MSS 2.0 Weekly Meeting - April 19th, 2011 - 5:30 PM (Paris Time)

[Jean Deruelle]

Anyone from the community interested in participating to the weekly meeting or in MSS development is welcomed to join every Tuesday at 5:30 PM Paris Time or comment offline on the chat log every week if time zone is an issue

Attendees : Jean, Vlad, Thibault, Sri, Shay

Summary

#1 MSS in Arquillian - Sri is starting MSS on AS5 remote integration in Arquillian. Also as part of the SIP test suite a java interface to play sipp script in under development in https://github.com/srinarayanan/jsipp As of now the parser and basics request creation is done.

#2 MSS in the cloud - MSS Images working on Amazon EC2 as cluster now integrating management for auto scaling. Code committed to http://code.google.com/p/mobicents/source/browse/trunk/cloud/sip-servlets and user guide http://code.google.com/p/mobicents/source/browse/trunk/cloud/sip-servlets/cirrasM-build/migration_to_the_cloud.pdf

#3 MSS 2.X and CTF - released today http://www.dzone.com/links/mobicents_sip_servlets_200alpha2_released_.html & http://www.dzone.com/links/cdi_telco_framework_100alpha1_is_out_.html

#4 Telemedecine application - moved from proxy application to a B2BUA application. sipdroid and jitsi are talking to each other with audio working OK, figuring out video now.

Location irc.freenode.net/8001 Channel #mobicents

Log

-----------------------------------------------------------------------------------------------------------------

