qfs集群安装之一 meta server安装
东门晓博
下载程序
下载centos 6的版本：
https://s3.amazonaws.com/quantcast-qfs/qfs-centos-6.7-master-x86_64.tgz
然后解压
Meta Server
负责保存所有chunk的位置，协调chunk的复制和恢复。
配置文件
需要创建一个配置文件，文件名为meta_server.prp，这是简要配置
# port used by clients to connect to the metaserver
metaServer.clientPort = 20000
# port used by chunk servers to connect to the metaserver
metaServer.chunkServerPort = 30000
# chunk placement groups by IP address or first three octets
# metaServer.rackPrefixes =
# create new file system if no transaction logs or checkpoints are found
metaServer.createEmptyFs = 1
# location to write transaction logs
metaServer.logDir = /your-path/qfs_logs
# location to write checkpoints, this needs be pruned periodically
metaServer.cpDir = /your-path/qfs_checkpoint
# unique cluster id
metaServer.clusterKey = dean-qfs
完整配置从这里下载。
注意完整配置文件分成两部分，一部分是静态配置，必须重启服务才生效，一部份是动态配置，可以通过发SIGHUP信号给meta server进程生效。两部分中间有分割说明
# ==============================================================================
# The parameters below this line can be changed at runtime by editing the
# configuration file and sending meta server process HUP signal.
# Note that to restore parameter to default at run time the default value must
# be explicitly specified in the configuration file. In other words commenting
# out the parameter will not have any effect until restart.
创建qfs用户
创建qfs用户，adduser qfs，然后将之前配置的所有目录都改成qfs为owner。初始化
切换到qfs用户，运行下面的命令
/your-path/qfs/bin/metaserver -c /your-path/qfs_conf/meta_server.prp10-13-2016 02:06:13.083 ERROR - (UserAndGroup.cc:648) getpwent duplicate user id: 0 name: a cur: root group: 0 cur: 0
10-13-2016 02:06:13.083 ERROR - (UserAndGroup.cc:648) getpwent duplicate user id: 0 name: b cur: root group: 0 cur: 0这样不能直接删除用户，需要(用root帐号)先到/etc/passwd中将a和b的uid改成两个新值，然后再用userdel 命令删除a和b
现在再试运行一次命令
继续报错：
10-13-2016 02:10:20.687 ERROR - (TcpSocket.cc:575) 0.0.0.0:30000 Address already in use 98
10-13-2016 02:10:20.687 ERROR - (Acceptor.cc:102) failed to bind to: 30000 error: Address already in use 98
Loading key metaServer.clientPort with value 20000
Loading key metaServer.chunkServerPort with value 40000
Loading key metaServer.rackPrefixes with value
Loading key metaServer.createEmptyFs with value 1
Loading key metaServer.logDir with value /your-path/qfs_logs
Loading key metaServer.cpDir with value /data/slot0/qfs_checkpoint
Loading key metaServer.clusterKey with value dean-qfs
Loading key metaServer.clusterKey with value dean-qfs
10-13-2016 02:15:47.171 INFO - (nofilelimit.cc:82) max # of open files: 600000
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:460) meta server client listner: 20000
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:479) meta server chunk server listener: 40000
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:488) path->fid cache disabled
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:351) min chunk servers that should connect: 1
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:359) min. # of replicas per file: 1
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:520) hard limits: open files: 600000 chunk servers: 4096 clients: 595648
10-13-2016 02:15:47.173 INFO - (LayoutManager.cc:1728) rack: prefix: id: 10
10-13-2016 02:15:47.173 INFO - (LayoutManager.cc:2060) max. response size: 268435456 minIoBufferBytesToProcessRequest: 278921216
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:541) creating empty file system
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:706) creating new empty file system
