streamLoad慢

【详述】streamLoad很慢,1条数据需要操作5~6s,导致实时流写入不可用,不知道streamLoad如何开启profile,使用insert into 排查问题
【背景】做过哪些操作?
【业务影响】
【是否存算分离】
【StarRocks版本】例如:3.0.7
【集群规模】测试环境(fe与be混部)
【表模型】聚合模型,场景约等于主键模型
【导入或者导出方式】streamLoad

【附件】
Query:
Summary:
- Query ID: e22f548f-b367-11ee-9c4f-005056ad08a8
- Start Time: 2024-01-15 13:35:25
- End Time: 2024-01-15 13:35:35
- Total: 10s561ms
- Query Type: Query
- Query State: OK
- StarRocks Version: 3.0.7-208c9db
- User: root
- Default Db: demo
- Sql Statement: /* ApplicationName=DBeaver 7.3.1 - SQLEditor <Script-351.sql> */ INSERT INTO demo.table_tm_fuel_demo_recently
(device_sn, data_time, recv_time, save_time, device_no, obdx_urea_cost, obdx_urea_inst_cost, obdx_dpf_egt, obdx_instant_fuel, obdx_engine_time, obdx_smoke_meter_state, obdx_pm_consistence, obdx_kvalue, obdx_opacity, login_type, login_order_num, login_sim_iccid, gps_fix_state, gps_latitude, gps_longitude, gps_altitude, gps_speed, gps_mileage, gps_oil_cost, gps_rev, gps_dir, gps_oil__percent, gps_running_time, gps_working_time, gps_longitude_gcj102, gps_latitude_gcj102, gps_longitude_bd09, gps_latitude_bd09, obd_detect_protocol, obd_mil_state, obd_detect_state, obd_detect_ready_state, obd_indentify_code, obd_id_version, obd_calibrate_verifiy, obd_iupr, obd_error_code_count, obd_list_error_code, df_oil_percent, df_air_input, df_atmo_pres, df_rev, df_dpf_different_press, df_upper_scr_nox_output, df_latitude, df_temper_coolant, df_down_scr_nox_outPut, df_speed, df_output_wrest, df_fix_state, df_temper_scr_output, df_rub_wrest, df_percent_reactant, df_temper_scr_input, df_fuel_velocity_flow, df_longitude, df_mileage, obdx_torque_mode, obdx_acc_pedal, obdx_oil_cost, obdx_urea_temper)
VALUES(‘xbtest’, ‘2024-01-12 09:02:38’, ‘2024-01-12 09:02:40’, ‘2024-01-12 09:06:34.646000000’, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, 2, 3, ‘89860621180001641810’, 3, 30.148159000, 120.087539000, 25.000000000, 45.900000000, 16599.100000000, 10568.500000000, 65535.000000000, 80, 98.000000000, 2147483647, 4294967295, 120.092346000, 30.145879000, 120.098854000, 30.151928000, 1.000000000, 0.000000000, 57067.000000000, 0.000000000, ‘LZ5N2CE52PB001351’, ‘HM09P370P185764000’, ‘9E1D71A630303030303030303030303030’, ‘005A00B00002000E004E005A00000000008D00040032005A000000000012005A007C00’, 0, ‘[570521324]’, 98.400000000, 215.200000000, 102.000000000, 712.500000000, 0.600000000, 245.000000000, 30.148401000, 83.000000000, 4.750000000, 10.391000000, 7.000000000, 0, 315.125000000, 5.000000000, 85.200000000, 329.718750000, 1.800000000, 120.088877000, 16599.200000000, NULL, NULL, NULL, NULL)

 - Variables: parallel_fragment_exec_instance_num=1,max_parallel_scan_instance_num=-1,pipeline_dop=0,enable_adaptive_sink_dop=true,enable_runtime_adaptive_dop=false
 - Collect Profile Time: 0