17:32:13 narayant: hi am I late
17:32:19 jeand: hi sri
17:32:19 narayant: just arrived from office
17:32:23 jeand: no problems
17:32:25 narayant: good evening everybody
17:32:33 tleruitte1: hi guys
17:32:36 shaym: hi
17:32:39 jeand: we will wait for 5 more minutes as georges has not arrived yet
17:32:43 jeand: hey guys
17:34:04 jeand: is anyone in a hurry to pass first or we stick to the regular order where narayant is first because of timezone ?
17:34:27 tleruitte1: i have plenty of time so do as it's best for you 
17:34:39 shaym: works for me
17:34:43 jeand: lucky you 
17:34:48 narayant: 
17:35:22 tleruitte1: well, i'll try to manage rhq in the meanwhile 
17:36:03 narayant: so shall I start off
17:36:13 jeand: #1 MSS in Arquillian
17:36:13 jeand: #2 MSS in the cloud
17:36:13 jeand: #3 MSS 2.X and CTF
17:36:13 jeand: #4 Telemed example
17:36:25 jeand: yep go ahead on #1 narayant
17:36:59 narayant: As of Arquillian in MSS , I started with one of the original approach that aslak had  suggested
17:37:19 narayant:  having the jboss  As 5.1
17:37:31 narayant: container support and pointing into to MSS
17:38:08 narayant: That is in progress
17:38:26 jeand: I noticed on the thread with aslak
17:38:32 jeand: that you're trying EJBs right away
17:38:38 jeand: I would suggest first
17:38:39 narayant: yes
17:38:48 jeand: to try with a simple sip application packaged as a war
17:38:49 narayant: its should work after aslaks last comments
17:38:59 jeand: to make sure this works correctly first
17:39:07 jeand: and don't need specific hacking
17:39:18 jeand: and then only add the EJB part and bundle it as EAR
17:39:23 jeand: or have 2 different apps
17:39:28 narayant: ok , Jean u had written a simple servlet test
17:39:32 jeand: to showcase kinda like a tutorial
17:39:47 narayant: can u send the link
17:39:56 narayant: I deleted the local copy
17:40:28 narayant: jeand is talking about the thread in arquillian forum
17:40:28 narayant: http://community.jboss.org/thread/165203?tstart=0
17:40:42 narayant: Also I have started to work on jsipp
17:40:56 narayant: which will be able to play SIPP script from within Junit
17:41:09 narayant: https://github.com/srinarayanan/jsipp
17:41:16 narayant: still some more work to do
17:41:30 narayant: I would like to update on what I have done as of now
17:42:04 jeand: kind of something similar to https://github.com/deruelle/arquillian/blob/ARQ-250/containers/mobicents-sip-servlets-embedded-1/src/test/java/org/jboss/arquillian/container/mobicents/servlet/sip/embedded_1/MobicentsSipServletsEmbeddedClientTestCase.java
17:42:14 narayant: ok jean
17:42:16 jeand: instead of Embedded it will be remote
17:42:21 jeand: on AS 5.1
17:42:34 narayant: i will take that
17:43:33 narayant: SIPP has send and receive actions basically
17:43:55 narayant: I planned to use jain sips - sip provider sender to send the request
17:44:15 narayant: and the SipUnits' sip phone as the receive mechanism
17:44:40 narayant: Also the SIPP script has variables like [call-id] ,
17:44:49 narayant: [last_from]
17:44:51 narayant: etc
17:45:07 narayant: which will be covered by a strategy pattern
17:45:17 jeand: why don't you go full JAIN SIP or full SipUnit instead of a mix ?
17:47:02 narayant: Sip unit has a better suited wait method
17:47:17 narayant: which will wait for specifica time fot the receive tag
17:47:29 jeand: so you go full SipUnit ?
17:47:51 jeand: or did I misunderstood ?
17:48:07 narayant: the MessageFactory will be needed to generate Request from String format
17:49:09 narayant: Request sipRequest = messageFactory.createRequest(filledSring.trim());
17:49:17 narayant: this is from jain sip
17:49:29 narayant: which is well suited for creating request from String
17:49:34 jeand: beware because you might need the same thing for subsequent requests
17:49:37 jeand: or even responses
17:49:54 narayant: but is there any problem in having both
17:49:54 jeand: depending on how you act as (UAC or UAS)
17:50:07 narayant: I mean sip unit and jain sip
17:50:13 jeand: I'm not sure never tried it
17:50:14 narayant: sip unit any way used jain sip
17:50:25 jeand: but if you don't use sip unit data structure from the beginning
17:50:35 jeand: it might not know about your call
17:50:39 jeand: and might not work
17:50:40 narayant: I have prevously used this approach in direct Junit testing
17:51:15 jeand: ok just putting a warning here on the fact that this may not match 1-1 to sipp scripts
17:51:29 narayant: yes , I agree
17:51:55 narayant: having to support complex sipp script will be a little complicated
17:52:11 narayant: but for mostly used scenarios it should be fine
17:52:37 jeand: it's always a delicate balance
17:52:47 jeand: but the more you will get community users on this project
17:52:55 jeand: the more you will need to support complex 
17:53:54 narayant: 
17:54:11 narayant: First I will get it work for my SIPP xmls
17:54:17 narayant: then make it generic
17:54:19 narayant: 
17:54:24 jeand: your call
17:54:58 narayant: jeand, any jain sip open forums
17:55:06 jeand: yes
17:55:13 narayant: can u post the link
