ES性能测试

主节点Es-1: 配置8核，6g

调整配置为8核，12g

 

 

Node-3：配置为4核，12g

Node-4,5：配置为8核，4g

 

Es-1,2、node-1,2：配置为8核，12g

 

 

集群中尽量避免水桶效应；master节点内存应尽量大，否则gc时间过长，

 

 

[2017-03-06T08:38:35,604][INFO ][o.e.i.e.InternalEngine$EngineMergeScheduler] [node-5] [mtest][3] now throttling indexing: numMergesInFlight=10, maxNumMerges=9

[2017-03-06T08:39:00,352][INFO ][o.e.i.e.InternalEngine$EngineMergeScheduler] [node-5] [mtest][3] stop throttling indexing: numMergesInFlight=8, maxNumMerges=9

 

"index.merge.policy.max_merged_segment": "1g"

 

段合并影响较大，默认最大为5g，此时合并时系统开销较大，改为1g

 

    

"index.translog.sync_interval": "60s",

    "index.translog.durability": "async",

    "index.translog.flush_threshold_size": "1g",

    "index.merge.scheduler.max_thread_count": 2,

"index.merge.scheduler.max_merge_count": 2,    

    "index.merge.policy.max_merged_segment": "1g",

    "refresh_interval": "180s",

"number_of_shards": 6,

"number_of_replicas": 0,

    "index.routing.allocation.total_shards_per_node": 1

6*8核*10g，3w/s

持续写入15h，大小1.25T