-
Bug
-
Resolution: Fixed
-
P3
-
11, 14
-
b21
Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
---|---|---|---|---|---|---|
JDK-8231994 | 11.0.6-oracle | Thomas Schatzl | P3 | Resolved | Fixed | b01 |
On our internal Cassandra benchmark setup once or twice per run pause times are exceptionally long.
This seems to be always caused by one thread in the Thread Roots scan phase, i.e. the log for that phase looks as follows:
[2019-09-05T13:22:07.788+0100][942.036s][1567686127788ms][trace][gc,phases ] GC(271) Thread Roots (ms): Min: 0.0, Avg: 85.2, Max: 1954.3, Diff: 1954.3, Sum: 1959.1, Workers: 23
[2019-09-05T13:22:07.788+0100][942.036s][1567686127788ms][trace][gc,phases,task ] GC(271) 0.0 0.3 0.3 0.2 0.2 1.0 0.3 1954.3 0.2 0.2 0.2 0.3 0.2 0.3 1.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Seems to occur since JDK11 (not jdk8u231 at least), and only seen G1, not others like ZGC.
On that setup, THP is off, and also occurs if logs are created on a ramdisk. Sys time is zero.
This seems to be always caused by one thread in the Thread Roots scan phase, i.e. the log for that phase looks as follows:
[2019-09-05T13:22:07.788+0100][942.036s][1567686127788ms][trace][gc,phases ] GC(271) Thread Roots (ms): Min: 0.0, Avg: 85.2, Max: 1954.3, Diff: 1954.3, Sum: 1959.1, Workers: 23
[2019-09-05T13:22:07.788+0100][942.036s][1567686127788ms][trace][gc,phases,task ] GC(271) 0.0 0.3 0.3 0.2 0.2 1.0 0.3 1954.3 0.2 0.2 0.2 0.3 0.2 0.3 1.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Seems to occur since JDK11 (not jdk8u231 at least), and only seen G1, not others like ZGC.
On that setup, THP is off, and also occurs if logs are created on a ramdisk. Sys time is zero.
- backported by
-
JDK-8231994 Waiting on completion of strong nmethod processing causes long pause times with G1
-
- Resolved
-
- relates to
-
JDK-8201313 Times of sub-phases of ext root scan may be larger than the sum of individual timings
-
- Closed
-
-
JDK-6672778 G1 should trim task queues more aggressively during evacuation pauses
-
- Resolved
-
-
JDK-8249528 Remove obsolete comment in G1RootProcessor::process_java_roots
-
- Resolved
-