum motivo n o papa, e pensa que se perdeu o BYE ... e volta a enviar.

os seja, o jain-sip tb pensa que é uma 'out-of-call message' !!!

ppelo menos o novo RA so me crasha ?s 130 e o antigo ?s 120 berra logo

pelo menos ja merece um esfor?ozito

tou mesmo com a ideia q s?o estes byes q mandam o stack abaixo

? muita msg

uma coisa que já reparei é que o stack funciona melhor com 100 INVITEs/seg do que com 50 INVITEs + 50 BYEs/seg

anda muita gente a brincar com o xdm ultimamente

pena ? n ter ajuda nenhuma

neste momento, o que está no roadmap ???

... além dos testes :P

os testes, os issues abertos, o sbb cliente do presence service e o rls

mudar pra este ra de sip

SBB cliente = interface local, certo ???

e mudar o ra de http pra n usar http sessions como activity obrigatio e em vez disso ir para algo do tipo do sip ra, usa-se requests como activity ate q o sbb cria a http-session

sim, mexer com o ps atraves de um sbb

com uma implementa??o primeiro interna, e depois uma externa tb

mas a ultima n ? tao prioritaria

esse sbb n ? tao pera doce como parece

n querias tb algo do tipo do enabler de presença ???

precisava de uma semanita pra me concentrar nisso

e talvez ter de alterar qq coisa no ps pra faciltar

sim, claro. Não é chegar à BD e alterar uns campos assim à papseco ...

... é uma interface nove de raiz, que em vez de receber mensagens da rede recebe invokacões de métodos.

bem ... tou a ter uns java.lang.OutOfMemoryError: GC overhead limit exceeded .... amanha vou despatchar isto e testar com menos carga.

tas com a flag no event router a true?

MONITOR_UNCOMMITTED_AC_ATTACHS = false;

hey guys

can't java use accented chars?

like: String soufl?Flavour = "...";

I don't think so

would be a pain to share code no ?

yeah.. There's a guy with problems with some accented chars in jDiameter (?). but it works fine for me..

I think best practice is to avoid that

yeah, those were there by mistake anyway... just finding odd that it fails in other env

hi team

back to good old Europe

the troubly in Bulgaria starts next week. watch the news

so let's get it on

hi.

say hi to extravaganza for me

#1 Mobicents All 1.0CR1 status

perfect timing

#2 MMS July Release date?

#3 MSS Release date

#4 Diameter + testsuite release?

#5 MSS PBX status

#6 QE update

#7 docs

other topics?

good to see so many releases :)

alexandre: wait until they are effectively released :-)

lets start then

#1 CR1 status - Eduardo?

hi

Hi Team

we are almost ready, have some critical issues to close

specifically? what's outstanding

278, 103

and 279

103 can be closed, I need only someone to verify

which means? matter of time or we need to make a decision?

Alex, can you please verify 103?

103 and 279 are patches just about to be committed

278 is the logging issue

for 278 I will test according to Martins latest instructions

is the one that needs a decision

I haven't able to reproduce the problem so far

278 is not critical IMO

Eduardo, what decision is needed on 278

I've checked #103, and reported to baranowb that it wasn't working, forgot to comment on issue :S

one thing martins, if you have if(isDebugEnabled) without logger.warn in if loop, still it goes in if loop?

a solution or drop our own logging config framework

abhayani: have you tried what I added in the issue?

not yet

thats what I am saying I will have to give it a shot

looks like all these 3 can be decided on promptly and we can attempt a CR1 release by Friday. Is that correct?

in my opinion I would drop it and leave for 2.0 and jboss 5

easiest solution

and we get time to work on that

drop what?

the inconvenience of restart or the custom log

our log4j config framework

restart is fixed

I don't think we can afford to drop our log4j config

the restart of log was another issue, already closed

so 103 is not restart?

103 is about the sbb object pools

278 is about isDebugEnabled returning true in core whatever config you throw in our log4j.xml config

?

