我正在以大约80,000 EPS的速度从Spark写入ES。 虽然早些时候运行良好,但最近索引率逐渐放缓,结果导致火花作业堆积起来。
我可以调整/验证哪些群集设置来解决此问题
群集设置: -
{
"persistent": {
"cluster": {
"routing": {
"allocation": {
"cluster_concurrent_rebalance": "160",
"node_concurrent_recoveries": "2",
"disk": {
"threshold_enabled": "true",
"watermark": {
"low": "85%",
"high": "95%"
}
},
"node_initial_primaries_recoveries": "40",
"enable": "all"
}
}
},
"indices": {
"breaker": {
"fielddata": {
"limit": "50%"
}
},
"recovery": {
"concurrent_streams": "80",
"max_bytes_per_sec": "50mb"
},
"store": {
"throttle": {
"max_bytes_per_sec": "500gb"
}
}
},
"threadpool": {
"bulk": {
"queue_size": "5000",
"size": "40",
"type": "fixed"
}
},
"discovery": {
"zen": {
"minimum_master_nodes": "2"
}
}
},
"transient": {
"cluster": {
"routing": {
"allocation": {
"enable": "all"
}
}
}
}
}
群集状态: -
{
"cluster_name": "**********",
"status": "green",
"timed_out": false,
"number_of_nodes": 105,
"number_of_data_nodes": 100,
"active_primary_shards": 7315,
"active_shards": 7330,
"relocating_shards": 0,
"initializing_shards": 0,
"unassigned_shards": 0,
"delayed_unassigned_shards": 0,
"number_of_pending_tasks": 0,
"number_of_in_flight_fetch": 0,
"task_max_waiting_in_queue_millis": 0,
"active_shards_percent_as_number": 100
}