CMS: abort preclean due to time

CMS: abort preclean due to time 2509869.358: [CMS-concurrent-abortable-preclean: 8.180/8.380 secs] [Times: user=8.20 sys=0.00, real=8.38 secs]

2509869.358: [GC (CMS Final Remark) [YG occupancy: 1322862 K (3774912 K)]2509869.359: [Rescan (parallel) , 0.1282081 secs]2509869.487: [weak refs processing, 0.0000395 secs]2509869.487: [class unloa$

2509869.516: [CMS-concurrent-sweep-start]

2509869.638: [CMS-concurrent-sweep: 0.122/0.122 secs] [Times: user=0.12 sys=0.00, real=0.12 secs]

2509869.638: [CMS-concurrent-reset-start]

2509869.651: [CMS-concurrent-reset: 0.013/0.013 secs] [Times: user=0.02 sys=0.00, real=0.01 secs]

2509871.650: [GC (CMS Initial Mark) [1 CMS-initial-mark: 498126K(4194304K)] 1820988K(7969216K), 0.1300750 secs] [Times: user=0.77 sys=0.00, real=0.13 secs]

2509871.780: [CMS-concurrent-mark-start]

2509872.022: [CMS-concurrent-mark: 0.241/0.241 secs] [Times: user=0.48 sys=0.00, real=0.24 secs]

2509872.022: [CMS-concurrent-preclean-start]

2509872.029: [CMS-concurrent-preclean: 0.007/0.007 secs] [Times: user=0.01 sys=0.00, real=0.01 secs]

2509872.029: [CMS-concurrent-abortable-preclean-start]

During our historical server being unresponsive yesterday I found this above over and over in the historical log.

The druid cluster has run very well for quite some time before yesterday when the broker queries to the historical started timing out.

Here is my jvm.config for the historical service:

-server

-Xmx8g

-Xms8g

-XX:NewSize=4g

-XX:MaxNewSize=4g

-XX:MaxDirectMemorySize=16g

-XX:+UseConcMarkSweepGC

-XX:+PrintGCDetails

-XX:+PrintGCTimeStamps

-Duser.timezone=UTC

-Dfile.encoding=UTF-8

-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager

-Djava.io.tmpdir=/data/tmp

Help and recommendations appreciated.