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

Re: [xmlblaster-devel] Slave Database Persistence



Hi again,

1. Your fix is now commited to svn, thanks for this detailed report.
   The slaves should now receive proper erase events and clean up
   its resources.

2. The missing forceDestroy at the slave:
  I have fixed this and plan to release it next week after some
 more testing.

regards
Marcel




Jonathan Clark wrote:
Marcel,

Yes, I did test with just the !forceDestroy removed and as far
as I can tell, it worked as planned.  One other thing that
I noticed is that the erases that are sent to the slaves always
Have the forceDestroy set to false.

Jonathan Clark wrote:
Marcel,

One other thing to note on this problem.  We are erasing with forceDestroy=true which is the functionality we need for our
project.  If I set forceDestroy=false, then the erase
notifications make it to the slave after toSoftErased is called
from fireMessageEraseEvent in TopicHandler.

fireMessageEraseEvent only calls toDead if forceDestroy=true, but the following lines appear in toDead such that the notifyList will
never be populated and thus the erase notifications are never
sent.


if (!forceDestroy && !isSoftErased()) {
notifyList = collectNotifySubscribersAboutErase(sessionName);
Ok, thanks for this analysis, i assume you have tested to remove the condition and it worked fine.
I'll run the testsuite with the change and do some code inspection - and if all looks well
i'll commit your fix.


regards
Marcel
}

Jonathan


Matthew Sargent wrote:

Marcel,

More information on this, we were publishing with a the domain of the master
specified, but when we erased the keys we were not specifying the domain. We
have changed the erase to specify the domain but it does not seem to impact
the cleanup on the slave at all.

Matt (Jonathan's co-worker)

----- Original Message ----- From: "Marcel Ruff" <mr at marcelruff.info>
To: <xmlblaster-devel at server.xmlBlaster.org>
Sent: Saturday, February 11, 2006 2:15 PM
Subject: Re: [xmlblaster-devel] Slave Database Persistence



Jonathan Clark wrote:
I am running a master/slave configuration where the messages are
published to the master. Several clients are connected to the slave and

subscribe to the domain of the master. The clients get the messages
with out any problem and delete the messages when the keys are

erased. The slave properties file is configured as follows:

StoragePlugin[JDBC][1.0]=${JdbcStorage[postgres]}

StoragePlugin[RAM][1.0]=org.xmlBlaster.engine.msgstore.ram.MapPlugin


StoragePlugin[CACHE][1.0]=org.xmlBlaster.engine.msgstore.cache.PersistenceCa
chePlugin,persistentQueue=JDBC,transientQueue=RAM
# Choose the plugins

persistence/topicStore/defaultPlugin=CACHE,1.0

persistence/msgUnitStore/defaultPlugin=CACHE,1.0

The problem that I am having is that the number of rows in the
xb_entries table continues to grow on the slave. It stays at ~800 rows on

the master, but has grown to over 10,000 on the slave. Any idea why
the entries wouldnât be deleted from the slave database?

I will check this,

thanks for reporting
Marcel
Jonathan Clark

Open Roads Consulting, Inc.

757-546-3401