8319876: Reduce memory consumption of VM_ThreadDump::doit #16598

yanglong1010 · 2023-11-10T09:13:06Z

I would like to fix this.

Create 4096 threads, and the stack depth of each thread is 256.
After running jmx.dumpAllThreads(true, true), the RSS reaches 5.3GiB.
After optimization, the RSS is 250MiB.

I would appreciate it if anyone could review this.

update

If the number of threads and stack depth are relatively large, we need to apply for more space in ResourceArea during the execution of jmx.dumpAllThreads(true, true).

The reason is that VM_ThreadDump::doit creates vframe for each frame of each thread.

jdk/src/hotspot/share/services/threadService.cpp

Line 704 in fe0ccdf

for (vframe* f = start_vf; f; f = f->sender() ) {

sizeof vframe is 4808 (bytes), and sizeof compiledVFrame is 4824 (bytes), mainly because the xmm registers in RegisterMap are relatively large. Assuming there are 4096 threads and each thread has 256 frames, the memory required is 4096 * 256 * 4824 = 4.7GiB。

These memories of all threads are released once by the the initial ResourceMark of VM_ThreadDump::doit.

jdk/src/hotspot/share/runtime/vmOperations.cpp

Line 269 in fe0ccdf

ResourceMark rm;

My solution is to add a ResourceMark for each thread.

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8319876: Reduce memory consumption of VM_ThreadDump::doit (Enhancement - P4)

Reviewers

David Holmes (@dholmes-ora - Reviewer)
Stefan Karlsson (@stefank - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/16598/head:pull/16598
$ git checkout pull/16598

Update a local copy of the PR:
$ git checkout pull/16598
$ git pull https://git.openjdk.org/jdk.git pull/16598/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 16598

View PR using the GUI difftool:
$ git pr show -t 16598

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/16598.diff

Webrev

Link to Webrev Comment

bridgekeeper · 2023-11-10T09:14:32Z

👋 Welcome back yanglong1010! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2023-11-10T09:15:08Z

@yanglong1010 The following label will be automatically applied to this pull request:

hotspot-runtime

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2023-11-10T09:18:05Z

Webrevs

shipilev · 2023-11-10T12:34:59Z

Please describe the fix a bit better.

My reading: vframe-s are allocated in resource arena during the stack walking, and we do not need them after the walk is completed. Safety: a) we do not persist the vframe-s after the walk, we only sometimes converting them to StackFrameInfo in ThreadStackTrace::add_stack_frame; b) all allocations on protected path are done in C heap, or under internal ResourceMark-s (e.g. in StackFrameInfo constructor).

Please run tier1 tier2 tier3 tier4 with fastdebug to catch any accidents.

dholmes-ora

Yes please explain the problem and the fix better. It is far from clear that every ResourceArea allocation that is made from this code can be safely discarded when this method returns. Which objects are the ones consuming so much space, and which can be erased? Maybe we should be avoiding these allocations rather than cleaning them up more promptly. ??

src/hotspot/share/services/threadService.cpp

yanglong1010 · 2023-11-13T09:53:32Z

Please describe the fix a bit better.

My reading: vframe-s are allocated in resource arena during the stack walking, and we do not need them after the walk is completed. Safety: a) we do not persist the vframe-s after the walk, we only sometimes converting them to StackFrameInfo in ThreadStackTrace::add_stack_frame; b) all allocations on protected path are done in C heap, or under internal ResourceMark-s (e.g. in StackFrameInfo constructor).

Please run tier1 tier2 tier3 tier4 with fastdebug to catch any accidents.

hi Aleksey,
After the execution of VM_ThreadDump::doit is completed, glibc (2.17) on my machine will not return the memory to the OS immediately. It takes a while to see RSS return to normal values.
To be precise and the most important, with 1024 threads and 1024 stack depth per thread, just before the destructor of ResourceMark of VM_ThreadDump::doit is invoked, the RSS reached 5.2GiB (detail below).
I'm running tier1 to tier4 in my environment, it may take some time.

yanglong1010 · 2023-11-13T10:00:11Z

hi David,

The situation we encountered is that in a Java process with a large number of threads, executing VM_ThreadDump::doit will cause Java to be killed by the Linux OOM Killer.

