This section describes known issues in Sun GlassFish Enterprise Server v3 and workarounds if available.
[JDK_Issue] Performance degradation caused by invoking setSoLinger or setReuseAddress (Issue 7109)
[JDK_Issue] Null pointer exception on server restart (Issue 8299)
[JDK_Issue] IO exception: invalid argument during longevity test (Issue 7529)
[JDK_Issue] EPoll null pointer exception at startup (Issue 9472)
[JDK_ISSUE] Richaccess: java.io.IOException: Invalid argument from doSelect (Issue 8573)
File permissions on domain /applications directory can cause NullProcessException (Issue 6545)
Access to statistics for new virtual servers requires server restart (Issues 6238 and 6422)
[Open Installer] Option -l to relocate log files ignored on Windows (Issue 10693)
Issues occur with ZIP distribution if UAC enabled on Windows Vista (Issue 10755)
Null pointer exception thrown from com.sun.xml.wss.NonceManager.getInstance (Issue 11138)
Standalone Update Tool fails with segmentation fault on Solaris (Issue 11222)
Ruby applications deployed on context root don't work with Admin Console (Issue 10854)
Launching an app client can give ClassNotFound error for the client's main class (Issue 11181)
Change to log file location requires server restart to take effect (Issue 11142)
[Update Center] Non-user directory access fails (Update Center Issue 1583)
Inline help and CLI man page list incorrect servlet version 2.4 in X-Powered-By (Issue 11011)
[Monitoring] Extra monitoring view for connector-connection-pools not available (Issue 11256)
[EclipseLink] Issues with ElementCollections of embeddables (EclipseLink Issue 296606)
EJB interop for remote EJBs broken when target EJB is on the same host (Issue 11152)
Installer hangs at 41% intermittently on Windows Vista and Windows 7 (Issue 11185)
Embedded ACC overly strict on current thread context class loader (Issue 11427)
Expired certificate in Enterprise Server truststore (Issue 6852796)
When the setSoLinger method or the setReuseAddess method is invoked, performance is degraded and the following exception is thrown:
[#|2009-01-26T00:33:56.325-0800|WARNING|sun-appserver9.1| javax.enterprise.system.container.web|_ThreadID=17; _ThreadName=SelectorReaderThread-8084; _RequestID=11ae0030-c392-4217-8408-cfa7efe0a879;|setSoLinger exception java.net.SocketException: Invalid argument |
This issue is caused by an issue with the JDK software. This issue is resolved in JDK version 7.
For more information, see Issue report.
None.
Restarting Enterprise Server sometimes causes a null pointer exception to be thrown.
SEVERE: doSelect exception java.lang.NullPointerException |
This issue is caused by an issue with the JDK software. This issue is resolved in JDK version 7.
For more information, see Issue report.
None.
During an HTTP longevity test, the following exception is thrown 42 hours into the run:
[#|2009-04-05T17:41:26.537-0700|SEVERE|glassfish|javax.enterprise.system.core| _ThreadID=15;_ThreadName=Thread-1;|doSelect exception java.io.IOException: Invalid argument |
The instance and application are still accessible during the run.
This issue is caused by an issue with the JDK software. This issue is resolved in JDK version 7.
For more information, see Issue report.
None.
At startup, Enterprise Server throws a null pointer exception:
java.lang.NullPointerException at sun.nio.ch.Util.atBugLevel(Util.java:326) at sun.nio.ch.SelectorImpl.<init>(SelectorImpl.java:40) at sun.nio.ch.EPollSelectorImpl.<init>(EPollSelectorImpl.java:47) |
This relates to Grizzly and is a JDK 6 issue. This issue is resolved in JDK 7.
For more information, see Issue report.
None. Grizzly tries to work around this issue but still may fail.
The following exception is seen:
[#|2009-06-20T06:05:57.942-0700|SEVERE|glassfish| com.sun.grizzly.config.GrizzlyServiceListener| _ThreadID=21;_ThreadName=Thread-2;|doSelect IOException java.io.IOException: Invalid argument |
This is a JDK issue, scheduled to be fixed in JDK 1.6.0_18.
