I'll ask Irina/Yuliya to review doc for cluster. Do you think we still need to mention cluster is in beta in release notes?
On Wed, Mar 13, 2013 at 9:25 AM, seba.wag...@gmail.com < seba.wag...@gmail.com> wrote: > Hi Maxim, > > your use case does not really make sense. > You should not login to different servers initially. > > You should login to the same server initially. Then also the overallChat > will work as epxected. > > The server will redirect you for the conference room to the appropriate > server automatically. But it is not intend to have 10 servers and you use > randomly one for your initial request. > > We might extend that to this later, however currently you have to use the > same server initially. > > I think I have written down that in my docs for Denis but it somehow did > find its way to > http://openmeetings.apache.org/Clustering.html > I think Denis had initially the same issue. > > Your issue is still valid but in another context: > Once the user is redirected to a conference room, he potentially could > receive a currently in use streamId. > I am not sure if that can become a problem when you are in the conference > room but it will require further thinking, review and testing. > > Sebastian > > > > 2013/3/12 Maxim Solodovnik <solomax...@gmail.com> > >> Hello Sebastian, >> >> while testing cluster I found an issue: >> >> Steps: >> 1) configure 2 servers to work in cluster mode ( >> http://openmeetings.apache.org/Clustering.html) >> 2) start both servers >> 3) login to each server with different users >> Result: one of the servers has following Exception in the logs: >> >> DEBUG 03-12 11:12:40.162 ScopeApplicationAdapter.java 295262 463 >> org.apache.openmeetings.remote.red5.ScopeApplicationAdapter >> [NioProcessor-2] - ----------- getPublicSID >> ERROR 03-12 11:12:40.214 SessionManager.java 295314 129 >> org.apache.openmeetings.session.SessionManager [NioProcessor-2] - >> [getClientByStreamId] >> java.lang.RuntimeException: more then one client was found streamId 0 >> server Server [id=1, name=master-105, address=192.168.72.105, port=5080, >> user=admin, pass=12345, protocol=http, active=true, webapp=openmeetings, >> deleted=false] >> at >> org.apache.openmeetings.data.conference.dao.ClientDao.getClientByServerAndStreamId(ClientDao.java:115) >> ~[openmeetings-2.1.0-SNAPSHOT.jar:na] >> >> I guess this is because streamid is not unique anymore (both client just >> entered to the different red5 instances have streamId == 0) >> From the user perspective rtmp connection get dropped and client switched >> to RTMPT (Try 1) >> >> >> Room chat works as expected (the fact messages are not stored in the DB >> is not so important until everybody left the room, after that room can be >> opened on another server by auto-balancer) >> >> According to the Administration->Connections tab connection dropped >> because of issue above still listed as active ..... >> >> As you have mentioned before global chat is not working in this case. But >> as you also have mentioned this limitation can be bypassed if "master" >> server will be used as entry point by OM users) >> >> I believe current cluster implementation has same limitations as previous >> version, but better implemented. >> >> -- >> WBR >> Maxim aka solomax >> > > > > -- > Sebastian Wagner > https://twitter.com/#!/dead_lock > http://www.webbase-design.de > http://www.wagner-sebastian.com > seba.wag...@gmail.com > -- WBR Maxim aka solomax
