master indexer not receiving slave requests

mmogley · **Posted:** Thu Sep 30, 2010 9:43 pm

I'm trying to set up a master/slave configuration using JGroups. I've turned on full tracing for both master and slave. On the slave side, when updating an indexed entity, I get:

Code:

TRACE [http-9080-2] (JGroupsBackendQueueProcessor.java:63) run -  Preparing 2 Lucene works to be sent to master node.
TRACE [http-9080-2] (JGroupsBackendQueueProcessor.java:73) run -  Filtering: optimized Lucene works are not going to be sent to master node. There is 2 Lucene works after filtering.
TRACE [http-9080-2] (JGroupsBackendQueueProcessor.java:92) run -  Lucene works have been sent from slave IWL65431O-17365 to master node.
TRACE [http-9080-2] (PostTransactionWorkQueueSynchronization.java:83) beforeCompletion -  Processing Transaction's beforeCompletion() phase: org.hibernate.search.backend.impl.PostTransactionWorkQueueSynchronization@1fa2bd1
TRACE [http-9080-2] (PostTransactionWorkQueueSynchronization.java:78) beforeCompletion -  Transaction's beforeCompletion() phase already been processed, ignoring: org.hibernate.search.backend.impl.PostTransactionWorkQueueSynchronization@1fa2bd1
TRACE [http-9080-2] (PostTransactionWorkQueueSynchronization.java:94) afterCompletion -  Processing Transaction's afterCompletion() phase for org.hibernate.search.backend.impl.PostTransactionWorkQueueSynchronization@1fa2bd1. Performing work.
TRACE [http-9080-2] (BatchedQueueingProcessor.java:287) performWorks -  Lucene WorkQueue to send to backend: 
   DeleteLuceneWork: org.csmc.mds.model.Physician#3990
   AddLuceneWork: org.csmc.mds.model.Physician#3990

The master does display that it's joined the same jgroups cluster on startup. But after that, it just sits there and displays no trace about any incoming slave requests.

Master configuration:

Code:

hibernate.search.default.optimizer.operation_limit.max=1000
hibernate.search.default.optimizer.transaction_limit.max=100
hibernate.search.default.indexBase=/index/master
hibernate.search.default.refresh=300
hibernate.search.default.sourceBase=/index/master-copy
hibernate.search.default.directory_provider=org.hibernate.search.store.FSMasterDirectoryProvider
hibernate.search.worker.backend=jgroupsMaster

Slave configuration:

Code:

hibernate.search.default.optimizer.operation_limit.max=1000
hibernate.search.default.optimizer.transaction_limit.max=100
hibernate.search.default.indexBase=/index/slave
hibernate.search.default.refresh=300
hibernate.search.default.sourceBase=/index/master-copy
hibernate.search.default.directory_provider=org.hibernate.search.store.FSSlaveDirectoryProvider
hibernate.search.worker.backend=jgroupsSlave

I'm using Hibernate Search 3.2.1, Hibernate 3.5.6, and JGroups 2.9.0.GA.

Should I be using some other version of JGroups?

Thanks for any help.

Michael

hardy.ferentschik · **Posted:** Fri Oct 01, 2010 5:53 am

Hi,

we are testing against JGroups 2.8.0.GA, but I am not aware of a change which would make it not work for 2.9.0.GA. How does the master log look like? Do you have any additional JGroups specific configuration?
You could also try to use a packet sniffer to see what goes over the wire.

--Hardy

hardy.ferentschik · **Posted:** Fri Oct 01, 2010 6:17 am

FYI, I just upgraded our build to 2.9.0.GA of JGroups and our JGroup setup test are passing.
To isolate the problem I suggest you see whether JGroups works within your setup. You can for example try to run twice: java -cp jgroups-2.9.0.GA.jar org.jgroups.demos.Draw
You should get two widgets where you can draw in one while the other one updates. If this is working it is a good sign that your JGroups setup is working.
On the other hand, does your application work in a non clustered environment using eg a default FSDirectoryProvider?

--Hardy

mmogley · **Posted:** Fri Oct 01, 2010 5:03 pm

Hardy, thanks for your reply. I ran the jgroups demo and it was not successful. That is, the two instances are not seeing each other. So I guess this means I have to enable multicasting on Windows XP. You wouldn't happen to know how/where to do that off-hand? A cursory look through Google didn't turn up anything useful.

sanne.grinovero · **Posted:** Wed Oct 13, 2010 6:54 am

hi mmogley,
windows should have multicast enabled.

The jgroups demo is known to work well, so I assume you're trying to have multiple nodes connected (different computers).
When playing with the jgroups demo, this was made to show quick&easy clustering of different applications on the same host.
The start scripts are defining the interface to bind to as 127.0.0.1, so you should change that to the network interface you want to use and also make sure the firewall and security settings of windows are not going to block packets.
Even so, I'd recommend directly connected nodes to keep it simple: if you have "clever routers" in the middle, you might have to change more options and I can't help :/

For more help on jgroups, please refer to the mailing lists

mmogley · **Posted:** Wed Oct 13, 2010 1:07 pm

Thanks, Sanne. I was actually trying to run both Jgroups instances off the same localhost and they still did not see each other. Rather than debugging this, I opted for the JMS approach, since I was under time pressure.