Hello,
We use a Jena TDB-backed dataset (release 3.1.0) accessed through a single
JVM multi-threaded application running generally on a Microsoft Windows
server.
The read and write accesses are made using transaction and read/write locks.
Here is the SPARQL update code we use:
dataset.begin(ReadWrite.WRITE);
> model.enterCriticalSection(Lock.WRITE);
> try {
> UpdateAction.parseExecute(sparql, model);
> if(writeMode) {
> dataset.commit();
> }
> } finally {
> model.leaveCriticalSection();
> model.close();
> dataset.end();
> }
Note that our SPARQL query code implements also read transactions and locks.
When multiple updates are made on our TDB-backed dataset, a
java.util.ConcurrentModificationException is sometimes thrown, leaving the
dataset in an inaccessible state.
Here is the stacktrace part:
> ...
> Caused by: java.util.ConcurrentModificationException: Reader = 1, Writer =
> 1
> at
> org.apache.jena.tdb.sys.DatasetControlMRSW.policyError(DatasetControlMRSW.java:157)
> at
> org.apache.jena.tdb.sys.DatasetControlMRSW.policyError(DatasetControlMRSW.java:152)
> at
> org.apache.jena.tdb.sys.DatasetControlMRSW.checkConcurrency(DatasetControlMRSW.java:79)
> at
> org.apache.jena.tdb.sys.DatasetControlMRSW.startUpdate(DatasetControlMRSW.java:60)
> at
> org.apache.jena.tdb.store.nodetupletable.NodeTupleTableConcrete.startWrite(NodeTupleTableConcrete.java:65)
> at
> org.apache.jena.tdb.store.nodetupletable.NodeTupleTableConcrete.sync(NodeTupleTableConcrete.java:249)
How can we avoid such a situation ?
Can we safely do without read/write locks when we use transactions in a
single JVM multi-threaded application ?
Thank you in advance for your help.
Sincerely,
Laurent