17:55:31 jeand: you need to sign up here with an account https://java.net/people/new then login with the username and password at https://java.net/people/login?original_uri=%2Fprojects%2Fjsip then subscribe to the mailing lists here http://java.net/projects/jsip/lists
17:55:40 jeand: Then you will be able to post there.
17:55:48 narayant: ok will do that
17:55:51 narayant: one more thing
17:55:58 narayant: is regarding shrikwrap
17:56:09 jeand: yes go ahead
17:56:20 narayant: I see that wee have to add individual classes to create a jar archive
17:56:32 narayant: but a war usually has 100s of classes
17:56:50 narayant: are there any simler apis to include all classed
17:56:58 narayant: classes * from the project
17:57:16 narayant: in short how do we create a full EAR programmatically
17:57:19 narayant: ?
17:57:37 narayant:  return ShrinkWrap.create(JavaArchive.class, "test.jar")
17:57:37 narayant:          .addClasses(GreetingManager.class, GreetingManagerBean.class);
17:58:34 jeand: you might be able to do addPackage
17:58:36 jeand: or something
17:59:16 narayant: createFromZipFile(Class<T> type, File archiveFile)
17:59:20 jeand: yes addPackage or addPackages
17:59:25 jeand: would work
17:59:27 jeand: on the latest one
17:59:31 narayant: does this support ear / war
17:59:43 jeand: http://docs.jboss.org/shrinkwrap/1.0.0-alpha-12/org/jboss/shrinkwrap/api/container/ClassContainer.html#addPackages(boolean, org.jboss.shrinkwrap.api.Filter, java.lang.Package...)
17:59:53 jeand: yes
18:00:06 jeand: use addPackages instead of addClasses
18:00:13 narayant: sure
18:00:38 jeand: well ear doesn't contain any classes
18:00:44 jeand: so this would work on war and jar
18:01:11 jeand: and ear will package those
18:01:58 narayant: so u mean to say use add packages repeatedly to create an EAR
18:02:28 narayant: ?
18:02:44 jeand: no
18:02:53 narayant: createFromZipFile
18:02:56 jeand: I mean use addPackages to add multiple classes
18:03:01 narayant: The file is expected to be encoded as ZIP (ie. JAR/WAR/EAR)
18:03:15 narayant: from the doc it seems to support ear
18:03:21 narayant: http://docs.jboss.org/shrinkwrap/latest/
18:03:34 narayant: so its should work even for EAR I guess
18:03:41 narayant: will update after a sample check
18:04:09 narayant: so thats it from me
18:04:35 narayant: I will have more interesting  updates next week
18:04:36 narayant: 
18:04:41 jeand: final EnterpriseArchive ear = ShrinkWrap.create(EnterpriseArchive.class, "test.ear")
18:04:42 jeand: .addModule(ejb)
18:04:42 jeand: .addModule(war)
18:05:06 narayant: ok
18:05:10 jeand: this is how you create an EAR pending the war and ejb are respectively JarArchive and WarArchive created previously
18:05:29 jeand: ok thanks for the update narayant
18:05:40 jeand: what's your expectation on where you can be next week ?
18:06:12 narayant: i should be able to get 5.1 container  checked with remote on MSS
18:06:36 narayant: we have a (company )release planned  next tuesday
18:06:50 narayant: I am not sure how much time I will be able to spend
18:07:12 jeand: ok
18:07:26 jeand: hopefully we can create a sip application archive
18:07:32 narayant: yes
18:07:36 jeand: and deploy it
18:07:38 narayant: thats the first task
18:07:47 narayant: i will take up jsipp after that
18:08:29 jeand: https://github.com/deruelle/arquillian/blob/ARQ-250/containers/mobicents-sip-servlets-embedded-1/src/test/java/org/jboss/arquillian/container/mobicents/servlet/sip/embedded_1/MobicentsSipServletsEmbeddedInContainerTestCase.java
18:08:38 jeand: was using org.jboss.shrinkwrap.sip.api.spec.ConvergedSipWebArchive
18:08:52 narayant: one more thing arquillian has beed refactored to core and containers
18:08:56 jeand: not sure if they included it in the latest releases of Shrinkwrap though
18:09:08 jeand: narayant, what does it changes for us ?
18:09:18 narayant: and that helped in resolvong the build issues
18:09:21 jeand: you mean different git repos ?
18:09:29 narayant: yes
18:09:36 narayant: we have to just pull our core
18:09:47 narayant: and create container support directly
18:09:51 jeand: I see
18:09:55 narayant: as of noww
18:10:10 narayant: I just ppulled core and AS5.1 support
18:10:19 jeand: ok
18:10:21 narayant: and both builds fine
18:10:31 jeand: good
18:11:01 narayant: jeand are there any planned dates for Sip servlet
18:11:03 narayant: suuport
18:11:16 narayant: I mean mobicent + arquillian
18:11:21 jeand: not yet
18:11:26 jeand: will depend on you 
18:11:34 narayant: ok , 
18:11:41 jeand: if you get it to work they can include it
18:11:51 jeand: I'm not able to spend any cycles on this currently
18:11:56 jeand: so I rely on community
18:12:08 jeand: ok thx
18:12:13 narayant: ok
18:12:19 jeand: #2 MSS in the cloud
18:12:26 tleruitte1: hello 
18:12:26 jeand: tleruitte1,  your turn 
18:12:57 tleruitte1: so, i've commited part of my work
18:12:58 tleruitte1: https://github.com/tleruitte/cirrasM-build
18:13:43 tleruitte1: i've called the project CirrASM, because it's kind the same than CirrAS, but with Mobicents instead of JBoss
18:14:07 tleruitte1: so with cirrasM-build, you will be able to build the appliances on ec2
