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

Re: FW: [xmlblaster-devel] Master/Slave Master Restart

Hi again,

i have now checked your setup and have fixed following:

o 14909# Cluster fixes
  - Setting retries=-1 as a default for remote cluster login
  - Default is now to NOT disconnect at the remote clusters during shutdown
    The old behavior can be forced by adding a <diconnect/> tag.
  - Bugfix to not forward the '__persistenceId' marker client property
from subscribeQos to the remote node

Please assure that your xmlBlasterPlugins.xml contains

  <!--========== SubscriptionPersistencePlugin =============-->
  <plugin create='true' id='subPersistence'
     <action do='LOAD' onStartupRunlevel='4' sequence='10' />
     <action do='STOP' onShutdownRunlevel='3' sequence='60'/>

and that your clients use persistent subscribes.

The only remaining problem is that i haven't the time to add an
automated test case
for your use case. If you find some spare time it would be nice if you
add one
under xmlBlaster/testsuite/src/java/org/xmlBlaster/test/cluster, there
are some
examples showing how this could be done. The big benefit is that like
this our future development
is assured to not harm your operational scenario (to run the cluster tests
you currently have to force CORBA with protocol=IOR).

To get the latest please use our subversion access.

best regards

PS: Here are the examples i've used for manual testing

You HAVE TO BE in the directory xmlBlaster/demo/javaclients/cluster to
run them successfully.
You can kill any or all of the participants and will never loose a
message or subscription.

java -Dcom.sun.management.jmxremote org.xmlBlaster.Main -propertyFile

java -Dcom.sun.management.jmxremote org.xmlBlaster.Main -propertyFile

java -Dcom.sun.management.jmxremote javaclients.HelloWorldPublish
-protocol IOR -bootstrapPort 7600 -numPublish 100 -interactivePublish
true -session.name pubisher/2 -persistent true -erase.forceDestroy true
-protocol IOR -oid PublishToSlave.RUGBY_NEWS -domain RUGBY_NEWS -content
"We win"

java -Dcom.sun.management.jmxremote javaclients.HelloWorldSubscribe
-protocol IOR -bootstrapPort 7601 -oid PublishToSlave.RUGBY_NEWS
-session.name subscriber/1 -domain RUGBY_NEWS -persistentSubscribe true
-dispatch/callback/retries -1 -multiSubscribe false

Sniff all messages passing through:

java javaclients.simplereader.SimpleReaderGui -protocol IOR
-bootstrapPort 7600

java javaclients.simplereader.SimpleReaderGui -protocol IOR
-bootstrapPort 7601


Jonathan Clark wrote:

This is pretty important as we have an installation in the field that we
are trying to fix. I realize that you have a lot going on, so if there
is any thing that Matt and I can do to help gather more info or somewhere
we should start looking, please let me know.



-----Original Message-----
From: owner-xmlblaster-devel at server.xmlBlaster.org [mailto:owner-xmlblaster-devel at server.xmlBlaster.org] On Behalf Of Marcel Ruff
Sent: Wednesday, March 08, 2006 8:01 AM
To: xmlblaster-devel at server.xmlBlaster.org
Subject: Re: [xmlblaster-devel] Master/Slave Master Restart


thanks for the details, i'm currently very busy but i will certainly
try to resolve the issue the next days,


PS: You can contact me directly if you have a very urgent time schedule ...

Jonathan Clark wrote:
As a follow-up to my previous message, I am testing by this starting a topic
and then breaking my connection between the master and the slave while the client is still connected to the slave but subscribed to the master domain.
If I make changes to the topic on the master while the connection is broken when I reconnect I don't get the updates that were processed by the master while the connection was broken.

-----Original Message-----
From: owner-xmlblaster-devel at server.xmlBlaster.org [mailto:owner-xmlblaster-devel at server.xmlBlaster.org] On Behalf Of Marcel Ruff
Sent: Tuesday, February 28, 2006 7:40 PM
To: xmlblaster-devel at server.xmlBlaster.org
Subject: Re: [xmlblaster-devel] Master/Slave Master Restart

Jonathan Clark wrote:

One other thing that I am noticing. I have a slave connected to the master for a domain and a client connected to the slave subscribing to the masterâs domain.

If the master gets restarted, my client does not get any more messages on the master domain until I restart the slave. The slave reconnects to the master

before the restart with the following message: âXmlBlaster.PingTimer ClusterNode-/node/ecc/] Connected to ecc, no backup messages to flushâ ; however, I

never get any more of the master domain messages. When I restart the slave, master messages appear that were generated well before the slave restart.

Is there a configuration item that would govern the reconnect? Any other thoughts?

Is this a persistent subscribe?

Jonathan Clark

Open Roads Consulting, Inc.