[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xmlblaster] connection problems



Michael Atighetchi wrote:

Marcel,

thanks for the help - I finally got connections via XmlBlasterAccess
working this morning. What I believe turned out to be the problem is
not related to the XmlBlasterConnection / Access issue at all. I'm
starting both a master and slave node on a single box, and had the
master node only start IOR listeners on a port different from the
slave (which was starting all listeners). This worked with a previous
version of xmlblaster by doing something similar to

 ClientProtocolPlugin[IOR][1.0]=org.xmlBlaster.client.protocol.corba.CorbaConnection
 ClientProtocolPlugin[SOCKET][1.0]=
 ClientProtocolPlugin[RMI][1.0]=
 ClientProtocolPlugin[XML-RPC][1.0]=

 ClientCbServerProtocolPlugin[IOR][1.0]=org.xmlBlaster.client.protocol.corba.CorbaCallbackServer
 ClientCbServerProtocolPlugin[SOCKET][1.0]=
 ClientCbServerProtocolPlugin[RMI][1.0]=
 ClientCbServerProtocolPlugin[XML-RPC][1.0]=

 CbProtocolPlugin[IOR][1.0]=org.xmlBlaster.protocol.corba.CallbackCorbaDriver
 CbProtocolPlugin[SOCKET][1.0]=
 CbProtocolPlugin[RMI][1.0]=
 CbProtocolPlugin[XML-RPC][1.0]=
 CbProtocolPlugin[JDBC][1.0]=org.xmlBlaster.protocol.jdbc.CallbackJdbcDriver

Now, it seems as if even though the master has empty drivers, it still
tries to start socket and xmlrpc listeners, which were colliding with
the slave. I've worked around this by adding

 socket.port=9090
 xmlrpc.port=8090

however, would prefer to turn of these xport protocols on the
master. How is that done ?

Salute Michael,

this settings are outdated, they are now handled by Micheles new Plugin manager, see

http://www.xmlblaster.org/xmlBlaster/doc/requirements/engine.runlevel.html

You need to change xmlBlasterPlugins.xml now.


Also, I noticed the following warnings when starting up my cluster:

djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:02 PM WARN  RequestBroker-/node/djmproxy1] Persistent and recoverable topics are switched of with '-useTopicStore false', topics are handled RAM based only.

You need to configure Oracle or Postgres to have persistency.

djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:05 PM WARN  RmiDriver-/node/djmproxy1] Removed another entry while binding authentication RMI server to registry with name 'rmi://xxx.bbn.com:1099/I_AuthServer'
djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:05 PM WARN  RmiDriver-/node/djmproxy1] Removed another entry while binding xmlBlaster RMI server to registry with name 'rmi://xxx.bbn.com:1099/I_XmlBlaster'

The RMI name service (registry) currently only supports one xmlBlaster cluster node per registry.
It needs to be extended similar to the CORBA approach for cluster usage, see:


http://www.xmlblaster.org/xmlBlaster/doc/requirements/protocol.corba.NameService.html

I have currently no plans to extend the RMI plugin, if you need this you would
need to code it yourself, it shouldn't be a big issue, see


 xmlBlaster/src/java/org/xmlBlaster/protocol/rmi/RmiDriver.java

djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:08 PM WARN  JdbcDriver-/node/djmproxy1] No JDBC driver in xmlBlaster.properties given, set 'JdbcDriver.drivers' to point to your DB drivers if wanted, e.g. JdbcDriver.drivers=oracle.jdbc.driver.OracleDriveri:org.gjt.mm.mysql.Driver:postgresql.Driver

If you need JDBC add the drivers here, else ignore the warning (probably
we should introduce a new log level called 'HINT').

djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:08 PM WARN  HtPasswd] Security risk, no access control: The passwd file is switched off with 'Security.Server.Plugin.htpasswd.secretfile=NONE'

Install for example OpenLdap and configure to use the LDAP plugin for authentication.

djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:09 PM WARN  RequestBroker-/node/djmproxy1] Cluster manager is not ready, handling message '__sys__UserList' locally
djmproxy_xxx.xxx.xxx.xxx.log:[Apr 28, 2003 1:04:09 PM WARN  RequestBroker-/node/djmproxy1] Cluster manager is not ready, handling message '__sys__Login' locally

This is an open issue, we know about it but we have not
investigated a solution yet. Probably it has no impact in your
setup.
This issue is probably solvable with Micheles runlevel manager
in  future.

I'm particularly worried about "Cluster manager is not ready". Also,
how do you turn of the JDBC warning in the slave djmproxy ?

Just add a dummy driver like:

 JdbcDriver.drivers=sun.jdbc.odbc.JdbcOdbcDriver

this is included in your JDK.

regards

Marcel