18:14:37 tleruitte1: the README explains how to build the appliances, it's quiet straightforward 
18:15:01 tleruitte1: more explanation are given on how to build the appliances on migration_to_the_cloud.pdf
18:15:14 tleruitte1: this doc also explains how to use the appliances
18:15:39 tleruitte1: (the LB is ready to use out of the box, and you need to upload your app to the mobicents appliance)
18:16:35 tleruitte1: any comment so far?
18:17:09 vralev: does it work already out of the box?
18:17:34 tleruitte1: vralev: what are you talking about?
18:18:06 jeand: let me read the thing a bit
18:18:11 vralev: LB+some nodes - do they work after installation
18:18:21 tleruitte1: jeand: sure 
18:18:51 tleruitte1: vralev: amost, you can read migration_to_the_cloud.pdf for more details
18:18:59 tleruitte1: basically, one step is required
18:19:09 tleruitte1: two actually
18:19:22 tleruitte1: 1) upload your app to the mobicents appliance
18:19:37 tleruitte1: 2) set the LB's ip in the mobicents appliance
18:19:43 tleruitte1: and then it works 
18:19:50 jeand: nice
18:19:54 jeand: just finished the reading
18:20:08 jeand: would be interesting to add a step by step guide
18:20:16 jeand: with instructions with a real application
18:20:19 jeand: and testing of it
18:21:02 tleruitte1: jeand: i'll do that! i'll be useful for the redaction of my thesis too
18:21:13 vralev: so no NAT hacks needed?
18:21:38 tleruitte1: well, there is an issue i'd like to talk about 
18:22:11 tleruitte1: i think it would be a good idea to allocate an elastic IP to the LB
18:23:06 jeand: can you expand on that ?
18:23:15 tleruitte1: the problem is that the LB wouldn't be bind on the elastic ip, but on the internal ip
18:23:25 tleruitte1: are you familiar with the networking config of ec2?
18:23:54 jeand: not so much
18:24:06 tleruitte1: basically, each instance has a public ip and a private ip
18:24:59 tleruitte1: the private ip is used for the communication between the instances, the traffic is free (no need to pay), and those ip are unreachable from outside the cloud
18:25:31 tleruitte1: the public ip is reachable outsite and inside the cloud, but this traffic isn't free
18:25:51 tleruitte1: each time you start an instance, you get a new public and private ip
18:25:57 jeand: I see
18:26:22 tleruitte1: the idea of elastic ip, it's a public ip that you can attach and detach to an instance, so it's always the same ip
18:27:32 tleruitte1: i think i would be good to have an elastic ip for the LB, and use the private ip for the communication between the LB and mobicents nodes
18:27:59 jeand: yes indeed
18:28:05 tleruitte1: the problem is that the LB would have to use two different IP
18:28:17 tleruitte1: in the configuration fo the LB, it is bound to one IP
18:28:24 vralev: i think there is no need for that - may be binding internal host to the provate adapter and the external to the outside will be enough
18:28:46 vralev: the LB binds to two IPs actually
18:29:01 vralev: by default 127.0.0.1:5060 (external,public)
18:29:17 vralev: and 127.0.0.1:5065 (internal, private) for communication with the nodes
18:29:39 jeand: http://hudson.qa.jboss.com/hudson/view/Mobicents/job/MobicentsBooks/lastSuccessfulBuild/artifact/sip-balancer/index.html#sslb-binary-SIP_Load_Balancer-Configuring
18:29:39 tleruitte1: vralev: what is the one in the config file?
18:29:51 jeand: # The binding address of the load balancer where clients should connect (if the host property is not specified)
18:29:51 jeand: #externalHost=127.0.0.1
18:29:56 jeand: # The binding address of the load balancer where SIP application servers should connect (if the host property is not specified)
18:29:56 jeand: #internalHost=127.0.0.1
18:29:56 tleruitte1: jeand: you gave a VPN link again 
18:30:01 jeand: ah shit
18:30:09 jeand: pardon my french 
18:30:17 jeand: http://hudson.jboss.org/hudson/view/Mobicents/job/MobicentsBooks/lastSuccessfulBuild/artifact/sip-balancer/index.html#sslb-binary-SIP_Load_Balancer-Configuring
18:31:01 tleruitte1: well, my bad, i should have read that file more carefully 
18:31:30 jeand: so no problems we should work fine with elastic IP
18:32:00 vralev: but we should avoid elastic IP - too expensive, no?
18:32:06 tleruitte1: so there are 3 ip to configure in the LB, host, externalHost and internalHost
18:32:24 jeand: you can discard host if you use both external and internal
18:32:28 tleruitte1: vralev: actually i think elastic ip is free as long as there are attached to an instance, you pay only if you don't use it
18:33:28 tleruitte1: and if there aren't used, i may be something like 1 or 2$ per month, so affordable 
18:33:37 vralev: you pay if you dont use it? that doesnt make sense  are you sure?
18:34:17 tleruitte1: actually it makes sense  when you start an instance you have already one public ip for free
18:34:27 tleruitte1: it doesn't cost a lot to amazon to give more
18:34:32 tleruitte1: but it cost if you don't use it
18:34:48 tleruitte1: but i'm not 100% sure but i think so
18:34:50 jeand: tleruitte1, it might be interesting in your doc to actually detail the cost for one call