I can see that 103 patch works

it has to be something different

SbbO are returned properly

when all entities are removed

and I uninstll DU

no active objects remain

baranowb: it shoudl work, it's just a line change and I don't see side effects from dropping a sbb object on a error codintion

I'm using it atm and have no issues

I will just propose a change of the object pool config to not be able to make objects forever and claim them when idle, but that's just apache pool config, nothing else

ok.. I've tried the patch with sipp and after ~9K SIP MESSAGE the server stopped responding... as it happened with patch

we will talk offile

*as it happened without patch

what sipp script

one baranowb sent me

Message >>>

but it's just sending MESSAGE

it may be that

<<< OK

try with the load test services

ok.

so ivelin, 278, the isDebugEnabled prob

drop our log4j config or look for a solution to make it work

martins: easiest fix isn't to add level to root ?

If we do that we'll have to modify the main log4 conf

y, same procedure as before

hold on

a copy of jboss xml with the mobicents categories

no good

one day is what I am asking for

Amit, do it

we're not going to modify main config files

that's asking for certain trouble

ok, so 1 real outstanding issues and 2 that are almost closed

as long we don't go for the release with that isDebugEnabled returning always true I'm fine with it

let's shoot for first internal release on Friday

:)

Amit will let us know by tomorrow

k

y

pretty good progress any way - from 60 to 3 in 10 business days is not bad

we are having some issues with the sip stack

sip stack again? or SIP RA?

but well, without not pushing too much the performance it's acceptable

sip stack

its akward handling of ACK

talked to Ranga

and he doesnt want to make changes

what's the problem ?

well, I don't really care about that one, we can pass over it in the ra

what are the problems then ?

that one is the fact that the stack

creates a fake ACK server transaction if a dialog is present

but not otherwise

if no dialog, no stx

secondly

but as I said we can, and we are passing it over by impleemnting a fake ack server tx

TX Term doesnt have references to objcts

baranowb: can that be because it doesn't have an internal app object?

it has

I mean it shoudl preserve reference to our wrapper

you mean the app data is null when the listener is called ?

but it doesnt

jd; yes

more important is the fact that our load tests sends a bye to sipp to close calls, sipp returns the 200 ok and the stack keeps sending byes

seems to have trouble to match the 200 ok with the bye

strange we use it and it seems we still have the app date in MSS

this results in a big load of msgs that hurts a lot performance

and may bring the stack down

can you reproduce it with a simple isolated jsip test case ?

but ok, I can live with it, 100 cps is handled without a prob

I can go 120 cps on the new ra

a little less on the old

100 cps is a nice value

but then the thousands of BYEs really kill it

maybe tonigh, in the night shift, can again brainstorm with barreiro on it

anyway, those are minor issues, the release should not wait for it

Is that something that only occurs in SIP RA but not MSS?

dunno

let me lauch a simple test case to verify that

in mss load tests who sends the bye

the app data to null

I mean

the app

jeand: what ?

baranowb: in MSS we don't have that problem it seems

martins: that should be easy to debug with wireshark..

jeand: that's only in ACK txs

may be missing tags (sipp does this sometimes)

sipp misses a lot

I've confirmed the BYE and 200 ok are good

so it's not sipp

maybe sipp doesn't handle correclty the load

just kidding :-)

hehe

jean, what I was aksing

in mss load tests

who temrinates the session

sipp or mss?

MSS application

sends the BYE

ok

then we need to check if it happens also there

but the maximum load we got was about 120 on QA labs

barreiro: can you share with me the load test project for the stack only?

today

martins, of course.

ok, if it can't be reproduced on MSS, let's not blame sipp

but its an offline topic it seems, please take it to the list

anything else around cr1?

martins: post the logs

n

baranowb: I don't have the null app data on processTXTerm in MSS

just ran a test

jeand: only happens in ack txs

oh ok

offline please

next topic

#2 MMS July release date?

ok

is it decided yet?

STUN and speex being two key features

we are going to next week

July 22?

I've commited regular tests for speex, found error with RTP timestamp

need to correct it

let July 25

Lets keep the release date close to July end

yes Oleg

thought it was end of month

ok, July 25 sounds good

Luis, is the MMS test working now at your end?

