刘珍 created IOTDB-4700:
-------------------------
Summary: Flush performance is slow
Key: IOTDB-4700
URL: https://issues.apache.org/jira/browse/IOTDB-4700
Project: Apache IoTDB
Issue Type: Improvement
Components: mpp-cluster
Reporter: 刘珍
Assignee: Haiming Zhu
Attachments: image-2022-10-20-17-12-15-915.png,
image-2022-10-20-17-21-01-580.png, image-2022-10-20-17-26-48-200.png,
image-2022-10-20-17-27-46-815.png
m_1019_f2ffb49
3rep，3C3D
{color:#DE350B}stop-datanode.sh , flushing metadata is slow :
At first , there is only one flush thread (pool-4-IoTDB-Flush-1 )
There are three Dataregions , so three flush threads should execute in
parallel{color}
!image-2022-10-20-17-12-15-915.png!
1. 192.168.10.72/73/74 48CPU 384GB
ConfigNode
MAX_HEAP_SIZE="8G"
schema_region_consensus_protocol_class=org.apache.iotdb.consensus.ratis.RatisConsensus
data_region_consensus_protocol_class=org.apache.iotdb.consensus.multileader.MultiLeaderConsensus
schema_replication_factor=3
data_replication_factor=3
DataNode
MAX_HEAP_SIZE="256G"
MAX_DIRECT_MEMORY_SIZE="32G"
max_connection_for_internal_service=300
enable_timed_flush_seq_memtable=true
seq_memtable_flush_interval_in_ms=600000
seq_memtable_flush_check_interval_in_ms=300000
enable_timed_flush_unseq_memtable=true
seq_memtable_flush_interval_in_ms=600000
seq_memtable_flush_check_interval_in_ms=300000
max_waiting_time_when_insert_blocked=3600000
query_timeout_threshold=3600000
2. bm
See the attachment
!image-2022-10-20-17-21-01-580.png!
3. benchmark write completed, ip74 datanode synchronization completed
4. stop ip74 datanode
Execute stop-datanode.sh
before stop datanode:
!image-2022-10-20-17-26-48-200.png!
after stop datanode(take snapshot):
!image-2022-10-20-17-27-46-815.png!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
