Flink生产配置

1、ssh端口号
修改默认ssh端口号:export FLINK_SSH_OPTS="-p 53742"

2、flink-conf.yaml

#---------------->memory begin<--------------
#flink 1.10 config start .............
taskmanager.memory.managed.fraction:  0.5
taskmanager.memory.jvm-metaspace.size: 256m
#flink 1.10 config stop ..............

#--------------->Failure Rate Restart Strategy<----------------
restart-strategy: failure-rate
restart-strategy.failure-rate.max-failures-per-interval: 5000
restart-strategy.failure-rate.failure-rate-interval: 6 min
restart-strategy.failure-rate.delay: 1 s

#---------------->checkpoint begin<--------------
#enable rocksdb to be the backend
state.backend: rocksdb

#filesystem to store checkpoint meta data (when external checkpoint enabled.)
state.checkpoints.dir: hdfs://{{cluster_name}}/user/flink_bu/xxx-fs-checkpoints/cpk/1_3

#filesystem to store checkpoint data
state.backend.fs.checkpointdir: hdfs://{{cluster_name}}/user/flink_bu/xxx-fs-checkpoints/cpk/1_3

#---------------->HA begin<--------------
#enable HA with zk
high-availability: zookeeper

#zk quorm
high-availability.zookeeper.quorum: {{zk_link}}

#only the directory point will stored in zk, data will be stored in the pointed file system
high-availability.zookeeper.storageDir: hdfs://{{cluster_name}}/user/flink_bu/HA/

#!!!important, this needed be unique per cluster.
high-availability.zookeeper.path.root: /flink_bu

#times container failed can be accepted.
yarn.maximum-failed-containers: 10

#times application failed can be accepted.
yarn.application-attempts: 10

# akka config
heartbeat.timeout: 180000

blob.service.cleanup.interval: 1800

akka.tcp.timeout: 60 s

akka.lookup.timeout: 60 s

akka.ask.timeout: 100 s

akka.client.timeout: 600 s

akka.framesize: 1048576000b
#--------------->blobStore begin<---------------
#
#only useful for non HA model
blob.storage.directory: /tmp

#--------------->Defines whether user-jars are included in the system class path for per-job-clusters as well as their positioning in the path<---------------------
yarn.per-job-cluster.include-user-jar: FIRST




state.backend.rocksdb.metrics.column-family-as-variable: true

state.backend.rocksdb.compaction.level.max-size-level-base: 512M
state.backend.rocksdb.block.cache-size: 512M
state.backend.rocksdb.compaction.level.target-file-size-base: 32M
state.backend.rocksdb.compaction.level.use-dynamic-size: true
state.backend.rocksdb.thread.num: 2
state.backend.rocksdb.writebuffer.size: 128M

security.kerberos.login.use-ticket-cache: true
security.kerberos.login.keytab: {{flink_bu_keytab_file}}
security.kerberos.login.principal: {{flink_bu_principal}}

最后编辑于
©著作权归作者所有,转载或内容合作请联系作者
平台声明:文章内容(如有图片或视频亦包括在内)由作者上传并发布,文章内容仅代表作者本人观点,简书系信息发布平台,仅提供信息存储服务。

推荐阅读更多精彩内容

  • --------"道路是曲折的,前途是光明的。"最近又涉及到了hadoop以及spark的安装,之前课题设计中有要...
    九七学姐阅读 3,092评论 3 5
  • 注:使用VMware+Centos7 安装步骤 安装Centos7虚拟机,参考网上教程,安装时网络模式选择NAT方...
    三丶七阅读 840评论 0 1
  • 肩上悬了死物,我捧着好几本书,在别人的称赞和欣赏中,生吞活剥下几千年文化历史。打了个饱嗝,头脑发胀,眼睛快要胀裂,...
    追一苇渡江阅读 384评论 0 3
  • *最近我成为了一名“令人骄傲”的初中生,我知道初中生活一定不会像小学那么愉快,至少不会像小学那样见义勇为打了个人还...
    矮狸科四阅读 745评论 1 3
  • 【@IT·互联网专题】是简书官方最大的专题之一,目前在简书中拥有超过57W+粉丝,收录了接近 30,000 篇IT...
    向右奔跑阅读 1,382评论 3 10