I mean with Hudson

Vladimir is STUN going to be ready?

abhayani, I haven't tried out yet.

ok

guys, I need feedback for mms

stun is ready, but i might add some extra features to imprive lookup time in a special case of NAT

what's the special case?

the case when there is no port mapping (a.k.a. PAT)

please elaborate a bit

i.e. when the NAT receives something on port 1024 it will forward it to port 1024 on the machine behind the NAT

thie case there is no port mapping

usually there is port mapping, - when the NAT machine receives something on 1024 it can go to port 3435 on the machine behind the NAT

are both machinces on the same subnet?

the NAT is supposed to have two network interfaces, one on the same subnet, one on the internet

this is not very precise though - subnet = private subnet

internet = internet subnet or AS

AS = autonomous system, but it;s not important , you get the point

yes, I do

why wouldn't there be mapping

assumed same ports?

but different IPs

?

connection problems?

lets move on

if you have a cable modem like me, the modem takes over the IP, but it is connected to exactly one machine so they can have different IPs but there is no reason to map ports

got it

how can you tell in this case that there is no mapping involved?

i will just let the user specify it

advanced users would know

but isn't STUN about auto-traversal?

Vlad, I guess port translation is not involved in NAT, addresses only

I think I was told at some point that NAT traversal for media requires a public bridge that the RTP stack knows about and responds to

STUn can determine the type of NAT with some degreee of certain-ness, but I'd try to avoid using it because of the overhead

so vlad you reimplemented STUN ?

i dont think you need another device to traverse NAT

well, Michael already reported that his app works behind the NAT, so there is no doubt

I dodnt reimplemnt STUN, I am just trying to save some STUN lookups by specifying if PAT is needed or not

if there is PAT, you have to do a new STUN lookup everytime you open a port

ok

but how does skype figure that out by itself?

If there is no PAT you can just do one lookup to determine the public IP of the NAT and use it everytime you open a port

what does a lookup consist of

'?

skype is easy, it should handle only a few RTP ports and it can do the mapping at anytime

MMS can have much more connections and has to do much more lookups

lookup tries to connecto to the stun server, the stun server sees the NAT IP and port and reports back to the client, then the client compares the NAT ip and port with it's own and sees if there is a mapping

then, there are other steps to determine if the NAT keeps the mapping for particular remote machine or every new remote machine must use different port, then it tries to check if the mapping are preserverd and for how long and a lot of other stuff

so Skype has a global STUN server that each client uses to find out the nat setting?

yes

but we rely on a global server too

you specify it in the settings

are there reliable free global stun servers that we can set by default?

yes, but by default STUN should be disabled

host names?

still it would be helpful to have a few default ones, do we list them in the MSS distro now?

yeah like stun.ekiga.net

in http://www.mobicents.org/stun.html

not sure if we list any in MSS, they are easy to fund

we provide a default one

stun01.sipphone.com

ok, that's cool

thank you

next topic

#3 MSS 0.5 release date?

and we provide a link to STUN servers

http://www.voip-info.org/wiki-STUN

currently still 1st of August

we still have a few big issues to close

such as concurrency control

which Fran is taking

clustering which is next on my list after fixes to laod balancer

vlad is on the Open IMS integration

Fran is usually quite busy with his projects. Is he commited to August 1 for concurrency?

it seems so yes

I checked with him this morning

saw you jbc mvcc message

that's great, but my concern is that its probably node level locking vs node map entry level

and he said that the refactoring of the SipApplicationDispatcher to break it down was very helpful

posted a question to the blog post. lets see what Manic says

ok :-)

well, good to hear. let's see

one question

will that include sipsession queuing?

not sure if he will have time for it

but vlad may help once he is done

BTW the bug in JBON prevents us from integrating into it right ?

