[11u] RFR 8185525: Add JFR event for DictionarySizes

Andrey Petushkov andrey.petushkov at gmail.com
Mon Jan 27 09:52:19 UTC 2020 

Hi Christoph, All,

Please excuse me for taking so long to respond.
Finally we got a list of JFR changes which are considered by Azul as
worth backporting to jdk11.

Features and changesets they depend on:
8185525: Add JFR event for DictionarySizes
   8223599: minimal build fails after JDK-8185525
   8225310: JFR crashed in
JfrPeriodicEventSet::requestProtectionDomainCacheTableStatistics()
8196341: Add JFR events for parallel phases of G1
   8211239: Build fails without JFR: empty JFR events signatures mismatch
8216041: [Event Request] - Deoptimization
8226897: Provide object age with JFR OldObjectSample event
8232594: Make the output of the JFR command duration more user friendly
8226511: Implement JFR Event Streaming
   8233700: EventStream not closed
   8233870: JFR TestSetEndTime.java times out - onClose() is never called
   8234059: Stress test fails with "Unexpected Exception in thread JFR
Event Stream"
   8234671: JFR api/consumer/recordingstream/TestStart.java failed due
to timeout at testStartTwice()
   8234684: JFR crashes when rotating the JFR output during assertion failure
   8234888: EventStream::close doesn't abort streaming thread
   8229209: [TESTBUG] test for cross-process JFR event streaming
   8234703: [TESTBUG] JFR TestOutOfProcessMigration.java should clean up files
   8235454: [TESTBUG] Basic test for JFR event streaming for jdk_jfr_sanity

Bugfixes and improvements:
8215237: jdk.jfr.Recording javadoc does not compile
8216565: Specifying the same path creates a new directory in JFR.configure
   8231317: jdk/jfr/jcmd/TestJcmdConfigure.java fails with
"java.lang.RuntimeException: assertTrue: expected true, was false"
8216995: Clean up JFR command line processing
8217089: JFR: Lazy install os interface components for improved startup
8219205: JFR file without license header
8230400: Missing constant pool entry for a method in stacktrace
8230767: FlightRecorderListener returns null recording
8220344: Build failures when using --with-jvm-features=-g1gc,-jfr
8233111: Epoch shift synchronization point for Compiler threads
8233197: Invert JvmtiExport::post_vm_initialized() and
Jfr:on_vm_start() start-up order for correct option parsing
8233375: JFR emergency dump does not recover thread state
8234060: Potential memory reordering problem in JfrBuffer flush mechanism
8235390: JfrEmergencyDump::on_vm_shutdown crashes
8235654: JFR leak profiler should not trace through the StringTable
8222379: JFR TestClassLoadEvent.java failed due to EXCEPTION_ACCESS_VIOLATION
8232905: JFR fails with assertion: assert(t->unflushed_size() == 0)
failed: invariant
8233416: JFR: assert((((((klass)->trace_id()) & (((1 << 1) << 8) |
(JfrTraceIdEpoch::in_use_this_epoch_bit()))) != 0))) failed: invariant
8210024: JFR calls virtual is_Java_thread from ~Thread()
8225004: Remove invalid assertion in jfr_conditional_flush()
8225797: OldObjectSample event creates unexpected amount of checkpoint data
   8231025: Incorrect method tag offset for big endian platform
   8231081: TestMetadataRetention fails due to missing symbol id
8209802: Garbage collectors should register JFR types themselves to
avoid build errors
(the last 6 are already have backport-request label)