For more information, see Issue report.
Increase the maximum number of file descriptors to 10000 or greater (the default on Solaris is 64000). Once JDK 1.6.0_18 becomes available, install it.
If a domain's /applications directory restricts access, or if you use directory deployment from a restricted directory, the server cannot read the files in the expanded directory. A NullProcessException error occurs during deployment.
For more information, see Issue report.
Change the file access settings for such directories to grant the server permission to read the directory contents.
The time-stamp-install.log file cannot be read, because all lines written to the file are concatenated into a single long string.
For more information, see Issue report.
Use a different editor, and manually open the installation log file created under the %TEMP% directory.
Monitoring statistics about newly added virtual servers are available only after the server is restarted.
For more information, see the reports for Issue 6238 and Issue 6422.
After adding a virtual server, restart the server to view monitoring data for the virtual server.
Option -l to relocate log files is ignored when used with options -a and -s and the log files are created in the default location.
For more information, see Issue report.
None.
Some features will not work well on Windows Vista with User Account Control (UAC) enabled. One example is the Administration Console, which cannot be launched.
For more information, see Issue report.
Disable UAC and reboot.
Testing a JAX-RPC web service, with GlassFish message security provider enabled, throws the following exception in the server logs:
[#|2009-11-23T11:16:58.375+0005|SEVERE|glassfishv3.0| javax.enterprise.resource.webservices.rpc.server.http|_ThreadID=25;_ ThreadName=http-thread-pool-8080-(2);|caught throwable java.lang.RuntimeException: com.sun.enterprise.security.jauth.AuthException |
The default value of the nonce property does not work.
For more information, see Issue report.
To use message security with JAX-RPC web services, disable the nonce property in the configuration. See the Issue report for the complete steps for the workaround.
The Start menu group for Enterprise Server is not displayed after installation is first completed. If you log out and then log back in, the menu group is displayed but it is empty.
For more information, see Issue report.
None.
The command asadmin get -m "server.*" returns all monitoring data pertaining to that server. When many applications are deployed, the amount of data is quite large and could take a long time to return. The client might time out with the following client-side error:
./asadmin get --monitor=true "server.*" I/O Error: Read timed out Command get failed. |
For more information, see Issue report.
Minimize the amount of data that is returned to the client.
Run the asadmin list -m "server.*" command, which returns the top-level elements.
Select the top-level element for which details are needed, and use this as a filter. For example:
> ./asadmin list -m "server.*" ... server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-cache server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods. create-int-int-[Lorg\.spec\.jappserver\.supplier\.helper\.ComponentOrder server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods. findByPrimaryKey-java\.lang\.Integer server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods.generateXml server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods.getEJBLocalHome ... >./asadmin get -m "server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods.*" ... server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods.remove. methodstatistic-name = MethodStatistic server.applications.SPECjAppServer.supplier\.jar.POEnt.bean-methods.remove. methodstatistic-starttime = 1259604209775 ... |
The standalone Update Tool started with the updatetool command fails with a segmentation fault on Solaris when installing add-on components.
For more information, see Issue report.
Ensure that your system conforms to the standalone Update Tool patch requirements as defined in the Update Center Release Notes.
Update Tool functionality in the Administration Console uses a different Java-based Update Center API and is not affected by this issue.
If a Ruby application is deployed at contextroot '/' and the Administration Console is then accessed, accessing the Ruby application produces a 404 error.
For more information, see Issue report.
None.
When using Java Web Start to launch an application client, any managed beans in the application client will not be recognized.
For more information, see Issue report.
Launch the application client using the appclient script. Managed beans in the application client will be supported normally.