The reason is that VM_ThreadDump::doit creates vframe for each frame of each thread, sizeof vframe is 4808 (bytes), and sizeof compiledVFrame is 4824 (bytes).
RegisterMap._location of vframe occupies 577 words, mainly because the xmm registers is relatively large.
Assuming there are 1024 threads and each thread has 1024 frames, the memory required is 1024 * 1024 * 4824 = 4.7GiB.

I saw that in StackFrameInfo::StackFrameInfo(javaVFrame* jvf, bool with_lock_info), Method*, bci, and MonitorInfo._owner of javaVFrame have been copied to CHeap, so I feel that ResourceMark can be released.

tstuefe · 2023-11-13T10:51:30Z

@yanglong1010

ThreadStackTrace::_jni_locked_monitors and ThreadStackTrace::_frames are both GrowableArray, and IIUC its backing memory lives in RA as well. At least _frames will get added to inside your ResourceMark via add_stack_frame.

Are we sure adding an RM at that place is okay?

BTW, should be obvious but if you run tests, pls make sure to test debug builds to get Arena zapping on chopped arena chunks.

stefank · 2023-11-13T11:09:37Z

ThreadStackTrace::_jni_locked_monitors and ThreadStackTrace::_frames are both GrowableArray, and IIUC its backing memory lives in RA as well. At least _frames will get added to inside your ResourceMark via add_stack_frame.

From what I can see these two are CHeap-allocated:

ThreadStackTrace::ThreadStackTrace(JavaThread* t, bool with_locked_monitors) {
  _thread = t;
  _frames = new (mtServiceability) GrowableArray<StackFrameInfo*>(INITIAL_ARRAY_SIZE, mtServiceability);
  _depth = 0;
  _with_locked_monitors = with_locked_monitors;
  if (_with_locked_monitors) {
    _jni_locked_monitors = new (mtServiceability) GrowableArray<OopHandle>(INITIAL_ARRAY_SIZE, mtServiceability);
  } else {
    _jni_locked_monitors = nullptr;
  }
}

The MEMFLAGS indicate that they are CHeap-allocated.

tstuefe · 2023-11-13T11:21:16Z

ThreadStackTrace::_jni_locked_monitors and ThreadStackTrace::_frames are both GrowableArray, and IIUC its backing memory lives in RA as well. At least _frames will get added to inside your ResourceMark via add_stack_frame.

From what I can see these two are CHeap-allocated:
ThreadStackTrace::ThreadStackTrace(JavaThread* t, bool with_locked_monitors) {
  _thread = t;
  _frames = new (mtServiceability) GrowableArray<StackFrameInfo*>(INITIAL_ARRAY_SIZE, mtServiceability);
  _depth = 0;
  _with_locked_monitors = with_locked_monitors;
  if (_with_locked_monitors) {
    _jni_locked_monitors = new (mtServiceability) GrowableArray<OopHandle>(INITIAL_ARRAY_SIZE, mtServiceability);
  } else {
    _jni_locked_monitors = nullptr;
  }
}
The MEMFLAGS indicate that they are CHeap-allocated.

Wait, how does this work on expansion?

ThreadStackTrace::add_stack_frame -> GrowableArray::append -> ... ->

template <typename E, typename Derived>
void GrowableArrayWithAllocator<E, Derived>::expand_to(int new_capacity) {
...
  E* newData = static_cast<Derived*>(this)->allocate();

Calls GrowableArray::allocate() without MEMFLAGS, which calls GrowableArrayResourceAllocator::allocate

What am I missing?

