hive创建目录时相关的几个hdfs中的类:
org.apache.hadoop.hdfs.DistributedFileSystem,FileSystem 的具体实现类 org.apache.hadoop.hdfs.DFSClient,client操作hdfs文件系统的类 org.apache.hadoop.fs.permission.FsPermission 文件权限相关类,主要的方法有getUMask和applyUMask方法
org.apache.hadoop.hdfs.DistributedFileSystem中需要注意的几个方法:
initialize,主要用来初始DFSClient的实例:
@Override public void initialize(URI uri, Configuration conf) throws IOException { super.initialize(uri, conf); setConf(conf); String host = uri.getHost(); if (host == null) { throw new IOException("Incomplete HDFS URI, no host: "+ uri); } this.dfs = new DFSClient(uri, conf, statistics); this.uri = URI.create(uri.getScheme()+"://"+uri.getAuthority()); this.workingDir = getHomeDirectory(); }
mkdir用来创建一个目录,mkdirs用来创建多个目录(类似于mkdir -p):
public boolean mkdir(Path f, FsPermission permission) throws IOException { statistics.incrementWriteOps(1); return dfs.mkdirs(getPathName(f), permission, false); } public boolean mkdirs(Path f, FsPermission permission) throws IOException { statistics.incrementWriteOps(1); return dfs.mkdirs(getPathName(f), permission, true); }
两者最终调用的都是DFSClient.mkdirs方法,org.apache.hadoop.hdfs.DFSClient的mkdirs方法:
final Conf dfsClientConf; ... public boolean mkdirs(String src, FsPermission permission, boolean createParent) throws IOException { if (permission == null) { //如果传入的权限为null permission = FsPermission.getDefault(); } FsPermission masked = permission.applyUMask(dfsClientConf.uMask); return primitiveMkdir(src, masked, createParent); //调用primitiveMkdir方法 }
这里需要注意 FsPermission.getDefault方法和Conf.uMask属性(Conf是DFSClient的内部类,主要用来设置默认配置)
Conf.uMask属性:
uMask = FsPermission.getUMask(conf); //由getUMask获取
getUMask方法:
public static final String DEPRECATED_UMASK_LABEL = "dfs.umask"; public static final String UMASK_LABEL = CommonConfigurationKeys.FS_PERMISSIONS_UMASK_KEY; //fs.permissions.umask-mode public static final int DEFAULT_UMASK = CommonConfigurationKeys.FS_PERMISSIONS_UMASK_DEFAULT; //0022 public static FsPermission getUMask(Configuration conf) { int umask = DEFAULT_UMASK; if(conf != null) { String confUmask = conf.get(UMASK_LABEL); int oldUmask = conf.getInt(DEPRECATED_UMASK_LABEL, Integer.MIN_VALUE); //老的配置项:dfs.umask,默认值为Integer.MIN_VALUE(-2147483648) try { if(confUmask != null) { //如果设置了fs.permissions.umask-mode,则按这个umask,否则为默认的umask(0022) umask = new UmaskParser(confUmask).getUMask(); } } catch(IllegalArgumentException iae) { // Provide more explanation for user-facing message String type = iae instanceof NumberFormatException ? "decimal" : "octal or symbolic"; String error = "Unable to parse configuration " + UMASK_LABEL + " with value " + confUmask + " as " + type + " umask."; LOG.warn(error); // If oldUmask is not set, then throw the exception if (oldUmask == Integer.MIN_VALUE) { throw new IllegalArgumentException(error); } } if(oldUmask != Integer.MIN_VALUE) { //如果手动设置了老的配置项dfs.umask if (umask != oldUmask) { //并且dfs.umask的值不等于0022 LOG.warn(DEPRECATED_UMASK_LABEL + " configuration key is deprecated. " + "Convert to " + UMASK_LABEL + ", using octal or symbolic umask " + "specifications."); // Old and new umask values do not match - Use old umask umask = oldUmask; //umask为默认值0022 } } } return new FsPermission((short)umask); }
在hive中创建hdfs的目录有两种方法
1)通过Utilities的createDirsWithPermission方法,这种方法会重设fs.permissions.umask-mode
2)直接通过DistributedFileSystem的mkdirs方法创建
两者最终都是调用了DFSClient的mkdirs方法,不同的是调用Utilities.createDirsWithPermission创建的目录权限在proxy时权限有可能是777(因为手动设置了权限为777),
比如:
Context类的构造函数中创建临时文件目录通过Context.getMRScratchDir调getLocalScratchDir(local job)或getScratchDir(非local job),其中getScratchDir中调用Utilities.createDirsWithPermission方法调用目录