When you invoke the appclient script on Mac OS X systems with Java from Apple installed, the following stack trace is seen twice (only the first few lines are shown here):
Intentionally suppressing recursive invocation exception! java.lang.IllegalStateException: recursive invocation at java.lang.ClassLoader.initSystemClassLoader(ClassLoader.java:1394) at java.lang.ClassLoader.getSystemClassLoader(ClassLoader.java:1377) at sun.security.jca.ProviderConfig$1.run(ProviderConfig.java:64) ... |
For more information, see Issue report.
None needed.
Despite the warning messages, the client will be launched successfully and run normally. These errors are from an issue in the Apple Java implementation.
A problem occurs when all of the following are true:
The application is an EAR that contains an app client.
The EAR is directory deployed.
The EAR contains an application client module myAppClient.jar, which, because this is a directory deployment, is pre-expanded into myAppClient_jar. (For the purposes of this example, myApp is the name of the EAR. This name can be anything.)
Attempts to launch the application client fail with the following error because one file generated on the server is placed in the wrong server directory and overwrites another generated file:
java.lang.ClassNotFoundException:(main-class-for-the-client) |
For more information, see Issue report.
Change the name of the application client so it is not myAppClient.jar. Specifically:
Edit myApp/META-INF/application.xml so that the declaration of the client is <java>myClient.jar</java>. Note that you can use any name other than myAppClient.jar.
Rename the directory myApp/myAppClient_jar to myApp/myClient_jar. Note that the client's subdirectory name must be the same as the application client URI in the application.xml file, with the .jar replaced with _jar.
Deploy the application:
asadmin deploy --retrieve localdir myApp |
Run the application client:
appclient -client localdir/myAppClient.jar |
Changes to server log values on the General tab of the Logger Settings page in the Administration Console do not immediately take effect.
For more information, see Issue report.
Restart the server. All changes to the values on this page require a server restart to take effect.
Installation log files cannot be opened by clicking the links on the Summary page that displays at the end of the installation process in the graphical installer.
For more information, see Issue report.
Access the files manually. The names of the installation log and summary files are timestamp-install.log and timestamp-install-summary.html. On Linux and Mac systems these files are generated under the $TMP directory.
If you reinstall Enterprise Server (with Update Tool) in the same installation directory with the same defaults and invoke Update Tool using the updatetool command, you receive a message saying that Update Tool is not installed and are asked if you want to install it. This occurs on Windows systems only.
For more information, see Issue report.
Following uninstallation, manually remove the remaining .org* directory before reinstalling.
This issue occurs intermittently on Windows and Mac OS systems. pkg(5) does not work on certain systems at certain times.
For more information, see Update Center Issue report.
None.
The inline help and CLI man page list servlet 2.4 in the X-Powered-By field. The correct version is servlet 3.0.
For more information, see Issue report.
None.
When using embedded with glassfish-embedded-all-3.0-b73.jar, deployment fails with the following error:
SEVERE: WEB9051: Error trying to scan the classes at /private/var/folders/CV/CVhj8DvqEwGK5bdJKK9TaE TI/-Tmp- /gfembed6991712842235699248tmp/applications/xwiki-enterprise-web-2.0/ WEB-INF/lib/activation-1.1.jar for annotations in which a ServletContainerInitializer has expressed interest java.util.zip.ZipException: error in opening zip file |
The issue occurs because embedded uses a folder with plus signs (+), and plus signs in a path are converted into space characters " " when the path is decoded.
For more information, see Issue report.
A system property has been added that enables you to place temporary directories. The property is glassfish.embedded.tmpdir and can be set so the temporary domain directory is not placed in the user directory.
create-service fails to create service without AS_ADMIN_USER in passwordfile on Solaris.
For more information, see Issue report.
None.
Enterprise Server uses a tree structure to track monitorable objects. Within that tree, the following view is not available to obtain connection pool statistics:
server.connector-service.resource-adapter-name.connection-pool-name.* or server.jms-service.connection-factories.connection-factory-name.* (for jms-ra-related pools).
For more information, see Issue report.
