无论在StorageGRID 11.1之前的版本上重新启动、Java堆使用量始终保持高水平

适用场景

• StorageGRID Webscale 10.3
• StorageGRID Webscale 10.4
• StorageGRID Webscale 11.0

问题描述

系统会不断报告CAHP (Java堆使用情况)警报。

'top -b -n 1' 在lumberjack的system_commands中、cassand+占用了大量内存和CPU：

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
  652 cassand+  20   0 63.437g 0.011t 2.889g S 125.3 47.1   1355:59 java

'nodetool cfstats'在lumberjack的Cassandra/system_commands中、显示存在较大的行key_prefix_by_bucket：

      Table: key_prefix_by_bucket
      SSTable count: 8
      Space used (live), bytes: 1880268774
      Space used (total), bytes: 1880268774
      Off heap memory used (total), bytes: 724800
      SSTable Compression Ratio: 0.3176106335841224
      Number of keys (estimate): 1024
      Memtable cell count: 2970
      Memtable data size, bytes: 808762
      Memtable switch count: 428
      Local read count: 4225600
      Local read latency: 6.692 ms
      Local write count: 86867
      Local write latency: 0.026 ms
      Pending tasks: 0
      Bloom filter false positives: 0
      Bloom filter false ratio: 0.00000
      Bloom filter space used, bytes: 10688
      Bloom filter off heap memory used, bytes: 10624
      Index summary off heap memory used, bytes: 384
      Compression metadata off heap memory used, bytes: 713792
      Compacted partition minimum bytes: 150
      Compacted partition maximum bytes: 1386179893   <<<< 1.4G
      Compacted partition mean bytes: 132014784
      Average live cells per slice (last five minutes): 1.0
      Average tombstones per slice (last five minutes): 0.0