public static boolean createDirsWithPermission(Configuration conf, Path mkdirPath, FsPermission fsPermission, boolean recursive) throws IOException { String origUmask = null; LOG.warn("Create dirs " + mkdirPath + " with permission " + fsPermission + " recursive " + recursive); if (recursive) { //如果recursive为true,设置fs.permissions.umask-mode为000, //默认情况下recursive = SessionState.get().isHiveServerQuery() &&conf.getBoolean(HiveConf.ConfVars.HIVE_SERVER2_ENABLE_DOAS.varname,HiveConf.ConfVars.HIVE_SERVER2_ENABLE_DOAS.defaultBoolVal); //即时来自hiveserver的请求,并且开启了doas,这里还会把权限设置为777(这里我增加了一个逻辑,如果设置了proxy,recursive也为true) /** boolean recursive = false; if (SessionState.get() != null) { recursive = (SessionState.get().isHiveServerQuery() && conf.getBoolean(HiveConf.ConfVars.HIVE_SERVER2_ENABLE_DOAS.varname, HiveConf.ConfVars.HIVE_SERVER2_ENABLE_DOAS.defaultBoolVal))||(HiveConf.getBoolVar(conf,HiveConf.ConfVars.HIVE_USE_CUSTOM_PROXY)); fsPermission = new FsPermission((short)00777); } */ origUmask = conf.get("fs.permissions.umask-mode"); conf.set("fs.permissions.umask-mode", "000"); } FileSystem fs = ShimLoader.getHadoopShims().getNonCachedFileSystem(mkdirPath.toUri(), conf); //这里是DFSClient的实例 boolean retval = false; try { retval = fs.mkdirs(mkdirPath, fsPermission); resetConfAndCloseFS(conf, recursive, origUmask, fs); } catch (IOException ioe) { try { resetConfAndCloseFS(conf, recursive, origUmask, fs); //调用resetConfAndCloseFS,reset fs.permissions.umask-mode的设置 } catch (IOException e) { // do nothing - double failure } } return retval; }
resetConfAndCloseFS方法用来重设fs.permissions.umask-mode的设置,这样如果后面创建目录不是使用Utilities.createDirsWithPermission就会使用这个重设的配置
private static void resetConfAndCloseFS (Configuration conf, boolean unsetUmask, String origUmask, FileSystem fs) throws IOException { if (unsetUmask) { //unsetUmask为true,即recursive为true的话,需要重设fs.permissions.umask-mode if (origUmask != null) { //如果有设置项的话,使用设置项 conf.set("fs.permissions.umask-mode", origUmask); } else { conf.unset("fs.permissions.umask-mode"); //这里虽然可以unset,后面会有默认值 } } fs.close(); }
通过查看DFSClient的源码,发现在DFSClient的构造函数中会初始化ugi的信息,默认为当前用户
final UserGroupInformation ugi; ... this.ugi = UserGroupInformation.getCurrentUser(); 如果更改成proxy用户,通过运行hadoop fs -mkdir测试,发现生成的文件目录属主还是当前登录用户 更改DFSClient的构造方法: //this.ugi = UserGroupInformation.getCurrentUser(); if(conf.getBoolean("use.custom.proxy",false)){ this.ugi = UserGroupInformation.createRemoteUser(conf.get("custom.proxy.user")); }else{ this.ugi = UserGroupInformation.getCurrentUser(); }
在hdfs-site.xml配置中增加:
dfs配置中增加:
<property> <name>use.custom.proxy</name> <value>true</value> </property> <property> <name>custom.proxy.user</name> <value>ericni</value> </propertMySQL主主互备模式(Keepalived) 阅读原文»MySQL主主互备模式(Keepalived)MySQL双主-高可用
单台数据库实例安装
请参考:二进制包安装MySQL
资源规划
主机名 os 版本 MySQL 版本 主机 IP MySQL VIP db01.lyk.com
centos 6.4
mysql-5.6.21-linux-glibc2.5-x86_64
172.31.30.12
172.31.30.222
db02.lyk.com
centos 6.4 mysql-5.6.21-linux-glibc2.5-x86_64 172.31.30.11 3.修改MySQL配置文件
修改DB01的配置文件:
#在[mysqld]添加如下内容#
server-
id
=100
log-bin=
/usr/local/mysql/data/ttpai-bin
binlog_format=MIXED
#非必需
relay-log=
/usr/local/mysql/data/ttpai-relay-bin
binlog-ignore-db=mysql
binlog-ignore-db=
test
binlog-ignore-db=information_schema
binlog-ignore-db=performance_schema
replicate-wild-ignore-table=mysql.%
replicate-wild-ignore-table=
test
.%
replicate-wild-ignore-table=information_schema.%
replicate-wild-ignore-table=performance_schema.%
修改DB02的配置文件:
#在[mysqld]添加如下内容#
server-
id
=110
log-bin=
/usr/local/mysql/data/ttpai-bin
binlog_format=MIXED
#非必需
relay-log=
/usr/local/mysql/data/ttpai-relay-bin
binlog-ignore-db=mysql
binlog-ignore-db=
test
binlog-ignore-db=information_schema
binlog-ignore-db=performance_schema
replicate-wild-ignore-table=mysql.%
replicate-wild-ignore-table=
test
.%
replicate-wild-ignore-table=information_schema.%
replicate-wild-ignore-table=performance_schema.%
4.手动同步数据库
如果DB01上已经有MySQL数据,那么执行主主互备之前,需要将DB01和DB02上的两个MySQL的数据保持同步,首先在DB01上备份MySQL数据,执行如下SQL:
mysql>FLUSHTABLES
WITH
READ
LOCK;
在不退出终端的情况下(推出锁失效),再开启一个session,直接打包MySQL的数据文件或者mysqldump工具导出:
cd
/usr/local/mysql/
tar
zcvfdata.
tar
.gzdata/
将data.tar.gz 传输到DB02,依次重启DB01和DB02。
其实,可以在不执行READ LOCK语句,直接使用mysqldump语句备份,最起码个人测试是数据不会丢失或者说出现同步异常。使用如下命令:
mysqldump--default-character-
set
=gbk--opt--triggers-R-E--hex-blob--single-transaction--master-data=2ttpai>ttpai.sql
其中-
订阅: 博文评论 (Atom)
没有评论:
发表评论