Use the server.resources.* view to obtain the statistics.
For more information about Monitoring, see Chapter 8, Administering the Monitoring Service, in Sun GlassFish Enterprise Server v3 Administration Guide.
With weaving enabled, updates on an element collection of embeddables can potentially throw a null pointer exception.
For more information, see EclipseLink Issue report.
Two workarounds are available:
Add the annotation @ChangeTracking(DEFFERED) on the embeddable and set the property eclipselink.weaving.internal to false, or
Set the following properties to false in persistence XML: eclipselink.weaving.changetracking and eclipselink.weaving.internal.
After starting the domain and accessing localhost:4848, the following messages are seen in the server log:
[#|2009-11-27T16:21:57.091+1100|INFO|glassfishv3.0| javax.enterprise.system.container.web.com.sun.enterprise.web| _ThreadID=20;_ThreadName=Thread-1;|Created virtual server server|#] |
[#|2009-11-27T16:21:57.091+1100|INFO|glassfishv3.0| javax.enterprise.system.container.web.com.sun.enterprise.web| _ThreadID=20;_ThreadName=Thread-20;|Created virtual server server|#] |
These messages give the impression that the virtual server, server, was started twice. This is not the case. Virtual servers are only started once, but messages are logged multiple times.
For more information, see Issue report.
None.
Debugging JPA is difficult because of limited messages from the server.
For more information, see Issue report.
Add the property org.eclipse.persistence.session.level=INFO to the logging.properties file. You can then use the Administration Console to control EclipseLink loggers.
EJB interoperability for remote EJBs is broken when the target EJB is on the same host (another Enterprise Server domain or another Enterprise Server v3 instance).
For more information, see Issue report.
Set the following property in jvm-options:
-Dorg.glassfish.orb.iiop.orbserverid=:
The Enterprise Server graphical installer hangs at 41% during installation on Windows Vista and Windows 7. This does not happen with every installation attempt.
For more information, see Issue report.
This issue is related to the auto-tuning feature in Windows Vista and Windows 7, which is enabled by default.
If you experience hangs when installing Enterprise Server initially, or when adding packages or applying updates, restrict or disable the auto-tuning feature.
By default, the default host name for the JMS service on Enterprise Server is localhost. To access the JMS service from another system, however, you must change the host name. You can change it to either the actual host name or to 0.0.0.0.
For more information, see Issue report.
To make the change, do one of the following:
Use the Administration Console: Expand the Configuration, Java Message Service, and JMS Hosts nodes, select default_JMS_host, and edit the Host field, or
Use an asadmin subcommand such as the following:
asadmin set server-config.jms-service.jms-host.default_JMS_host.host="0.0.0.0", or
asadmin set server-config.jms-service.jms-host.default_JMS_host.host="hostname"
When Enterprise Server is installed using the graphical installer, the installation completes successfully but only the top-level GlassFish v3 entry is added in the Windows system menu, and it is empty. This issue occurs with both the localized and English installers.
For more information, see Issue report.
None.
mysql is not listed in the list of supported databases in --dbvendorname in the deploy command man page. This is incorrect. MySQL is a supported database and should be listed.
For more information, see Issue report.
None.
The ACC expects the current thread's context class loader to be an ACCClassLoader. This is overly restrictive. Although this condition is met for appclient script and Java Web Start launches, it might not be met for the embedded case. Other functions inside the ACC require the class loader to be a URLClassLoader (or an instance of a subclass of URLClassLoader), but the loader does not need to be an ACCClassLoader.
For more information, see Issue report.
Two workarounds are available:
Set -Djava.system.class.loader=org.glassfish.appclient.client.acc.ACCClassLoader, or
In your Java program, instantiate an ACCClassLoader and set it to be the current thread's context class loader using Thread.currentThread().setContextClassLoader before using the embedded ACC classes and interfaces.