18:34:52 tleruitte1: i should check
18:34:55 vralev: but the traffic is not free isnt that right?
18:35:00 jeand: in this archietcure on amazon
18:35:05 tleruitte1: no, you still pay the traffic 
18:35:26 jeand: I wonder how much does it cost and compare
18:35:36 tleruitte1: jeand: indeed, but how can i compute that?
18:35:50 jeand: not sure 
18:36:03 vralev: ok, so by binding the LB internalHost to the private IP you will be paying 2x less traffic
alexandrem [~unk...@bl14-28-27.dsl.telepac.pt] entered the room. (18:37:01)
mode (+o alexandrem) by ChanServ (18:37:01)
18:37:06 tleruitte1: the amount of call processed is not dependent of the time the cluster is up
18:37:12 tleruitte1: vralev: yep
18:37:31 jeand: you pay both for instances up and traffic right ?
18:37:37 tleruitte1: jeand: yes
18:37:49 tleruitte1: fix cost and variable cost
18:37:57 tleruitte1: it remind me my economics class
18:37:59 jeand: is it per day/ week / month ?
18:38:02 jeand: lol
18:38:02 tleruitte1: maybe i can talk a bit about that 
18:38:07 tleruitte1: per hour
18:38:11 jeand: ok good
18:38:21 jeand: let's assume we have 1 call per second
18:38:33 jeand: you can script it through sipp
18:38:39 jeand: and target your cluster
18:38:49 jeand: and after one hour see how much it costs
18:38:56 jeand: or you avoid the paying
18:39:05 jeand: and see how traffic fee is computed
18:39:26 tleruitte1: that's a good idea 
18:39:26 jeand: and see if you can do the maths
18:40:09 tleruitte1: i could do it even theorically whitout testing
18:40:43 jeand: yeah I think so too
18:40:57 jeand: you can compute locally the size of the SIP traffic for one call
18:41:15 jeand: one question can you use UDP or TCP
18:41:26 jeand: or one the transports is mandatory like only TCP
18:41:36 tleruitte1: i could use both i guess
18:41:45 tleruitte1: i should check that
18:42:17 jeand: ok I was thinking of creating a git repo
18:42:23 tleruitte1: i'll be back to you later with the answer 
18:42:24 jeand: for mobicents
18:42:33 jeand: or mobicents-sip-servlets
18:42:46 jeand: so that you could push your project to it
18:43:02 jeand: so that if you leave or go to other things it would not be lost in the wild
18:43:03 tleruitte1: yeah that would be good, otherwise i can use svn as well ;
18:43:05 tleruitte1: 
18:43:06 jeand: on your git account
18:43:15 jeand: ah if you can use svn that would be even better
18:43:25 jeand: that would avoid syncing headaches
18:43:32 jeand: between our git repo
18:43:37 jeand: and svn repo
18:43:49 tleruitte1: not a problem for me 
18:43:55 tleruitte1: i would just need some credentials 
18:44:01 jeand: ok I will give you access to mobicents repo
18:44:12 jeand: write acces
18:44:15 shaym: me2 plz 
18:44:18 jeand: access so that you can commit your stuff
18:44:25 jeand: shaym, how much can you pay for that ?
18:44:27 jeand: :-
18:44:29 jeand: 
18:44:36 shaym: a bottle of scotch 
18:44:44 jeand: deal 
18:44:53 jeand: by the way is any of you going to JUDCon ?
18:45:05 jeand: http://www.jboss.org/events/JUDCon
18:45:09 jeand: in Boston
18:45:14 jeand: in case we can meet face to face
18:45:23 shaym: is it a good show?
18:45:27 tleruitte1: next point i'd like to talk about: rhq
18:45:41 tleruitte1: i'd like to but not sure if my promoter would pay me the trip 
18:45:55 jeand: tleruitte1,  don't hurt to ask 
18:46:12 jeand: say it is for strategic planning 
18:46:34 tleruitte1: indeed 
18:47:47 tleruitte1: can i go on rhq?
18:48:30 jeand: you both have been granted write access to the mobicents rep
18:48:32 jeand: repo
18:48:35 jeand: congrats !
18:48:45 tleruitte1: thank you 
18:48:47 shaym: tx 
18:48:50 tleruitte1: it's a day to remember 
18:48:59 shaym: indeed
18:49:02 jeand: ok go ahead on rhq
18:49:05 jeand: 
18:49:26 tleruitte1: well, not much to say, anyway i've started to take a look at it
18:49:39 tleruitte1: i've some trouble to connect the agent and the mobicent instance
18:50:20 tleruitte1: oh yeah, first i might have found a bug in http://hudson.jboss.org/hudson/view/Mobicents/job/MobicentsBooks/lastSuccessfulBuild/artifact/sip-servlets/index.html#emom-Enterprise-Monitoring-Operations-Management
18:50:39 jeand: configuring RHQ is a pain
18:51:01 jeand: what kind of trouble ?
18:51:03 jeand: and which bug ?
18:51:31 tleruitte1: when i test the jopr for development locally, everything is goof and i can access to the monitoring in localhost:8080
18:52:08 tleruitte1: but if mobicents is started with the -b option, i can't get the management page of mobicents anymore
18:52:26 tleruitte1: i mean the page doesn't show that part
18:52:36 tleruitte1: and i get some error in the console
18:52:55 tleruitte1: org.mc4j.ems.connection.EmsConnectException: Could not connect [jnp://127.0.0.1:1099] javax.naming.CommunicationException: Could not obtain connection to any of these urls: 127.0.0.1:1099 [Root exception is javax.naming.CommunicationException: Failed to connect to server /127.0.0.1:1099 [Root exception is javax.naming.ServiceUnavailableException: Failed to connect to server /127.0.0.1:1099 [Root exception is java.net.ConnectException: Connection refused]]]