Most of those we'd like to backport to jdk8 as well

Thanks,
Andrey


On Thu, Dec 19, 2019 at 5:06 PM Langer, Christoph
<christoph.langer at sap.com> wrote:
>
> Hi Andrey,
>
> thanks for your comments about these JFR event backports.
>
> Can you compile a list of items that you’d like to backport to both, 8u JFR and 11u? Then we can sync together with Mario and Andrew Haley and take a decision about what we want to approve.
>
> Cheers
> Christoph
>
> From: Andrey Petushkov <andrey at azul.com>
> Sent: Montag, 9. Dezember 2019 13:08
> To: Denghui Dong <denghui.ddh at alibaba-inc.com>; Erik Gahlin <erik.gahlin at oracle.com>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net>; Mario Torre <neugens at redhat.com>; Langer, Christoph <christoph.langer at sap.com>
> Subject: Re: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi guys, All,
>
> No, there is not special reasoning for back-porting of this particular event type. Feel free to reject the request. A few comments though
> - we at Azul see JFR as a very useful and successful tool for solving customer issues. So in general we try to keep up with new fixes and features there, unless we consider fix as too risky or inappropriate
> - all these JFR backports proposed are undergo full test cycle, involving all tests we have (including of course all jtreg) on all platforms supported by Azul (well, a lot)
> - @Christoph regarding fix not yet pushed to 8u. We (Marrio and team involved into JFR to 8u backporting)  have a policy that we don't push into 8u until it's pushed into 11u. Hence it's not pushed
>
> Regards,
> Andrey
>
>
> On 9 Dec 2019, at 12:33, Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>> wrote:
>
> I think we could wait for Andrey's reply to this issue before we decide whether to reject the backport.
>
> Thanks,
> Denghui Dong
> ------------------------------------------------------------------
> From:Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com>>
> Send Time:2019年12月9日(星期一) 17:17
> To:董登辉(卓昂) <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>>; Erik Gahlin <erik.gahlin at oracle.com<mailto:erik.gahlin at oracle.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net>>; Andrey Petushkov <andrey at azul.com<mailto:andrey at azul.com>>; Mario Torre <neugens at redhat.com<mailto:neugens at redhat.com>>
> Subject:RE: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Denghui,
>
> ok, I see. So, taking a glance on http://hg.openjdk.java.net/jdk8u/jdk8u-jfr-incubator/hotspot/, I can’t see that the bug is pushed there. So, if that’s true, I’d rather reject the backport to 11u for now. Would you agree?
>
> Thanks
> Christoph
>
> From: Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>>
> Sent: Montag, 9. Dezember 2019 10:06
> To: Erik Gahlin <erik.gahlin at oracle.com<mailto:erik.gahlin at oracle.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net>>; Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com>>; Andrey Petushkov <andrey at azul.com<mailto:andrey at azul.com>>; Mario Torre <neugens at redhat.com<mailto:neugens at redhat.com>>
> Subject: Re: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Christoph,
>   I was on vacation last week, sorry for the late reply.
>   In fact, the backport of 8185525 for 8u was requested by @Andrey at first (https://mail.openjdk.java.net/pipermail/jdk8u-dev/2019-September/010204.html)
>   For the sake of consistency, I made this backport for 11u.
>   And in my opinion, these events are not very useful for users indeed.
>
> Thanks,
> Denghui Dong
> ------------------------------------------------------------------
> From:Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com>>
> Send Time:2019年12月6日(星期五) 16:32
> To:Erik Gahlin <erik.gahlin at oracle.com<mailto:erik.gahlin at oracle.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net>>
> Cc:董登辉(卓昂) <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>>
> Subject:RE: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Erik,
>
> thanks for you input on this. I guess you have a very good point here – the bar for backporting new events should be quite high.
>
> @Denghui: Is there any motivation for this backport from your end that comes from support requirements/customer incidents?
>
> Thanks
> Christoph
>
> From: Erik Gahlin <erik.gahlin at oracle.com<mailto:erik.gahlin at oracle.com>>
> Sent: Freitag, 6. Dezember 2019 05:14
> To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net>>
> Cc: Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>>; Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com>>
> Subject: Re: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi,
>
> What’s nice about jFR is that the file format is self-describing, so a program like JMC can see what events exist or does not exist in a recording.
>
> Still, I think the bar should be high when backporting events. Not just for stability reasons, but If different JDK distributions or update releases have different events, it will confuse users, or lead to bugs, when users switch from one JDK to another.
>
> In my opinion, events should only be backported if support engineers truly need them to diagnose a certain type of bugs in the JVM/JDK. For this happen, there should be at least a few incidents where they have come to the conclusion that the events are needed. If there are no such real cases, the events should not be backported.
