Community
Participate
Working Groups
Build Identifier: Enterprise Edition IWAT4039E The Test Execution Harness reported the following problems: org.eclipse.hyades.internal.execution.local.control.InactiveProcessException. I am running eclipse helios (3.6) using tptp 4.7.2 plugin. I setup a junit test and connected it to a remote deployment running on Windows 2008 server (VMWare Virtual Server). Normally the test runs fine however the server is under some stress (disk io) and I get this message. Also happens when I am testing with a large amount of jars (160mb). I noticed the ramp up period before it begins testing is several minutes. This seems extremely slow when compared to running the same local test using IAC on the Windows 7 desktop running eclipse. Is there anything I can do to improve this? Also I notice the tptp junit tests run alot faster on a physical machine rather than virtual. Please see screenshot for error message. I am running 32bit jvm (1.6.0_11) and 32bit agent controller on win2k8 server. The code I am invoking is invoking API's on a third party Oracle product. Reproducible: Always Steps to Reproduce: 1. Setup deployment on Windows 7 Enterprise in eclipse. Point it to Win 2k8 server running 32bit jvm jdk 1.6.0_11 and 32bit edition of Agent controller. 2. Put win 2k8 server under stress. 3. Run junit test by setting up a run configuration in eclipse.
Created attachment 202370 [details] error msg that shows in eclipse
Heiner, can you please have a look at this?
The exception usually just tells that there has been a timeout. I guess the timeout occurs during deployment. How much percent does the launcher show, when you get the exception? Paul Slauenwhite did a lot of experiments regarding a veriy similar issue some time ago - the problem has never been pinned down though: https://bugs.eclipse.org/bugs/show_bug.cgi?id=299942 You could try another JVM (IBMs, or a jrockit or a more current Sun^H^H^HOracle JVM). Paul mentioned the IBM jvm was a lot faster in a TPTP remote test setup. Raising timeout and remote JVM heap memory seem promising as well, especialy with large classpaths. Cannot tell exactly how and where off the top of my head. Increasing initial heap memory might even make things worse, if it is not needed (especialy in a VM env). Regarding physical vs virtual machine: To know for sure if it is the VM or the fact that the agent controller runs remotely, one should try to run on a local agent controller in a VM and/or on a remote agent controller on a physical machine.
I have tried running 64bit jvm and 64bit ac (1.6.0_11 jvm). Still nothing. I have tried copying eclipse into the VM environment and tried testing locally.. still slow and times out if server under load.
does the log file on agent controller side (<agentCtrlInstDir>/config/servicelog.log show anything helpful (common base event log file)? If not, you can try to change the log level to DEBUG in that line: <Logging directory="%TPTP_AC_CONFIG_HOME%" format="CBE" level="INFORMATION"/> I still had no time to find a convenient way to raise the timeout of the launcher. I saw that it is done in automated TPTP tests by extending org.eclipse.hyades.execution.harness.TestExecutionHarnessExecutorStub and overriding the method protected int getLaunchTimeout(); to return e.g. 6 Minutes (360000 milliseconds). The help topic "Launching tests from scripts and applications" under "Testing Applications" will surely cover the missing parts.