18:53:03 jeand: ah I think you need to modify something
18:53:08 jeand: in the descriptor
18:54:01 tleruitte1: i'll keep that in mind, for now i've skip that part and i tried the jopr for production option
18:54:18 tleruitte1: there the trouble is that the agent doesn't know about the mobicents instance
18:54:30 tleruitte1: it doesn't automaticaly discover it
18:54:37 tleruitte1: i though i could add it manually
18:54:49 jeand: yes in deploy/admin-console.war/plugins/jopr-mobicents-sip-servlets-as-5-plugin-1.6.jar/rhq-plugin.xml
18:55:01 tleruitte1: but i discover that rhq has a new gui since 4.0, and i can't find that option in the new gui
18:55:13 tleruitte1: so i might use the 3.0 version to give it a try
18:55:20 jeand: ah you need to use 3.x version
18:55:42 jeand: RHQ4 is still in beta and we don't support it yet
18:55:50 jeand: and probably on't before AS7 migration
18:55:55 jeand: won't*
18:56:16 tleruitte1: that's was i though, so there i am i'll try rhq 3.0
18:56:35 tleruitte1: that's good because i'll be able to take the management appliance file from the cirras project 
18:56:43 jeand: cool 
18:56:59 jeand: ok feel free to ask any questions related to RHQ and mobicents on the google group or IRC
18:57:02 jeand: if you are stuck
18:57:09 jeand: I know this one is painful
18:57:14 tleruitte1: thanks 
18:58:11 tleruitte1: one last point maybe for today if we have time, what point should i consider when monitoring mobicents, in order to decide to launch or shutdown a instance?
18:58:35 tleruitte1: should i go only with hardware, such as cpu and memory use
18:58:46 tleruitte1: or should i consider number of sip requests processed
18:58:47 tleruitte1: ?
18:59:39 tleruitte1: (i don't know if i will get to this point until next week so it can wait if it's late )
18:59:58 jeand: you should consider number of active sip sessions or sip application sessions I think
19:00:19 jeand: it gives you an idea of the number of concurrent calls
19:00:40 jeand: and usually I think a given instance is able to handle as much concurrent calls
19:00:44 jeand: before failing
19:01:05 jeand: that is if you know the machine you're running on
19:01:17 jeand: and if you already did the computation
19:01:21 jeand: beforehand
19:01:34 jeand: if you don't know and runnning on shared or virtualized
19:01:40 tleruitte1: the good thing about cloud is that al the machine have basically the same config
19:01:48 jeand: you may want to take CPU and memory
19:02:06 jeand: I think either one of the mentionned could work
19:02:21 jeand: CPU, memory, active sip sessions, active sip application sessions
19:02:36 jeand: if one go over a threshold
19:02:38 jeand: start a new instance
19:03:03 tleruitte1: that seem like a good idea
19:03:18 tleruitte1: same think if one go under a threshold, shutdown an instance i guess
19:03:20 jeand: you might need a dedicated plugin to start a new amazon instance from RHQ
19:03:25 jeand: though not sure how it can be done
19:03:31 jeand: right
19:03:39 jeand: or if the 3 of them are below a threshold
19:04:05 jeand: like if 1 is at 15%
19:04:08 jeand: one at 50%
19:04:18 jeand: and one at 80%
19:04:48 jeand: shutting down the 15% one
19:05:09 jeand: might make incresing the one at 80% above a threshold that will make it start an instance
19:05:29 jeand: thus going into a deadly loop of starting a new instance and shutting down
19:05:42 tleruitte1: indeed that would be a problem
19:05:57 jeand: so instead if all are below 60 or 70%
19:06:11 jeand: then you can shutdown one instance
19:06:17 jeand: or have an algo compute that
19:06:29 jeand: to choose which one should be shutdown
19:06:37 jeand: based on the load of the overall system
19:07:31 tleruitte1: yes that seems good to me 
19:07:49 jeand: ok thanks for the update
19:08:02 jeand: you can now port your git repo to svn repo
19:08:10 jeand: let me check where we can host this
19:08:14 tleruitte1: thank you all 
19:08:16 jeand: in which folder
19:08:30 tleruitte1: actually, i've 3 git repo so far
19:08:30 jeand: I will create a folder and let you know
19:08:40 tleruitte1: but i'll be 3 subfolder in sn i guess
19:08:42 tleruitte1: svn*
19:08:47 narayant: just a little info before we close
19:08:51 narayant: whats rhq
19:09:09 jeand: I would think trunk/cloud/sip-servlets would do
19:09:09 narayant: and what r we trying to achive with mobicents in the cloud
19:09:20 jeand: and then stuff your 3 subfolder in it
19:09:25 jeand: hey narayant
19:09:26 narayant: just pointers to both would be fine
19:09:34 narayant: waiting to ask
19:09:36 narayant: 
19:09:36 jeand: rhq is the monitoring and management system
19:09:41 jeand: for JBoss
19:09:51 narayant: ok
19:09:53 jeand: http://rhq-project.org/display/RHQ/Home
19:10:05 jeand: we have a plugin that allows to monitor Mobicents
19:10:21 jeand: http://hudson.jboss.org/hudson/view/Mobicents/job/MobicentsBooks/lastSuccessfulBuild/artifact/sip-servlets/index.html#emom-Enterprise-Monitoring-Operations-Management
19:10:30 jeand: mobicents in the cloud
19:10:34 jeand: is an initiative
jeffprestes left the room (quit: Ping timeout: 276 seconds). (19:10:47)
19:10:54 jeand: to allow our container to be deployed easily in the cloud
19:11:01 jeand: either public or private
19:11:13 jeand: public : amazon EC2
19:11:16 jeand: rackspace etc
