Details
-
Bug
-
Resolution: Not a Bug
-
Major
-
None
-
AMQ 7.2.0.GA
-
None
Description
Artemis broker-client-threads and others seem to maintain large (16mb) buffer caches at the thread-local level. This seems to be as-designed, but results in large amounts of memory usage in environments with lots of threading.
Class Name | Ref. Objects | Shallow Heap | Ref. Shallow Heap | Retained Heap -------------------------------------------------------------------------------------------------------------------------------------------------- java.lang.Thread @ 0x60a96f750 default task-223 Thread | 1 | 120 | 88 | 16,304 '- threadLocals java.lang.ThreadLocal$ThreadLocalMap @ 0x60a995250 | 1 | 24 | 88 | 15,928 '- table java.lang.ThreadLocal$ThreadLocalMap$Entry[128] @ 0x73cca9e90 | 1 | 528 | 88 | 15,904 '- [15] java.lang.ThreadLocal$ThreadLocalMap$Entry @ 0x6050d4418 | 1 | 32 | 88 | 1,280 '- value io.netty.util.internal.InternalThreadLocalMap @ 0x6050d4438 | 1 | 136 | 88 | 1,248 '- indexedVariables java.lang.Object[32] @ 0x6050d44c0 | 1 | 144 | 88 | 1,112 '- [1] io.netty.buffer.PoolThreadCache @ 0x600258408 | 1 | 72 | 88 | 195,832 '- heapArena io.netty.buffer.PoolArena$HeapArena @ 0x5e8fc3398| 1 | 144 | 88 | 2,760 '- qInit io.netty.buffer.PoolChunkList @ 0x5e8fc3ea0 | 1 | 40 | 88 | 40 '- head io.netty.buffer.PoolChunk @ 0x60dc7e098 | 1 | 88 | 88 | 16,795,048 --------------------------------------------------------------------------------------------------------------------------------------------------
This behavior was observed in the embedded broker in JBoss EAP. There do seem to be some global / JVM flags available for tuning netty caching behavior, but the ramifications of adjusting these settings globally is unknown. Are there any settings for tuning this usage at the broker level?
Attachments
Issue Links
- is related to
-
-
- Verified
-
-
-
- Closed
-
