JAVA_DIAGNOSTICS for s2i java 11 image uses deprecated and removed JVM options

[muff1nman]

Options such as PrintGCDateStamps are no longer valid.

[rhuss]

@vorburger I guess we need to updatre run-java.sh for this, right ? I.e. we probably need a version detection in this startup script.

[vorburger]

Options such as PrintGCDateStamps are no longer valid.

@muff1nman but that's just a warning, it "works" anyway - or doesn't it for you? Curious.

Would you (@muff1nman) be willing to contribute a PR to https://github.com/fabric8io-images/run-java-sh (perhaps inspired by https://github.com/fabric8io-images/run-java-sh/pull/73/files) which does the trick?

we probably need a version detection in this startup script

yeah I guess, but in run-java-sh? From a (quick) look, it would seem that https://github.com/fabric8io-images//pull/73/files already added a if [ "${JAVA_MAJOR_VERSION:-0}" -ge "10" ]; in some places... perhaps someone could build on top of that? (I'm unlikely to get to that at least this week.)

[rhuss]

BTW, I just submitted #214 which updates run-java.sh to not calculated defaults when running on Java 11.

[vorburger]

BTW, I just submitted #214 which updates run-java.sh to not calculated defaults when running on Java 11.

but #214 doesn't skip PrintGCDateStamps (this issue), does it?

[vorburger]

but #214 doesn't skip PrintGCDateStamps (this issue), does it?

yup, I could (still, so not fixed by #214) reproduce this bug like this.

So we need a if(Java 11) similar to run-java-sh/pull/73/ also here in run-java.sh ...

Who would be willing to have a closer look at which of -XX:NativeMemoryTracking=summary -XX:+PrintGC -XX:+PrintGCDateStamps -XX:+PrintGCTimeStamps -XX:+UnlockDiagnosticVMOptions are NOK for Java 11?

Some of these options perhaps can / need to be replaced with something else now, or (which of them exactly?) should just be skipped, and raise a PR against run-java-sh for it? Then we'll need a new release of https://github.com/fabric8io-images/run-java-sh and a bump to it here (like in #214) and then a rebase #216 which should then pass.

[rhuss]

Yeah, would be interesting which option combo to use for diagnostics for Java 11. Unfortunately, I'm overloaded, but if someone finds out a good combination and creates a PR for run-java.sh I can do a release and rebase quickly.

[muff1nman]

I have opened a potential PR, but my biggest concern is that I want to be able to override the -Xlog:gc option. I.e. the following are options that allow me to log gc to a pvc for later viewing:

-XX:+PrintFlagsFinal -Xlog:gc*:file=/logs/gc.log:time:filecount=5,filesize=25M  -XX:NativeMemoryTracking=summary -XX:+UnlockDiagnosticVMOptions -XX:+PrintNMTStatistics -XX:CompressedClassSpaceSize=150m -XX:MaxMetaspaceSize=150m -XX:MetaspaceSize=75m

The options for overriding -Xlog:gc as proposed in #217 are to:

1. rely on the behavior specified here (https://bugs.openjdk.java.net/browse/JDK-8142952?focusedCommentId=13866870&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-13866870) and move JAVA_OPTIONS to be applied after JAVA_DIAGNOSTICS
2. keep JAVA_DIAGNOSTICS off and instead specify all options via JAVA_OPTIONS and hope that JAVA_DIAGNOSTICS doesn't drive other valuable logic.
3. Build logic into JAVA_DIAGNOSTICS so that it omits any duplicate options already specified in JAVA_OPTIONS