Out Java team tried to debug this issue using Java tools. The problem we
encountered is that this extra memory is not tracked by Java at all (off
heap memory). We realised this using NMT -
https://docs.oracle.com/javase/8/docs/technotes/guides/troubleshoot/tooldescr007.html
.
There was not change in NMT tracked memory with and without annotations
enabled. Please see the output of "jcmd <PID> VM.native_memory summary
scale=MB" below:
Glassfish + annotations disabled (*551* MB RES on startup):
Total: reserved=2397MB, committed=*1159MB*
- Java Heap (reserved=1024MB, committed=1024MB)
(mmap: reserved=1024MB, committed=1024MB)
- Class (reserved=1077MB, committed=60MB)
(classes #10470)
(malloc=1MB #15133)
(mmap: reserved=1076MB, committed=59MB)
- Thread (reserved=23MB, committed=23MB)
(thread #80)
(stack: reserved=23MB, committed=23MB)
- Code (reserved=249MB, committed=28MB)
(malloc=5MB #8188)
(mmap: reserved=244MB, committed=23MB)
- GC (reserved=3MB, committed=3MB)
(mmap: reserved=3MB, committed=3MB)
- Internal (reserved=3MB, committed=3MB)
(malloc=3MB #13793)
- Symbol (reserved=14MB, committed=14MB)
(malloc=10MB #111702)
(arena=4MB #1)
- Native Memory Tracking (reserved=2MB, committed=2MB)
(tracking overhead=2MB)
Glassfish + annotations (*1909* MB RES on startup):
Total: reserved=2400MB, committed=*1162MB*
- Java Heap (reserved=1024MB, committed=1024MB)
(mmap: reserved=1024MB, committed=1024MB)
- Class (reserved=1079MB, committed=61MB)
(classes #10552)
(malloc=1MB #16204)
(mmap: reserved=1078MB, committed=59MB)
- Thread (reserved=22MB, committed=22MB)
(thread #74)
(stack: reserved=22MB, committed=22MB)
- Code (reserved=249MB, committed=30MB)
(malloc=6MB #8349)
(mmap: reserved=244MB, committed=25MB)
- GC (reserved=3MB, committed=3MB)
(mmap: reserved=3MB, committed=3MB)
- Internal (reserved=3MB, committed=3MB)
(malloc=3MB #13773)
- Symbol (reserved=14MB, committed=14MB)
(malloc=11MB #112818)
(arena=4MB #1)
- Native Memory Tracking (reserved=3MB, committed=3MB)
(tracking overhead=2MB)
- Arena Chunk (reserved=1MB, committed=1MB)
(malloc=1MB)
Apache + annotations (*1900* MB RES on startup):
Total: reserved=2400MB, committed=*1163MB*
- Java Heap (reserved=1024MB, committed=1024MB)
(mmap: reserved=1024MB, committed=1024MB)
- Class (reserved=1079MB, committed=61MB)
(classes #10679)
(malloc=1MB #16766)
(mmap: reserved=1078MB, committed=60MB)
- Thread (reserved=23MB, committed=23MB)
(thread #80)
(stack: reserved=23MB, committed=23MB)
- Code (reserved=249MB, committed=30MB)
(malloc=6MB #8492)
(mmap: reserved=244MB, committed=25MB)
- GC (reserved=3MB, committed=3MB)
(mmap: reserved=3MB, committed=3MB)
- Internal (reserved=3MB, committed=3MB)
(malloc=3MB #14043)
- Symbol (reserved=14MB, committed=14MB)
(malloc=11MB #113587)
(arena=4MB #1)
- Native Memory Tracking (reserved=3MB, committed=3MB)
(tracking overhead=2MB)
Thanks,
Ciprian
On Fri, Aug 28, 2015 at 11:20 AM, Christoph Läubrich <[email protected]>
wrote:
> In general you should use a proper tool (Eclipse MAT or JProfiler for
> example) to investigate where how much memory is used.
>
> Depending on your application, annotation parsing might involve a lot of
> (extra) class-loading and information caching... but this is just a wild
> guess!
>
>
> Am 28.08.2015 10:03, schrieb Ciprian Hacman:
>
> Hi,
>
> Just wanted to make things a little more clear.
>
> We changed from Apache to Glassfish, but we disabled the "annotations"
> module to reduce the memory usage. We had to move to Glassfish because the
> Apache engine did not work without "annotations".
>
> To make things easier to follow, here are some numbers for our webapp
> (numbers represent RES memory):
>
> 1. apache + annotations - 2.5GB
> 2. apache + annotations disabled - not working
> 3. glassfish + annotations - 2.5GB
> 4. glassfish + annotations disabled - 750MB
>
>
> Not sure why annotations make this huge difference for us and how to debug
> further.
> I see some other people bumped into similar issues (which gave us this
> idea):
>
> https://www.eclipse.org/forums/index.php/t/1067502/
>
>
> If anyone has any idea here, we would be happy to give it a try.
>
> Thanks,
> Ciprian
>
>
>
> _______________________________________________
> jetty-users mailing list
> [email protected]
> To change your delivery options, retrieve your password, or unsubscribe
> from this list, visit
> https://dev.eclipse.org/mailman/listinfo/jetty-users
>
_______________________________________________
jetty-users mailing list
[email protected]
To change your delivery options, retrieve your password, or unsubscribe from
this list, visit
https://dev.eclipse.org/mailman/listinfo/jetty-users
