【知识图谱系列】- ongdb因果集群部署

一、安装包准备

• 下载ongdb-3.5.22企业稳定版.zip，并将安装包分别上传相关待部署服务器。
• 解压中文文件，unzip -O CP936 ongdb-3.5.22企业稳定版.zip 。
• 进入[安装包]，根据实际需求选择解压unix版本或者windows版本。
• cd ongdb-enterprise-3.5.22/conf
• vim neo4j.conf

二、配置文件修改

以下为集群用到的配置，无具体说明的可直接删除“#”，将该行打开。

dbms.security.auth_enabled=false  # 控制数据库是否免密登录
dbms.memory.heap.initial_size=512m  # java heap初始值 (24g)
dbms.memory.heap.max_size=512m  # java heap最大值，一般不要超过可用物理内存的80％ (24g)
dbms.memory.pagecache.size=10g  # pagecache大小，官方建议设为：(总内存-dbms.memory.heap.max_size)/2 (28g)
dbms.backup.enabled=true  # 是否允许在线备份
dbms.backup.address=0.0.0.0:6362
dbms.connectors.default_listen_address=0.0.0.0
dbms.connectors.default_advertised_address=localhost  # localhost改为当前服务器ip
dbms.connector.bolt.listen_address=:7687
dbms.connector.http.listen_address=:7474
dbms.connector.https.listen_address=:7473
dbms.logs.http.enabled=true
dbms.logs.gc.enabled=true
dbms.logs.query.enabled=true
dbms.logs.query.threshold=0
dbms.mode=CORE
causal_clustering.minimum_core_cluster_size_at_formation=3  # 集群核心服务器数量 (2)
causal_clustering.minimum_core_cluster_size_at_runtime=3  # 集群运行最小服务器数量 (2)
causal_clustering.initial_discovery_members=localhost:5000,localhost:5001,localhost:5002  # 将集群内所有服务器的ip+端口号写到这里，注意端口号需一致，比如都是5000。
causal_clustering.discovery_listen_address=:5000
causal_clustering.transaction_listen_address=:6000
causal_clustering.raft_listen_address=:7000
causal_clustering.raft_advertised_address=:7000  # 等于号后面补充本机ip:7000
causal_clustering.transaction_advertised_address=:6000  # 等于号后面补充本机ip:6000
causal_clustering.leader_election_timeout=7s  # 设置触发选举时间,可跟据neo4j所运行的网络环境和业务负载情况进行调整 (60s)

在neo4j.conf末尾加上以下内容，主要包括APOC、ALGO等配置

dbms.jvm.additional=-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=5005
# algo|apoc|* plugins
dbms.security.procedures.unrestricted=algo.*,apoc.*,*
apoc.export.file.enabled=true
apoc.import.file.enabled=true
#********************************************************************
# Prometheus monitor
#********************************************************************
# Enable the Prometheus endpoint. Default is 'false'.
metrics.prometheus.enabled=true
# The default is localhost:2004.
metrics.prometheus.endpoint=localhost:2004
#********************************************************************
## ONgDB ElasticSearch Integration
##********************************************************************
## elasticsearch.discovery=true
#elasticsearch.host_name=https://vpc-knowledgegraph-4zarhbj33zcjjkqfo3afso45la.cn-north-1.es.amazonaws.com.cn
#elasticsearch.index_spec=pre_org_cn_node:PRE公司中文名称(name,hcode,pcode,hupdatetime,cluster_id),pre_org_cn_thin_node:PRE中文简称(name),pre_org_en_node:PRE英文名称(name),pre_org_en_thin_node:PRE英文简称(name),pre_org_cn_spell_node:PRE中文拼音简称(name),horg_sharehold:HORGShareHold(name,hcode)

#unsupported.dbms.tx_log.fail_on_corrupted_log_files=false
causal_clustering.discovery_type=LIST

三、部署

• 在各节点服务器上进行安装，修改neo4j.conf等操作。
• 开启服务：./bin/neo4j start
• 监控日志：tail -f ./logs/neo4j.log，若出现类似以下日志内容证明集群部署完成

2021-03-17 01:51:43.746+0000 INFO  Discovering other core members in initial members set: [node-1:5000, node-2:5000]
2021-03-17 01:51:44.000+0000 INFO  Waiting for a total of 2 core members...
2021-03-17 01:51:54.029+0000 INFO  Waiting for a total of 2 core members...
2021-03-17 01:52:00.599+0000 INFO  Neo4j Server shutdown initiated by request
2021-03-17 01:52:04.038+0000 INFO  Waiting for a total of 2 core members...
2021-03-17 01:52:14.047+0000 INFO  Waiting for a total of 2 core members...
nohup: ignoring input
Listening for transport dt_socket at address: 5005
2021-03-17 01:58:02.227+0000 INFO  ======== Neo4j 3.5.22 ========
2021-03-17 01:58:02.238+0000 INFO  Starting...
2021-03-17 01:58:03.984+0000 INFO  Initiating metrics...
2021-03-17 01:58:04.023+0000 INFO  My connection info: [
        Discovery:   listen=0.0.0.0:5000, advertised=node-1:5000,
        Transaction: listen=0.0.0.0:6000, advertised=node-1:6000,
        Raft:        listen=0.0.0.0:7000, advertised=node-1:7000,
        Client Connector Addresses: bolt://node-1:7687,http://node-1:7474,https://node-1:7473

• ./bin/cypher-shell 访问图数据库
• call dbms.cluster.overview();查看集群成员角色信息（确定leader节点，follower节点）

四、遇到的问题

• 部署集群的时候千万不要先启动图数据库，这会导致图数据库进行编译，当配置好集群后，集群间会出现无法通信问题。