19:11:24 tleruitte1: narayant: this is my master's thesis matter, i'll give you my thesis for reading if you're interested 
19:11:28 jeand: private : openstack, eucalyptus, ...
19:11:32 narayant: sure
19:11:37 jeand: and allow Mobicents to autoscale
19:11:41 jeand: depending on the load
19:11:44 narayant: I saw some forks in jeand's git as well
19:11:58 jeand: ie adding/removing nodes as the load grow up or go down
19:12:15 narayant: ok , so mobicent runs inside jboss normally
19:12:25 jeand: that would still be the case
19:12:33 narayant: in the clound does it agin run inside jboss
19:12:51 jeand: yes but you have images pre loaded with everything
19:13:06 jeand: and ready to be launch
19:13:12 narayant: how different is this than a jboss/MSS cluster
19:13:18 jeand: to accomodate the increasing load
19:14:12 jeand: well for now it is allow an MSS cluster to be deployed easily on those cloud providers
19:14:15 tleruitte1: narayant: it's different on 2 things: image are available, so it's easy to deploy in the cloud, and monitoring: new instance can be created automatically
19:14:29 jeand: but it can autoscale
19:14:38 jeand: like tleruitte1 just mentionned
19:14:39 tleruitte1: well, point 2 is under development 
19:14:59 narayant: ok ondemand ,without we having to maintain any hardware servers
19:15:14 narayant: or soft instances for that matter
19:15:19 jeand: for MSS 2.X, we plan to change the underlying clustering infrastructure to make use of data grid solutions like infinispan
19:15:34 jeand: narayant, yes by example if on public cloud
19:15:46 jeand: but you can still do it on premises through a private cloud
19:16:26 jeand: http://www.google.com/url?q=http%3A%2F%2Fsilverlighthack.com%2Fpost%2F2011%2F02%2F27%2FIaaS-PaaS-and-SaaS-Terms-Explained-and-Defined.aspx&sa=D&sntz=1&usg=AFQjCNE4HUiBGGn4KMnaVaqxu-v2EL5YSw
19:16:28 narayant: so the main advantage is in the maintenance front
19:16:36 narayant: also costwise ?
19:17:21 jeand: costwise you need to do the computation yourself depending on the provider you choose
19:17:50 jeand: and it gives you more flexibility
19:18:03 narayant: ok
19:18:29 jeand: think google infrastructure applied to telco
19:18:35 jeand: by example
19:18:44 narayant: mm
19:18:51 narayant: makes sense now
19:19:14 jeand: ok let me update on #3
19:19:36 jeand: I'm preparing the release of MSS 2.0.0.ALPHA2 and CTF 1.0.0.ALPHA1
19:19:47 jeand: and it should be out this evening or tomorrow morning
19:19:52 jeand: that's it
19:19:57 jeand: thanks for joining guys
19:20:01 narayant: <tleruitte1> , the thesis u mentioned is are doing them for some doctorate /phd s
19:20:47 narayant: let me know if we can chat / talk on this offline
19:20:48 tleruitte1: narayant: it's for my mmaster
19:20:58 narayant: ok
19:21:02 shaym: one update from me unles its too late for you guys
19:21:25 tleruitte1: sure, here is my jabber/gmail address: tler...@gmail.com
19:21:41 narayant: ok thanks
19:22:14 shaym: jean , vralev you there?
19:22:38 jeand: #4 telemed
19:22:40 jeand: sorry shay
19:22:45 shaym: np 
19:22:53 shaym: ill make it quick
19:23:31 jeand: well we should try to make it quicker as the topic list grow up ( we will try to keep it in the 1h next week)
19:23:39 jeand: sorry about thayt
19:23:40 shaym: i started of using proxy mode, and customized the speed-dial app ,but it looks like jitsi and some of the other phones
19:24:11 shaym: will only accept msgs from the sip they have on file, and nobody else
19:24:27 shaym: so i switched to B2BUA  based on call-forwarding
19:25:01 shaym: and had lots of fun with binding the right ports .....  they  all seemed elastic to me 
19:25:14 shaym: now i got the voice work, but video has issues
19:25:24 shaym: if im reading the logs correctly
19:25:36 jeand: ok making good progress then 
19:25:54 shaym: jitsi does not launch the video automatically , when you click the button it sends an invite
19:26:00 jeand: weird doubt it has to do with Mobicents since it is involved only in the SIP traffic here
19:26:05 shaym: which in turn blows up the servlet
19:26:10 jeand: ah a reinvite
19:26:13 shaym: yes
19:26:18 jeand: maybe the B2BUA example
19:26:23 jeand: doesn't support it
jeffprestes [~jeffprest@nat/redhat/x-pmmkoediabkbniyu] entered the room. (19:26:25)
19:26:28 shaym: invite from the receiver
19:26:30 jeand: you can add support for that easily
19:26:39 jeand: just check if the request.isInitial(à
19:26:43 jeand: ()*
19:26:51 jeand: you do the existing ode
19:26:53 jeand: code*
19:26:56 shaym: if not?
19:27:13 jeand: otherwise you reuse the sipsession to create the subsequent request
19:27:22 shaym: a new session?
19:27:30 shaym: reuse, sorry
19:27:45 shaym: so just fire it as another request in the same session?
19:28:00 jeand: SipServletRequest forkedRequest = b2buaHelper.createRequest(origSession, request, headers);
19:28:25 jeand: you can take a look at CallForwardingB2BUASipServlet in sip servlet testsuite
19:28:29 jeand: it does handle the reinvite
19:28:33 shaym: cool , ill give it a whirl
19:28:41 shaym: cant wait to see video 
19:28:41 jeand: but the code is a bit messy because we test plenty of stuff
19:28:48 jeand: in there
19:28:54 jeand: me neither ;-
19:28:58 jeand: me neither 
19:29:07 shaym: thats it for me
19:29:14 jeand: ok thanks all for joining
19:29:28 jeand: I'll post the chat log
19:31:48 narayant: ok thanks & bye