Update, never mind, I see the big switch in GrowableArray::allocate now.

dcubed-ojdk · 2023-11-13T18:16:05Z

/label add serviceability

openjdk · 2023-11-13T18:16:31Z

@dcubed-ojdk
The serviceability label was successfully added.

yanglong1010 · 2023-11-16T02:06:10Z

Hi
I ran tier1, tier2, tier3, tier4 on my host machine.
tier1, tier2, and tier3 all passed.
Because my host does not have a display device, I added export JTREG_KEYWORDS="!headful" before running tier4.
Finally, some tests in tier4 that depend on the printing device failed, and the rest were successful.

dholmes-ora · 2023-11-16T09:29:28Z

Okay I'm running this through our CI testing as well. I can't see any easy way to check for escaping resource objects other than testing.

dholmes-ora · 2023-11-17T05:57:37Z

Our testing passed tiers 1-5. Thanks

openjdk · 2023-11-17T05:59:58Z

⚠️ @yanglong1010 the full name on your profile does not match the author name in this pull requests' HEAD commit. If this pull request gets integrated then the author name from this pull requests' HEAD commit will be used for the resulting commit. If you wish to push a new commit with a different author name, then please run the following commands in a local repository of your personal fork:

$ git checkout 8319876
$ git commit --author='Preferred Full Name <you@example.com>' --allow-empty -m 'Update full name'
$ git push

openjdk · 2023-11-17T05:59:59Z

@yanglong1010 This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8319876: Reduce memory consumption of VM_ThreadDump::doit

Reviewed-by: dholmes, stefank

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 82 new commits pushed to the master branch:

bbf52e0: 8319897: Move StackWatermark handling out of LockStack::contains
129c470: 8311932: Suboptimal compiled code of nested loop over memory segment
369bbec: 8319896: Remove monitor deflation from final audit
1588dd9: 8319567: Update java/lang/invoke tests to support vm flags
9727f4b: 8320199: Fix HTML 5 errors in java.math.BigInteger
d6aa7c8: 8314621: ClassNotFoundException due to lambda reference to elided anonymous inner class
52e2878: 8319987: compilation of sealed classes leads to infinite recursion
b05e69f: 8320209: VectorMaskGen clobbers rflags on x86_64
f3ed275: 8319103: Popups that request focus are not shown on Linux with Wayland
9e7a3ae: 8319630: Monitor final audit log lacks separator
... and 72 more: https://git.openjdk.org/jdk/compare/9cce9fe06780aa095b3aabdfa421f376ca7bfd08...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

As you do not have Committer status in this project an existing Committer must agree to sponsor your change. Possible candidates are the reviewers of this PR (@dholmes-ora, @stefank) but any other Committer may sponsor as well.

➡️ To flag this PR as ready for integration with the above commit message, type /integrate in a new comment. (Afterwards, your sponsor types /sponsor in a new comment to perform the integration).

yanglong1010 · 2023-11-17T07:05:24Z

@dholmes-ora David Thanks.

yanglong1010 · 2023-11-17T07:05:34Z

Could I get another review ?

yanglong1010 · 2023-11-17T08:40:53Z

/integrate

D-D-H · 2023-11-17T08:41:37Z

/sponsor

openjdk · 2023-11-17T08:42:20Z

@yanglong1010
Your change (at version b485290) is now ready to be sponsored by a Committer.

openjdk · 2023-11-17T08:43:23Z

Going to push as commit 8ec6b8d.
Since your change was applied there have been 82 commits pushed to the master branch:

bbf52e0: 8319897: Move StackWatermark handling out of LockStack::contains
129c470: 8311932: Suboptimal compiled code of nested loop over memory segment
369bbec: 8319896: Remove monitor deflation from final audit
1588dd9: 8319567: Update java/lang/invoke tests to support vm flags
9727f4b: 8320199: Fix HTML 5 errors in java.math.BigInteger
d6aa7c8: 8314621: ClassNotFoundException due to lambda reference to elided anonymous inner class
52e2878: 8319987: compilation of sealed classes leads to infinite recursion
b05e69f: 8320209: VectorMaskGen clobbers rflags on x86_64
f3ed275: 8319103: Popups that request focus are not shown on Linux with Wayland
9e7a3ae: 8319630: Monitor final audit log lacks separator
... and 72 more: https://git.openjdk.org/jdk/compare/9cce9fe06780aa095b3aabdfa421f376ca7bfd08...master

Your commit was automatically rebased without conflicts.