Able to create and store EJB timer in MySQL as user, but when trying to configure EJB Timer Service for MySQL, get the following exception (SQLException executing statement):
"CREATE TABLE EJB__TIMER__TBL (TIMERID VARCHAR(255) NOT NULL, BLOB BLOB(64000), INITIALEXPIRATIONRAW BIGINT, SCHEDULE VARCHAR(255), INTERVALDURATION BIGINT, OWNERID VARCHAR(255), STATE INTEGER, LASTEXPIRATIONRAW BIGINT, PKHASHCODE INTEGER, CREATIONTIMERAW BIGINT, CONTAINERID BIGINT, PRIMARY KEY (TIMERID))": com.mysql.jdbc.exceptions.jdbc4.MySQLSyntaxErrorException: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'BLOB BLOB(64000), INITIALEXPIRATIONRAW BIGINT, SCHEDULE VARCHAR(255), INTERVALDU' at line 1|#] |
For more information, see Issue report.
Make sure that the datasource-classname specified for the timer pool represents an XA DataSource.
MySQL-specific steps:
Unpack glassfishv3/glassfish/lib/install/applications/ejb-timer-service-app.war.
Add the following to WEB-INF/classes/__ejb_timer_mappings.xml after <persistence-unit-metadata>:
<persistence-unit-defaults> <delimited-identifiers/> </persistence-unit-defaults>
Repackage the WEB-INF/classes/__ejb_timer_mappings.xml file.
(This step required only if the current installation has already used the EJB Timer Service.) Manually create the EJB__TIMER__TBL in the database using this create statement:
CREATE TABLE `EJB__TIMER__TBL` (`TIMERID` VARCHAR(255) NOT NULL, `BLOB` BLOB(64000), `INITIALEXPIRATIONRAW` BIGINT, `SCHEDULE` VARCHAR(255), `INTERVALDURATION` BIGINT, `OWNERID` VARCHAR(255), `STATE` INTEGER, `LASTEXPIRATIONRAW` BIGINT, `PKHASHCODE` INTEGER, `CREATIONTIMERAW` BIGINT, `CONTAINERID` BIGINT, PRIMARY KEY (`TIMERID`))
The deploy subcommand fails against a secure server unless the --secure option is used.
For more information, see Issue report.
Use the --secure option when issuing the deploy subcommand against a secure server.
One of the authority certificates in the Enterprise Server truststore expired on January 7, 2010. The certificate is cacerts.jks. An error message is generated on startup indicating that the certificate has expired:
Version: V1 Subject: OU=Secure Server Certification Authority, O="RSA Data Security, Inc.", C=US Signature Algorithm: MD2withRSA, OID = 1.2.840.113549.1.1.2 Key: SunPKCS11-Solaris RSA public key, 1000 bits (id 17891456, session object) modulus: public exponent: Validity: [From: Tue Nov 08 19:00:00 GMT-05:00 1994, To: Thu Jan 07 18:59:59 GMT-05:00 2010] Issuer: OU=Secure Server Certification Authority, O="RSA Data Security, Inc.", C=US SerialNumber: [ 02ad667e 4e45fe5e 576f3c98 195eddc0] |
For more information, see Issue report.
The expired authority certificate was removed in update 18 of Java SE 6. It will also be removed from the Enterprise Server truststore in a future update.
You can ignore the error messages and use the update, or you can eliminate the error messages. To stop receiving the error messages, use keytool to remove the certificate from the Enterprise Server truststore:
=> cd domains/domainX/config => cp cacerts.jks cacerts.jks.save => keytool -delete -keystore cacerts.jks -alias verisignserverca Enter keystore password: |
To prevent the expired certificate from reappearing in subsequently created domains, the certificate should also be removed from the template truststore:
=> cd glassfish/lib/templates => cp cacerts.jks cacerts.jks.save => keytool -delete -keystore cacerts.jks -alias verisignserverca Enter keystore password: |
For more information about the keystore password, see the information about master passwords and keystores in Authentication in Sun GlassFish Enterprise Server v3 Administration Guide.