Planner:
- Analyzer: 0ms / 1
- CTEUniqueChecker: 0ms / 1
- ColumnReuseChecker: 0ms / 1
- CoordDeliverExec: 8ms / 1
- CoordPrepareExec: 0ms / 1
- InputDependenciesChecker: 0ms / 1
- InsertPlanner: 1ms / 1
- Optimizer: 0ms / 1
- PlanBuilder: 0ms / 1
- Total: 3ms / 1
- Transform: 0ms / 1
- TypeChecker: 0ms / 1
Optimizer:
- CostBaseOptimize: 0ms / 1
- PhysicalRewrite: 0ms / 1
- PlanValidate: 0ms / 1
- RuleBaseOptimize: 0ms / 1
- preprocessMvs: 0ms / 1
Execution Profile e22f548f-b367-11ee-9c4f-005056ad08a8:
- QueryAllocatedMemoryUsage: 160.86 KB
- QueryCumulativeCpuTime: 423.324us
- QueryCumulativeOperatorTime: 1.258ms
- QueryDeallocatedMemoryUsage: 153.36 KB
- QueryPeakMemoryUsage: 48.63 KB
Fragment 0:
- BackendAddresses: 172.18.1.22:9060
- BackendNum: 1
- FragmentInstancePrepareTime: 2.130ms
- prepare-fragment-ctx: 32.615us
- prepare-pipeline-driver: 431.170us
- prepare-query-ctx: 12.371us
- prepare-runtime-state: 431.170us
- InstanceAllocatedMemoryUsage: 160.86 KB
- InstanceDeallocatedMemoryUsage: 153.36 KB
- InstanceNum: 1
- InstancePeakMemoryUsage: 13.39 KB
- QueryMemoryLimit: 10.00 GB
Pipeline (id=1):
- ActiveTime: 219.188us
- __MAX_OF_ActiveTime: 793.374us
- __MIN_OF_ActiveTime: 21.300us
- BlockByInputEmpty: 0
- BlockByOutputFull: 0
- BlockByPrecondition: 0
- DegreeOfParallelism: 5
- DriverPrepareTime: 625.982us
- __MAX_OF_DriverPrepareTime: 1.12ms
- __MIN_OF_DriverPrepareTime: 478.858us
- DriverTotalTime: 1s626ms
- __MAX_OF_DriverTotalTime: 8s58ms
- __MIN_OF_DriverTotalTime: 13.92ms
- GlobalScheduleCount: 961
- __MAX_OF_GlobalScheduleCount: 927
- __MIN_OF_GlobalScheduleCount: 7
- GlobalScheduleTime: 372.288ms
- __MAX_OF_GlobalScheduleTime: 1s855ms
- __MIN_OF_GlobalScheduleTime: 1.517ms
- OverheadTime: 11.389us
- __MAX_OF_OverheadTime: 49.599us
- __MIN_OF_OverheadTime: 0ns
- PendingTime: 1s623ms
- __MAX_OF_PendingTime: 8s57ms
- __MIN_OF_PendingTime: 7.635ms
- InputEmptyTime: 21.803us
- __MAX_OF_InputEmptyTime: 26.166us
- __MIN_OF_InputEmptyTime: 16.816us
- FirstInputEmptyTime: 21.803us
- __MAX_OF_FirstInputEmptyTime: 26.166us
- __MIN_OF_FirstInputEmptyTime: 16.816us
- FollowupInputEmptyTime: 0ns
- OutputFullTime: 682.318us
- __MAX_OF_OutputFullTime: 694.233us
- __MIN_OF_OutputFullTime: 658.22us
- PendingFinishTime: 1s623ms
- __MAX_OF_PendingFinishTime: 8s56ms
- __MIN_OF_PendingFinishTime: 6.915ms
- PreconditionBlockTime: 0ns
- ScheduleCount: 5
- __MAX_OF_ScheduleCount: 1
- __MIN_OF_ScheduleCount: 1
- ScheduleTime: 2.342ms
- __MAX_OF_ScheduleTime: 5.435ms
- __MIN_OF_ScheduleTime: 168.643us
- TotalDegreeOfParallelism: 5
- YieldByPreempt: 0
- YieldByTimeLimit: 0
OLAP_TABLE_SINK:
CommonMetrics:
- CloseTime: 79.187us
- __MAX_OF_CloseTime: 100.278us
- __MIN_OF_CloseTime: 59.381us
- OperatorAllocatedMemoryUsage: 116.87 KB
- __MAX_OF_OperatorAllocatedMemoryUsage: 84.55 KB
- __MIN_OF_OperatorAllocatedMemoryUsage: 8.08 KB
- OperatorDeallocatedMemoryUsage: 84.84 KB
- __MAX_OF_OperatorDeallocatedMemoryUsage: 75.40 KB
- __MIN_OF_OperatorDeallocatedMemoryUsage: 2.36 KB
- OperatorPeakMemoryUsage: 25.97 KB
- __MAX_OF_OperatorPeakMemoryUsage: 25.97 KB
- __MIN_OF_OperatorPeakMemoryUsage: 5.72 KB
- OperatorTotalTime: 232.916us
- __MAX_OF_OperatorTotalTime: 825.565us
- __MIN_OF_OperatorTotalTime: 61.196us
- PrepareTime: 569.4us
- __MAX_OF_PrepareTime: 958.518us
- __MIN_OF_PrepareTime: 414.681us
- PullChunkNum: 0
- PullRowNum: 0
- PullTotalTime: 0ns
- PushChunkNum: 1
- __MAX_OF_PushChunkNum: 1
- __MIN_OF_PushChunkNum: 0
- PushRowNum: 1
- __MAX_OF_PushRowNum: 1
- __MIN_OF_PushRowNum: 0
- PushTotalTime: 75.310us
- __MAX_OF_PushTotalTime: 376.551us
- __MIN_OF_PushTotalTime: 0ns
- SetFinishedTime: 133ns
- __MAX_OF_SetFinishedTime: 246ns
- __MIN_OF_SetFinishedTime: 44ns
- SetFinishingTime: 78.285us
- __MAX_OF_SetFinishingTime: 389.481us
- __MIN_OF_SetFinishingTime: 296ns
UniqueMetrics:
- AutomaticPartition: false
- IndexNum: 1
- ReplicatedStorage: true
- TxnID: 8273909
- AllocAutoIncrementTime: 54ns
- __MAX_OF_AllocAutoIncrementTime: 272ns
- __MIN_OF_AllocAutoIncrementTime: 0ns
- CloseWaitTime: 5.97ms
- __MAX_OF_CloseWaitTime: 5.179ms
- __MIN_OF_CloseWaitTime: 5.70ms
- OpenTime: 0ns
- PrepareDataTime: 25.839us
- __MAX_OF_PrepareDataTime: 129.199us
- __MIN_OF_PrepareDataTime: 0ns
- ConvertChunkTime: 12.663us
- __MAX_OF_ConvertChunkTime: 63.316us
- __MIN_OF_ConvertChunkTime: 0ns
- ValidateDataTime: 12.94us
- __MAX_OF_ValidateDataTime: 60.471us
- __MIN_OF_ValidateDataTime: 0ns
- RowsFiltered: 0
- RowsRead: 1
- __MAX_OF_RowsRead: 1
- __MIN_OF_RowsRead: 0
- RowsReturned: 1
- __MAX_OF_RowsReturned: 1
- __MIN_OF_RowsReturned: 0
- RpcClientSideTime: 1s611ms
- __MAX_OF_RpcClientSideTime: 8s51ms
- __MIN_OF_RpcClientSideTime: 1.740ms
- RpcServerSideTime: 1s609ms
- __MAX_OF_RpcServerSideTime: 8s46ms
- __MIN_OF_RpcServerSideTime: 1000ns
- SendDataTime: 43.341us
- __MAX_OF_SendDataTime: 216.708us
- __MIN_OF_SendDataTime: 0ns
- PackChunkTime: 6.886us
- __MAX_OF_PackChunkTime: 34.431us
- __MIN_OF_PackChunkTime: 0ns
- SendRpcTime: 59.846us
- __MAX_OF_SendRpcTime: 112.741us
- __MIN_OF_SendRpcTime: 12.24us
- CompressTime: 0ns
- SerializeChunkTime: 5.901us
- __MAX_OF_SerializeChunkTime: 29.507us
- __MIN_OF_SerializeChunkTime: 0ns
- WaitResponseTime: 0ns
LOCAL_EXCHANGE_SOURCE (pseudo_plan_node_id=-100):