[Ivelin Ivanov]

On #2, if RHQ is too much pain in the neck, an alternatives to
consider: http://www.makara.com/tour/
Its a new product under RHT, so I'm not sure if it will be easier at
this stage to use for cloud instance management.

Just noticed this new product from SpringSource. Does anyone know how
it compares to boxgrinder and Makara?
http://cloudfoundry.com/

The Amazon Beanstalk and Cloudbees services are interesting too. Is it
possible to deploy an upgrade to fully functional MSS on top of their
pre-configured Tomcat instances? Of course that doesn't help with
deployment images for private clouds. Probably a lower priority point
for investigation.
http://www.ibm.com/developerworks/library/j-paasshootout/

[Jean Deruelle]

I don't think PaaS are the right target to look into for us right now as we would need to wait for them to add support for Mobicents as they keep control of the middleware stack. What we are trying to provide so far is enablers (image, architecture, etc...) for PaaS or SaaS providers or private clouds to autoscale Mobicents.

See inline.

On Wed, Apr 20, 2011 at 12:30 AM, Ivelin Ivanov <ivelin.atan...@gmail.com> wrote:

On #2, if RHQ is too much pain in the neck, an alternatives to
consider: http://www.makara.com/tour/
Its a new product under RHT, so I'm not sure if it will be easier at
this stage to use for cloud instance management.

It seems interesting for HTTP/java EE applications

It has out of the box support for HTTP for autoscaling, httpd and mod cluster it seems but I'm not sure how to add support for SIP autoscaling and the SIP Load Balancer. The code source is still not available to tweak it and there is no mention as to when it will be available but that could be reused if it was the case to add support for Mobicents there. Also they seem to offer it as a service but not as a downloadable image that you could run in your private cloud.

Here is the list of supported software so far :

Apache, MySQL, PHP, Zend Framework, Java, Tomcat, JBoss, Memcached

and cloud providers :

Amazon EC2, Amazon VPC, vCloud, Terremark

 

Alternatively there was http://steamcannon.org/news/2011/01/19/steamcannon-vs-elastic-beanstalk/ but it has been killed by the makara acquisition I think. It was still in alpha but definitely interesting.

Just noticed this new product from SpringSource. Does anyone know how
it compares to boxgrinder and Makara?
http://cloudfoundry.com/

Boxgrinder is not a Paas.

It is a PaaS so it compares more to makara, cloudbees and amazon elasticbean than boxgrinder.

It seems to be free from the underlying cloud provider as compared to makara and can be used on existing infrastructure (desktop, datacenter and private clouds) 

It is not clear which app servers they support 

"At this time, Cloud Foundry supports Spring for Java applications, Rails and Sinatra for Ruby apps, Node.js apps and apps for other JVM frameworks including Grails" 

although they claim to be open source so we could potentially reuse this to add support for Mobicents... but at the difference of makara, the source code is available https://github.com/cloudfoundry so it might be interesting to integrate with it indeed. Something to look into.

The Amazon Beanstalk and Cloudbees services are interesting too. Is it
possible to deploy an upgrade to fully functional MSS on top of their
pre-configured Tomcat instances? Of course that doesn't help with
deployment images for private clouds. Probably a lower priority point
for investigation.
http://www.ibm.com/developerworks/library/j-paasshootout/

Amazon BeanStalk/CloudBees/Makara/GAE don't allow you to upgrade the tomcat instance, they keep the control of that... Only CloudFoundry seems open enough so far.

[Ivelin Ivanov]

Agree that PaaS integration is a distraction at the moment. 

Ivelin

[Ivelin Ivanov]

Just learned something interesting on the server management front. A

S7 now ships with a new management console in each Beta. The latest -
Beta 3 went out his morning:
http://planet.jboss.org/view/post.seam;jsessionid=D152746BF34269095EADFECE970F0534?post=jboss_7_web_console_1_0_0_beta3

Heiko Braun leads the project effort:
http://relative-order.blogspot.com/

It was recommended to me that future management plug-ins be developed
for the AS7 embedded management console instead of RHQ. Heiko can be
engaged for questions and guidance on the AS7 public forums. RHQ will
be refactored to connect to the AS7 management API. Production release
is targeted for the end of 2011.

Ivelin

[Jean Deruelle]

Yeah I saw that recently as well on http://relative-order.blogspot.com/2011/04/managing-jboss7-staging-applications.html

Ok so it fits the current strategy of using RHQ for MSS on JBoss AS 5.

We shall move forward with MSS 2.X on JBoss AS 7 now that we cleared the ALPHA2 release and will look into those as well as we gain more "cloud knowledge" on our current infrastructure.

On Wed, Apr 20, 2011 at 4:05 PM, Jean Deruelle <jean.d...@gmail.com> wrote:

Ok so it fits the current strategy of using RHQ for MSS on JBoss AS 5.

We shall move forward with MSS 2.X on JBoss AS 7 now that we cleared the ALPHA2 release and will look into those as well.