FYI guys, i will be busy with EAP installers today and tomorow

if so I'll reschedule this issue to 0.7

vralev: do you think you can be done with the Open IMS by end of Friday ?

or end of week end

jeand: yes

cool

btw it is working right now with "the hack"

if you really need it

lol "the hack" :)

no I meant working with the clean stuff To tag

:)

what number was the jbon bug

I'll bring up again to the JBON team

125

ok

sorry that is the MSS issue of JBON integration

RHQ-622

regarding PBX, I talked with Thomas

he had not the time to continue much with it. It is still a big prioirity to them but customer projects prevails and vacation time so I don't think we will have it for end of month

ok, good to know at least what's the deal

I need to step out. Jean, please continue and send notes at the end if you don't mind.

ok no problem

enjoy your time in bulgaria

I'm sure you will ;-)

ok that's it for MSS

#4 Diameter + testsuite release?

alexandre:

that's me.

Diameter expert :-)

ok, have been working on it, JUnit tests have revealed some issues with the RA (as expected)

why expected, you're ajbossian, jbossian guys don't do bugs

been fixing them, while cleaning most of the FIXME: and TODO: so a much stable and coherent version should be available in this release

only features

undocumented, uncharted

bad news - is that we were supposed to get CCA from Erick

what is the expected release date ?

yeah, that. but "the others" are not ready for it :D

but we will have to do it by ourselves

also

what is CCA already ?

after looking a litle bit more into "pplugin" RA types into base impl

it seems not asuch a good idea

expected to release it by the end of this week, beggining of next

Credit Controll

george: why not a good idea ?

I dont like idea

alexandre: has it new features or only a bug fix release ?

of having exclusion list :)

george: guess we can go with Sh and wait for Erick to do CCA and after that, we move to Ro/Rf

some ra types use same events - to which we will deliver event?

Im not looking into waiting again

jeand: some new features, but on internal level... like the AVP Dictionary

especially after he dissapears during each chat

he is not reliable to wait for CCA

ok, but as I've told you, I'd like to have a more stable base before moving on to next

I will be doing a lot of opuce

and MMS now :P

so theere is some time

btw i really need feedback: http://groups.google.com/group/mobicents-core/browse_thread/.../c472a3bf9bf8759

http://groups.google.com/.../e5f35de252f5d9f8?lnk=gst&q=pool#e5f35de252f5d9f8

alexandre: if the base is more stable, will it improve the speed of the extensions ? or is it not related ?

I guess

we can simply extend BASE RA Impl

to have super handling of base

in extensions

should be better way

jeand: not worried on the speed of it, but more on the arch of it. there are some decisions that were taken and now with the usage I can see better options. but yeah, it will improve :)

the simple load tests I've made showed a good response

rate ?

I meant the speed of dev but that's good too :-)

250~300 msg/s

alexandre: can those load tests be automated and run in hudson ?

jeand: ah :) pretty much also, there's a lot of things making the usage of the RA more difficult than what is expectable

yets, I've been doing them using ericsson diameter sdk, but they can easily be replicated into JUnit

cool

do we already have a hudosn job with a testsuite for diameter ra ?

testing diameter RA features

not yet

how much time do you think is needed to refactor the Base RA so that it is convenient to use ?

the JUnit tests that are being dev will be for that

ok so with the release next week, the testsuite will also be completed ?

yes

great

perf tests also ?

hopefully :)

the release wll include the refactoring of the Base RA so that it is more convenient to use ?

or it's scoped for another release ?

sure, that's what I'm most worried about.

This first release isnt very stable, and I've noticed that even more with the tests

ok I see

what's planned after the release ?

that's it from me :)

advancing with other RAs, such as CCA and Sh

and Diameter Servlets ;)

:-)

which will benefit from these refactoring being made also

ok great stuff, all this work will benefit Sip Servlets too :-)

thanks alex

thanks

oh just another question

CCA and Sh are other RA or the same base one extended ?

iirc they have different ratypes for these

should be new RAs

ok

bartosz seems to have some ideas to do them as plugins, we'll discuss it later and see what gives

ok I guess you can share your findings on core or public then

y

ok thanks

#5 was already done as part of #3

so #6

#6 QE update

barreiro: ?

me.

Last week: performance tests with MSS and Load Balancer, that shown some problems on LB

jeand is working on that.

yes I'll do some profiling today and tomorrow

and give you a new version to test tomorrow

Next were the release scripts for MMS ... and the matching hudson job.

mow we have nightly builds of