CommonMetrics:
- CloseTime: 727ns
- __MAX_OF_CloseTime: 1.275us
- __MIN_OF_CloseTime: 347ns
- OperatorAllocatedMemoryUsage: 5.63 KB
- __MAX_OF_OperatorAllocatedMemoryUsage: 1.13 KB
- __MIN_OF_OperatorAllocatedMemoryUsage: 1.13 KB
- OperatorDeallocatedMemoryUsage: 5.63 KB
- __MAX_OF_OperatorDeallocatedMemoryUsage: 1.13 KB
- __MIN_OF_OperatorDeallocatedMemoryUsage: 1.13 KB
- OperatorPeakMemoryUsage: 0.00
- OperatorTotalTime: 4.466us
- __MAX_OF_OperatorTotalTime: 11.219us
- __MIN_OF_OperatorTotalTime: 1.359us
- PrepareTime: 21.142us
- __MAX_OF_PrepareTime: 26.69us
- __MIN_OF_PrepareTime: 17.921us
- PullChunkNum: 1
- __MAX_OF_PullChunkNum: 1
- __MIN_OF_PullChunkNum: 0
- PullRowNum: 1
- __MAX_OF_PullRowNum: 1
- __MIN_OF_PullRowNum: 0
- PullTotalTime: 1.560us
- __MAX_OF_PullTotalTime: 7.804us
- __MIN_OF_PullTotalTime: 0ns
- PushChunkNum: 0
- PushRowNum: 0
- PushTotalTime: 0ns
- SetFinishedTime: 1.962us
- __MAX_OF_SetFinishedTime: 4.363us
- __MIN_OF_SetFinishedTime: 616ns
- SetFinishingTime: 215ns
- __MAX_OF_SetFinishingTime: 390ns
- __MIN_OF_SetFinishingTime: 92ns
UniqueMetrics:
Pipeline (id=0):
- ActiveTime: 421.128us
- BlockByInputEmpty: 0
- BlockByOutputFull: 0
- BlockByPrecondition: 0
- DegreeOfParallelism: 1
- DriverPrepareTime: 100.564us
- DriverTotalTime: 484.185us
- GlobalScheduleCount: 1
- GlobalScheduleTime: 421.128us
- OverheadTime: 0ns
- PendingTime: 0ns
- InputEmptyTime: 0ns
- FirstInputEmptyTime: 0ns
- FollowupInputEmptyTime: 0ns
- OutputFullTime: 0ns
- PendingFinishTime: 0ns
- PreconditionBlockTime: 0ns
- ScheduleCount: 1
- ScheduleTime: 63.57us
- TotalDegreeOfParallelism: 1
- YieldByPreempt: 0
- YieldByTimeLimit: 0
LOCAL_EXCHANGE_SINK (pseudo_plan_node_id=-100):
CommonMetrics:
- CloseTime: 419ns
- OperatorAllocatedMemoryUsage: 0.00
- OperatorDeallocatedMemoryUsage: 0.00
- OperatorPeakMemoryUsage: 0.00
- OperatorTotalTime: 16.511us
- PrepareTime: 21.719us
- PullChunkNum: 0
- PullRowNum: 0
- PullTotalTime: 0ns
- PushChunkNum: 1
- PushRowNum: 1
- PushTotalTime: 15.20us
- SetFinishedTime: 79ns
- SetFinishingTime: 993ns
UniqueMetrics:
- Type: Passthrough
- ShuffleNum: 5
- LocalExchangePeakMemoryUsage: 2.13 KB
UNION_CONST_SOURCE (plan_node_id=0):
CommonMetrics:
- CloseTime: 14.541us
- OperatorAllocatedMemoryUsage: 40.94 KB
- OperatorDeallocatedMemoryUsage: 27.55 KB
- OperatorPeakMemoryUsage: 13.38 KB
- OperatorTotalTime: 405.110us
- PrepareTime: 22.848us
- PullChunkNum: 1
- PullRowNum: 1
- PullTotalTime: 390.185us
- PushChunkNum: 0
- PushRowNum: 0
- PushTotalTime: 0ns
- RuntimeBloomFilterNum: 0
- RuntimeInFilterNum: 0
- SetFinishedTime: 101ns
- SetFinishingTime: 283ns
UniqueMetrics:

节点的负载高吗?看起来主要是rpc的时间占用__MAX_OF_RpcClientSideTime、__MAX_OF_RpcServerSideTime,网络压力大吗?

网络压力不大
只要不写的话负载都是很低的,一开始写磁盘io就满了,(数据量不大)

这个IO只有几十M,看下iostat -x 1的输出呢
另外这个表本身有多大?

这个表本身不大,新建的
一开始是聚合模型,我理解和主键模型是一样的,看了其他贴改成了主键模型
io看着也不大,我奇怪的是只要我一开始写他util就变成了100,测试是一条一条数据写

HDD盘吗,麻烦发下建表语句。另外写入的时候建议攒批写,不建议1条1条写入的。

实际是攒批写入的,sparkstraming,只不过在测试环境发现数据写入很慢,所以本地验证的时候发现1条数据写入也很慢

CREATE TABLE table_tm_fuel_demo_tmp (
device_sn varchar(32) NOT NULL COMMENT “device_sn”,
data_time datetime NOT NULL COMMENT “data_time”,
recv_time datetime NULL COMMENT “recv_time”,
save_time datetime NULL COMMENT “save_time”,
device_no bigint(20) NULL COMMENT “device_no”,
obdx_urea_cost decimal128(27, 9) NULL COMMENT “obdx_urea_cost”,
obdx_urea_inst_cost decimal128(27, 9) NULL COMMENT “obdx_urea_inst_cost”,
obdx_dpf_egt decimal128(27, 9) NULL COMMENT “obdx_dpf_egt”,
obdx_instant_fuel decimal128(27, 9) NULL COMMENT “obdx_instant_fuel”,
obdx_engine_time decimal128(27, 9) NULL COMMENT “obdx_engine_time”,
obdx_smoke_meter_state decimal128(27, 9) NULL COMMENT “obdx_smoke_meter_state”,
obdx_pm_consistence decimal128(27, 9) NULL COMMENT “obdx_pm_consistence”,
obdx_kvalue decimal128(27, 9) NULL COMMENT “obdx_kvalue”,
obdx_opacity decimal128(27, 9) NULL COMMENT “obdx_opacity”,
login_type bigint(20) NULL COMMENT “login_type”,
login_order_num bigint(20) NULL COMMENT “login_order_num”,
login_sim_iccid varchar(512) NULL COMMENT “login_sim_iccid”,
gps_fix_state bigint(20) NULL COMMENT “gps_fix_state”,
gps_latitude decimal128(27, 9) NULL COMMENT “gps_latitude”,
gps_longitude decimal128(27, 9) NULL COMMENT “gps_longitude”,
gps_altitude decimal128(27, 9) NULL COMMENT “gps_altitude”,
gps_speed decimal128(27, 9) NULL COMMENT “gps_speed”,
gps_mileage decimal128(27, 9) NULL COMMENT “gps_mileage”,
gps_oil_cost decimal128(27, 9) NULL COMMENT “gps_oil_cost”,
gps_rev decimal128(27, 9) NULL COMMENT “gps_rev”,
gps_dir bigint(20) NULL COMMENT “gps_dir”,
gps_oil__percent decimal128(27, 9) NULL COMMENT “gps_oil__percent”,
gps_running_time bigint(20) NULL COMMENT “gps_running_time”,
gps_working_time bigint(20) NULL COMMENT “gps_working_time”,
gps_longitude_gcj102 decimal128(27, 9) NULL COMMENT “gps_longitude_gcj102”,
gps_latitude_gcj102 decimal128(27, 9) NULL COMMENT “gps_latitude_gcj102”,
gps_longitude_bd09 decimal128(27, 9) NULL COMMENT “gps_longitude_bd09”,
gps_latitude_bd09 decimal128(27, 9) NULL COMMENT “gps_latitude_bd09”,
obd_detect_protocol decimal128(27, 9) NULL COMMENT “obd_detect_protocol”,
obd_mil_state decimal128(27, 9) NULL COMMENT “obd_mil_state”,
obd_detect_state decimal128(27, 9) NULL COMMENT “obd_detect_state”,
obd_detect_ready_state decimal128(27, 9) NULL COMMENT “obd_detect_ready_state”,
obd_indentify_code varchar(512) NULL COMMENT “obd_indentify_code”,
obd_id_version varchar(512) NULL COMMENT “obd_id_version”,
obd_calibrate_verifiy varchar(512) NULL COMMENT “obd_calibrate_verifiy”,
obd_iupr varchar(512) NULL COMMENT “obd_iupr”,
obd_error_code_count bigint(20) NULL COMMENT “obd_error_code_count”,
obd_list_error_code varchar(512) NULL COMMENT “obd_list_error_code”,
df_oil_percent decimal128(27, 9) NULL COMMENT “df_oil_percent”,