openjdk · 2023-11-17T08:43:33Z

@D-D-H @yanglong1010 Pushed as commit 8ec6b8d.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

yanglong1010 · 2023-11-17T08:44:31Z

Thank you all ~

8319876: Reduce memory consumption of VM_ThreadDump::doit

7d144c9

openjdk bot added the rfr Pull request is ready for review label Nov 10, 2023

openjdk bot added the hotspot-runtime hotspot-runtime-dev@openjdk.org label Nov 10, 2023

move ResourceMark to before start_vf

0ee0e17

dholmes-ora reviewed Nov 13, 2023

View reviewed changes

src/hotspot/share/services/threadService.cpp Outdated Show resolved Hide resolved

Use VMThread::vm_thread() to avoid the need to call Thread::current()

b485290

openjdk bot added the serviceability serviceability-dev@openjdk.org label Nov 13, 2023

dholmes-ora approved these changes Nov 17, 2023

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Nov 17, 2023

stefank approved these changes Nov 17, 2023

View reviewed changes

openjdk bot added the sponsor Pull request is ready to be sponsored label Nov 17, 2023

openjdk bot added the integrated Pull request has been integrated label Nov 17, 2023

openjdk bot closed this Nov 17, 2023

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review sponsor Pull request is ready to be sponsored labels Nov 17, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8319876: Reduce memory consumption of VM_ThreadDump::doit #16598

8319876: Reduce memory consumption of VM_ThreadDump::doit #16598

yanglong1010 commented Nov 10, 2023 •

edited by openjdk bot

Loading

bridgekeeper bot commented Nov 10, 2023

openjdk bot commented Nov 10, 2023

mlbridge bot commented Nov 10, 2023 •

edited

Loading

shipilev commented Nov 10, 2023

dholmes-ora left a comment

yanglong1010 commented Nov 13, 2023

yanglong1010 commented Nov 13, 2023

tstuefe commented Nov 13, 2023

stefank commented Nov 13, 2023

tstuefe commented Nov 13, 2023 •

edited

Loading

dcubed-ojdk commented Nov 13, 2023

openjdk bot commented Nov 13, 2023

yanglong1010 commented Nov 16, 2023

dholmes-ora commented Nov 16, 2023

dholmes-ora commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023 •

edited

Loading

yanglong1010 commented Nov 17, 2023

yanglong1010 commented Nov 17, 2023 •

edited

Loading

yanglong1010 commented Nov 17, 2023

D-D-H commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

yanglong1010 commented Nov 17, 2023

8319876: Reduce memory consumption of VM_ThreadDump::doit #16598

8319876: Reduce memory consumption of VM_ThreadDump::doit #16598

Conversation

yanglong1010 commented Nov 10, 2023 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Reviewing

Webrev

bridgekeeper bot commented Nov 10, 2023

openjdk bot commented Nov 10, 2023

mlbridge bot commented Nov 10, 2023 • edited Loading

Webrevs

shipilev commented Nov 10, 2023

dholmes-ora left a comment

Choose a reason for hiding this comment

yanglong1010 commented Nov 13, 2023

yanglong1010 commented Nov 13, 2023

tstuefe commented Nov 13, 2023

stefank commented Nov 13, 2023

tstuefe commented Nov 13, 2023 • edited Loading

dcubed-ojdk commented Nov 13, 2023

openjdk bot commented Nov 13, 2023

yanglong1010 commented Nov 16, 2023

dholmes-ora commented Nov 16, 2023

dholmes-ora commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023 • edited Loading

yanglong1010 commented Nov 17, 2023

yanglong1010 commented Nov 17, 2023 • edited Loading

yanglong1010 commented Nov 17, 2023

D-D-H commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

openjdk bot commented Nov 17, 2023

yanglong1010 commented Nov 17, 2023

yanglong1010 commented Nov 10, 2023 •

edited by openjdk bot

Loading

mlbridge bot commented Nov 10, 2023 •

edited

Loading

tstuefe commented Nov 13, 2023 •

edited

Loading

openjdk bot commented Nov 17, 2023 •

edited

Loading

yanglong1010 commented Nov 17, 2023 •

edited

Loading