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:729) replaying logs
10-13-2016 02:15:47.173 INFO - (metaserver_main.cc:747) updating space utilization启动服务
将刚才的命令去掉 -c，保存为脚本文件，然后运行脚本：(-c只能用一次，以后集群正常运转的时候使用-c会丢失数据）
$ ./start-qfs-meta-server.sh
Loading key metaServer.clientPort with value 20000
Loading key metaServer.chunkServerPort with value 40000
Loading key metaServer.rackPrefixes with value
Loading key metaServer.createEmptyFs with value 1
Loading key metaServer.logDir with value /your-path/qfs_logs
Loading key metaServer.cpDir with value /data/slot0/qfs_checkpoint
Loading key metaServer.clusterKey with value dean-qfs
Loading key metaServer.clusterKey with value dean-qfs
10-13-2016 02:18:26.080 INFO - (nofilelimit.cc:82) max # of open files: 600000
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:460) meta server client listner: 20000
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:479) meta server chunk server listener: 40000
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:488) path->fid cache disabled
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:351) min chunk servers that should connect: 1
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:359) min. # of replicas per file: 1
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:520) hard limits: open files: 600000 chunk servers: 4096 clients: 595648
10-13-2016 02:18:26.081 INFO - (LayoutManager.cc:1728) rack: prefix: id: 10
10-13-2016 02:18:26.081 INFO - (LayoutManager.cc:2060) max. response size: 268435456 minIoBufferBytesToProcessRequest: 278921216
10-13-2016 02:18:26.081 INFO - (metaserver_main.cc:541) starting metaserver
10-13-2016 02:18:26.082 INFO - (Restorer.cc:99) restoring from checkpoint of 2016-10-12T18:15:47.174078Z
10-13-2016 02:18:26.082 INFO - (Replay.cc:73) open log file: /your-path/qfs_logs/log.0
10-13-2016 02:18:26.082 INFO - (Restorer.cc:99) restoring from checkpoint of 2016-10-12T18:15:47.174088Z
10-13-2016 02:18:26.082 INFO - (metaserver_main.cc:729) replaying logs
10-13-2016 02:18:26.082 INFO - (Replay.cc:73) open log file: /your-path/qfs_logs/log.0
10-13-2016 02:18:26.082 INFO - (Replay.cc:750) log time: 2016-10-12T18:15:47.174020Z
10-13-2016 02:18:26.082 INFO - (metaserver_main.cc:747) updating space utilization
10-13-2016 02:18:26.082 INFO - (Logger.cc:124) log append: int base: 16 file: /your-path/qfs_logs/log.0
10-13-2016 02:18:26.082 INFO - (metaserver_main.cc:548) start servicing
10-13-2016 02:18:27.083 INFO - (NetDispatch.cc:667) ===request=counters: 1476296307083673 9998 3999 1 0 9 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 9 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
10-13-2016 02:18:27.083 INFO - (NetDispatch.cc:684) ===rusage=self: utime: 9998 stime: 3999 maxrss: 6444 ixrss: 0 idrss: 0 isrss: 0 minflt: 1716 majflt: 0 inblock: 0 oublock: 16 msgsnd: 0 msgrcv: 0 nsignals: 0 nvscw: 18 nivscw: 4
10-13-2016 02:18:27.083 INFO - (NetDispatch.cc:689) ===rusage=chidren: utime: 0 stime: 0 maxrss: 0 ixrss: 0 idrss: 0 isrss: 0 minflt: 0 majflt: 0 inblock: 0 oublock: 0 msgsnd: 0 msgrcv: 0 nsignals: 0 nvscw: 0 nivscw: 0
10-13-2016 02:18:27.083 INFO - (LayoutManager.cc:9391) Initiating a replication check of all chunks
#!/bin/bash
#-c can be used once, do not use it when have impoartand data in cluster
#/your-path/qfs/bin/metaserver -c /your-path/qfs_conf/meta_server.prp
nohup /your-path/qfs/bin/metaserver /your-path/qfs_conf/meta_server.prp &注意观察nohup.out的内容，都是INFO就算正常。
chunk server的事下篇再说吧。
官方文档
https://github.com/quantcast/qfs/wiki/Deployment-Guide
参考meta server章节
类似资料：