>
> My $.02
>
> Erik
>
> On 5 Dec 2019, at 23:28, Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com>> wrote:
>
> Hi Denghui,
>
> since no objections were raised, I’m approving JDK-8185525. I’ll push it, once I get approval for JDK-8223599 as well. Testing in our system doesn’t show regressions.
>
> Cheers
> Christoph
>
> From: Langer, Christoph
> Sent: Freitag, 22. November 2019 00:30
> To: Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net>>
> Cc: Andrew Haley <aph at redhat.com<mailto:aph at redhat.com>>
> Subject: RE: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi,
>
> we had a discussion on this list regarding JFR feature requests. As a summary, I take we should
>
>  1.  Generally reject backports of new features from higher releases if these are too extensive or require lots of prerequisites. Exceptions might be made if the feature to be backported exists in lower releases (e.g. OpenJDK8u after the JFR backport has been integrated)
>  2.  Be very careful when it comes to extensive bugfixes for existing features
>
> This fix is actually on the edge of category a), since it is a new feature and it is not trivial in its nature. It needs some manual adaption but fortunately no other prerequisites.
>
> Given that Denghui has done the effort of backporting it already and the patch together with another small follow up fix runs through our regression testing at SAP successfully, I could imagine to allow it going in to 11u. I would then like to see it submitted at the early stage of the 11.0.7 cycle.
>
> So, @all, please raise objections to this decision until next Friday, 29th of November. If I don’t hear any, I’ll approve it after that date.
>
> Thanks
> Christoph
>
> From: Langer, Christoph
> Sent: Mittwoch, 30. Oktober 2019 15:41
> To: Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com><mailto:denghui.ddh at alibaba-inc.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net><mailto:jdk-updates-dev at openjdk.java.net>>
> Subject: RE: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Denghui,
>
> sorry for the late reply.
>
> I was discussing with Andrew Haley about JFR backports. He encouraged to do a general discussion on the mailing list about the policy regarding non-trivial JFR related backports, which I started just today:
> https://mail.openjdk.java.net/pipermail/jdk-updates-dev/2019-October/002053.html
> Let’s see if there’s any feedback…
>
> Best regards
> Christoph
>
> From: Denghui Dong <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com><mailto:denghui.ddh at alibaba-inc.com>>
> Sent: Dienstag, 22. Oktober 2019 04:09
> To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net><mailto:jdk-updates-dev at openjdk.java.net>>; Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com><mailto:christoph.langer at sap.com>>
> Subject: Re: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Christoph,
>  What's the status of this backport ?
>  By the way, the previous webrev didn't contain the original commit, so I uploaded a new one in http://cr.openjdk.java.net/~ddong/8185525/webrev/
>
> Thanks,
> Denghui Dong
> ------------------------------------------------------------------
> From:Langer, Christoph <christoph.langer at sap.com<mailto:christoph.langer at sap.com><mailto:christoph.langer at sap.com>>
> Send Time:2019年10月8日(星期二) 22:34
> To:董登辉(卓昂) <denghui.ddh at alibaba-inc.com<mailto:denghui.ddh at alibaba-inc.com><mailto:denghui.ddh at alibaba-inc.com>>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net><mailto:jdk-updates-dev at openjdk.java.net>>
> Subject:RE: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi Denghui,
>
> this is quite a large JFR enhancement.
>
> Looking at the 11u webrev, I didn't see any obvious issues although there are obviously quite some alterations from the original patch. I'll run it through our regression testing.
>
> However, in any case, we need to carefully decide if and when we allow this patch to jdk11u-dev. I'll get back to you soon.
>
> Best regards
> Christoph
> -----Original Message-----
> From: jdk-updates-dev <jdk-updates-dev-bounces at openjdk.java.net<mailto:jdk-updates-dev-bounces at openjdk.java.net><mailto:jdk-updates-dev-bounces at openjdk.java.net>> On
> Behalf Of Denghui Dong
> Sent: Donnerstag, 26. September 2019 17:03
> To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net<mailto:jdk-updates-dev at openjdk.java.net><mailto:jdk-updates-dev at openjdk.java.net>>
> Subject: [11u] RFR 8185525: Add JFR event for DictionarySizes
>
> Hi all,
>  Please review this backport, original patch doesn't apply cleanly, because
> SymbolTable has made some changes in upstream (I think it's not necessary
> to backport those changes)
> and class ClassLoaderDataGraph is located in different files.
> 11u-webrev:
> http://cr.openjdk.java.net/~wzhuo/8185525/webrev.00/
> Original bug:
> https://bugs.openjdk.java.net/browse/JDK-8185525
> http://hg.openjdk.java.net/jdk/jdk/rev/865ec913f916
> Testing: test/jdk/jdk/jfr/event/runtime/TestTableStatisticsEvent.java
>
> Thanks,
> Denghui Dong
>

More information about the jdk-updates-dev mailing list