df_air_input decimal128(27, 9) NULL COMMENT “df_air_input”,
df_atmo_pres decimal128(27, 9) NULL COMMENT “df_atmo_pres”,
df_rev decimal128(27, 9) NULL COMMENT “df_rev”,
df_dpf_different_press decimal128(27, 9) NULL COMMENT “df_dpf_different_press”,
df_upper_scr_nox_output decimal128(27, 9) NULL COMMENT “df_upper_scr_nox_output”,
df_latitude decimal128(27, 9) NULL COMMENT “df_latitude”,
df_temper_coolant decimal128(27, 9) NULL COMMENT “df_temper_coolant”,
df_down_scr_nox_outPut decimal128(27, 9) NULL COMMENT “df_down_scr_nox_outPut”,
df_speed decimal128(27, 9) NULL COMMENT “df_speed”,
df_output_wrest decimal128(27, 9) NULL COMMENT “df_output_wrest”,
df_fix_state bigint(20) NULL COMMENT “df_fix_state”,
df_temper_scr_output decimal128(27, 9) NULL COMMENT “df_temper_scr_output”,
df_rub_wrest decimal128(27, 9) NULL COMMENT “df_rub_wrest”,
df_percent_reactant decimal128(27, 9) NULL COMMENT “df_percent_reactant”,
df_temper_scr_input decimal128(27, 9) NULL COMMENT “df_temper_scr_input”,
df_fuel_velocity_flow decimal128(27, 9) NULL COMMENT “df_fuel_velocity_flow”,
df_longitude decimal128(27, 9) NULL COMMENT “df_longitude”,
df_mileage decimal128(27, 9) NULL COMMENT “df_mileage”,
obdx_torque_mode decimal128(27, 9) NULL COMMENT “obdx_torque_mode”,
obdx_acc_pedal decimal128(27, 9) NULL COMMENT “obdx_acc_pedal”,
obdx_oil_cost decimal128(27, 9) NULL COMMENT “obdx_oil_cost”,
obdx_urea_temper decimal128(27, 9) NULL COMMENT “obdx_urea_temper”
) ENGINE=OLAP
UNIQUE KEY(device_sn, data_time)
COMMENT “OLAP”
PARTITION BY RANGE(data_time)
(PARTITION p_202307 VALUES [(“2023-07-01 00:00:00”), (“2023-08-01 00:00:00”)),
PARTITION p_202308 VALUES [(“2023-08-01 00:00:00”), (“2023-09-01 00:00:00”)),
PARTITION p_202309 VALUES [(“2023-09-01 00:00:00”), (“2023-10-01 00:00:00”)),
PARTITION p_202310 VALUES [(“2023-10-01 00:00:00”), (“2023-11-01 00:00:00”)),
PARTITION p_202311 VALUES [(“2023-11-01 00:00:00”), (“2023-12-01 00:00:00”)),
PARTITION p_202312 VALUES [(“2023-12-01 00:00:00”), (“2024-01-01 00:00:00”)),
PARTITION p_202401 VALUES [(“2024-01-01 00:00:00”), (“2024-02-01 00:00:00”)),
PARTITION p_202402 VALUES [(“2024-02-01 00:00:00”), (“2024-03-01 00:00:00”)),
PARTITION p_202403 VALUES [(“2024-03-01 00:00:00”), (“2024-04-01 00:00:00”)),
PARTITION p_202404 VALUES [(“2024-04-01 00:00:00”), (“2024-05-01 00:00:00”)))
DISTRIBUTED BY HASH(device_sn) BUCKETS 64
PROPERTIES (
“replication_num” = “1”,
“dynamic_partition.enable” = “true”,
“dynamic_partition.time_unit” = “MONTH”,
“dynamic_partition.time_zone” = “Asia/Shanghai”,
“dynamic_partition.start” = “-2147483648”,
“dynamic_partition.end” = “3”,
“dynamic_partition.prefix” = “p_”,
“dynamic_partition.buckets” = “11”,
“dynamic_partition.history_partition_num” = “3”,
“dynamic_partition.start_day_of_month” = “1”,
“in_memory” = “false”,
“storage_format” = “DEFAULT”,
“enable_persistent_index” = “true”,
“replicated_storage” = “true”,
“compression” = “LZ4”
);

只insert一条ioutil也会打满吗?

是的,一条一条写io也会满,
实际试过上传文件,正常写io不会满

数据量大和数据量小耗时都是这么久

scheduling delay是什么?另外本身不建议使用HDD,写入太频繁HDD调度会很快到达阈值导致实际写入量不大ioutil打满排队

哦,这是sparkstreaming的ui图,大概就是在队列等待的时间,
但是我也频率也不高啊,有什么可以优化的方向嘛,不然测试环境数据量少,还一直处于堆积状态

测试环境数据量不多的话,可以考虑建表时的BUCKETS调小一些,写入降低些频率30秒一次攒批写入