Investigate if we can run on normal non 6gb image

[cdietrich]

Investigate if we can run on normal non 6gb image on jenkins

[LorenzoBettini]

@cdietrich what would that imply? As far as I know, we only need Xvnc for UI and SWTBot tests

[cdietrich]

i assume we could have more than 2 parallel builds ... the question can the tests handle it or will we oom. dont know with which xmxes we start plugin tests and the build itself

[LorenzoBettini]

@cdietrich I think GitHub Actions provides 7Gb of RAM. Do you know how much RAM the normal JIRO images provide?

[cdietrich]

4 For Xtend we used therefore the 6gb image but I don’t know if this is really needed

[LorenzoBettini]

@cdietrich let's see what happens: https://github.com/eclipse/xtext/tree/lb_2202_centos-8

[LorenzoBettini]

first impressions: the job started immediately... UI tests are running... let's see what happens with Xtend...

[LorenzoBettini]

@cdietrich unfortunately, it didn't work:

org.eclipse.xtend.core.tests.build.XtendIncrementalBuilderPerformanceTest.testPerf51200(org.eclipse.xtend.core.tests.build.XtendIncrementalBuilderPerformanceTest)
19:47:53    Run 1: XtendIncrementalBuilderPerformanceTest.testPerf51200 » OutOfMemory Java heap s...
19:47:53    Run 2: XtendIncrementalBuilderPerformanceTest.testPerf51200 » OutOfMemory Java heap s...
19:47:53    Run 3: XtendIncrementalBuilderPerformanceTest.testPerf51200 » OutOfMemory Java heap s...
19:47:53    Run 4: XtendIncrementalBuilderPerformanceTest.testPerf51200 » OutOfMemory Java heap s...
19:47:53 

[cdietrich]

the question is: can and should we make the test smaller cc @szarnekow

[LorenzoBettini]

just an experiment, I tried commenting out testPerf51200, let's see how it goes...

[LorenzoBettini]

unfortunately, we then get another failure later:

Running org.eclipse.xtend.ide.tests.quickfix.AddJunitLibToClasspathQuickfixTest

With this reason

[ERROR] Failed to execute goal org.eclipse.tycho:tycho-surefire-plugin:2.7.5:test (default-test) on project org.eclipse.xtend.ide.tests: An unexpected error occurred while launching the test runtime (process returned error code 137(SIGKILL received?)). The process logfile /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/target/work/data/.metadata/.log might contain further details. Command-line used to launch the sub-process was /opt/tools/java/temurin/jdk-11/latest/bin/java -Dosgi.noShutdown=false -Dosgi.os=linux -Dosgi.ws=gtk -Dosgi.arch=x86_64 -Xmx1500m -Dosgi.clean=true -jar /home/jenkins/.m2/repository/p2/osgi/bundle/org.eclipse.equinox.launcher/1.6.400.v20210924-0641/org.eclipse.equinox.launcher-1.6.400.v20210924-0641.jar -data /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/target/work/data -install /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/target/work -configuration /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/target/work/configuration -application org.eclipse.tycho.surefire.osgibooter.uitest -testproperties /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/target/surefire.properties -pluginCustomization /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests/../releng/org.eclipse.xtend.tycho.tests.parent/pluginCustomization.ini in working directory /home/jenkins/agent/workspace/xtext_lb_2202_centos-8/org.eclipse.xtend.ide.tests

and, as you might guess, error code 137 indicates that the container was terminated due to an out of memory issue.

[cdietrich]

yes i assume so. the question is what eats so much

[LorenzoBettini]

The archived log file does not seem to show any helpful trace from the first inspection. In any case, I'd say 4 Gb is too few in general...

[cdietrich]

@LorenzoBettini maybe we can experiment with something like https://gist.github.com/rednaxelafx/1513061 and archive the histogram maybe this gives an indication (e.g. too many open editors ...) am also no sure how much memory the host maven process takes. is it also 2gb. if yes do we need so much.

[LorenzoBettini]

@cdietrich I'm afraid I have no experience with memory dumps, profilers and the likes :( you might want to try to reduce the MAVEN_OPTS memory, or remove it completely. You can do that on top of the branch I pushed yesterday. In general, is there any way to beg... ehm... ask the Eclipse foundation to give us more memory or parallel jobs with 6Gb? O:)

[cdietrich]

i assume it wont be possible to have more machines. this would require a memory company to assign "their" budget to the project to my knowledge.

[cdietrich]

hmmm in test.parent we seem to set Xmx to 1500m anyway. so i wonder if the host maven then really eats too much und then the test eclipe actually starts to really allocate all of it

[LorenzoBettini]

Will you do some experiments? I'm experimenting with the toolchain stuff.

[cdietrich]

Will you do some experiments? yes

[LorenzoBettini]

by the way, I wonder if using a local virtual machine with only 4Gb would make experiments easier

[cdietrich]

yes but i dont have such a one. nor a usable docker setup. the images for cbi should be available there

[LorenzoBettini]

@cdietrich I saw that it once worked on Jenkins! Was it by chance?

[cdietrich]

its failing at random places. e.g. also the war plugin for the jetty example

[cdietrich]

this i also can see local if i run all project with 512m when i look into the war i at least see a ton of compile time deps like mwe and emf codegen

[cdietrich]

locall 1024 work fine (with no tests). will also try 768 @LorenzoBettini do we run maven surefire in fork mode or in the maven vm

[LorenzoBettini]

@cdietrich I don't think we touch the defaults under that respect. I wouldn't even know what would happen...

[cdietrich]

Of course on my local linux box it works perfectly fine with Xmx1G set in Maven Opts