问题及背景:同一用户的三家公司的物理集群合并,合并后用dataspace+kerberos控制不同公司对集群资料的访问权限,三家公司分别使用独立的kerberos票据访问,特定的namespace,而生产环境部署的kylin-2.0/2.1只能保存cuboid到hbase 的 default namespace。
解决的过程及方案:
1、经多方查找发现, Kylin源码github项目,最新的master分支已经支持更改存储的hbase的默认namespace,但是官方对外发布的最新安装包才更新到Kylin-2.1版本,且最新源码并没有经过生产的验证,不敢冒然部署,于是想办法在现有的版本上重新编译源码、打包。
2、经多方搜索确认更改hbase namespace的commit是kylin-2846,于是查找相关实现的细节。
解决思路:
1、首先按照github上提供的修改方法更改了https://github.com/apache/kylin/commit/ffcbdb7f4288fa6011f018163bdce9602415cf10
中的五个类,重新编译打包,执行到convert cuboid Data to HFile这一步的时候抛出异常,找不到Hbase表
org.apache.hadoop.hbase.TableNotFoundException:ISW:KYLIN_F3KD3ZC1LU
at org.apache.hadoop.hbase.client.HBaseAdmin.getTableDescriptor(HBaseAdmin.java:575)
而这时,我在hbase shell指定的namespace查看,已经生成了hbase表,为什么会找不到呢?继续查找原因
2、继续找Kylin-2846相关的内容,发现有人遇到同样的问题,地址:https://issues.apache.org/jira/browse/KYLIN-2846
原来github上给出的代码少了很关键的一步:
---.../main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java | 2 +-1 file changed, 1 insertion(+), 1 deletion(-)diff --git a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
index 9f9b382..7a583ca 100644
--- a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
+++ b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
@@ -92,7 +92,7 @@ public class CubeHFileJob extends AbstractHadoopJob {attachCubeMetadata(cube, job.getConfiguration());Configuration hbaseConf = HBaseConfiguration.create(getConf());
- HTable htable = new HTable(hbaseConf, getOptionValue(OPTION_HTABLE_NAME).toUpperCase());
+ HTable htable = new HTable(hbaseConf, getOptionValue(OPTION_HTABLE_NAME));// Automatic config !HFileOutputFormat.configureIncrementalLoad(job, htable);
--
1.9.1
具体的修改步骤如下:From 004e1bcb983442a1a2c69f7d1d6c09dc29d4301e Mon Sep 17 00:00:00 2001
From: Liu Shaohui <liushaohui@xiaomi.com>
Date: Tue, 5 Sep 2017 15:50:29 +0800
Subject: [PATCH] Add a config of hbase namespace for cube storage---.../src/main/java/org/apache/kylin/common/KylinConfigBase.java | 4 ++++core-common/src/main/resources/kylin-defaults.properties | 3 +++core-cube/src/main/java/org/apache/kylin/cube/CubeManager.java | 2 ++.../main/java/org/apache/kylin/storage/hbase/steps/BulkLoadJob.java | 2 +-.../main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java | 2 +-5 files changed, 11 insertions(+), 2 deletions(-)diff --git a/core-common/src/main/java/org/apache/kylin/common/KylinConfigBase.java b/core-common/src/main/java/org/apache/kylin/common/KylinConfigBase.java
index 749b515..3f53023 100644
--- a/core-common/src/main/java/org/apache/kylin/common/KylinConfigBase.java
+++ b/core-common/src/main/java/org/apache/kylin/common/KylinConfigBase.java
@@ -712,6 +712,10 @@ abstract public class KylinConfigBase implements Serializable {return StorageURL.valueOf(url);}+ public String getHBaseStorageNameSpace() {
+ return getOptional("kylin.storage.hbase.namespace", "default");
+ }
+public String getHBaseClusterFs() {return getOptional("kylin.storage.hbase.cluster-fs", "");}
diff --git a/core-common/src/main/resources/kylin-defaults.properties b/core-common/src/main/resources/kylin-defaults.properties
index a3277ce..34fb5a0 100644
--- a/core-common/src/main/resources/kylin-defaults.properties
+++ b/core-common/src/main/resources/kylin-defaults.properties
@@ -65,6 +65,9 @@ kylin.source.hive.redistribute-flat-table=true# The storage for final cube file in hbasekylin.storage.url=hbase+# The namespace for hbase storage
+kylin.storage.hbase.namespace=default
+# Compression codec for htable, valid value [none, snappy, lzo, gzip, lz4]kylin.storage.hbase.compression-codec=nonediff --git a/core-cube/src/main/java/org/apache/kylin/cube/CubeManager.java b/core-cube/src/main/java/org/apache/kylin/cube/CubeManager.java
index 043993c..484b977 100644
--- a/core-cube/src/main/java/org/apache/kylin/cube/CubeManager.java
+++ b/core-cube/src/main/java/org/apache/kylin/cube/CubeManager.java
@@ -684,11 +684,13 @@ public class CubeManager implements IRealizationProvider {}private String generateStorageLocation() {
+ String namespace = config.getHBaseStorageNameSpace();String namePrefix = IRealizationConstants.CubeHbaseStorageLocationPrefix;String tableName = "";Random ran = new Random();do {StringBuffer sb = new StringBuffer();
+ sb.append(namespace).append(":");sb.append(namePrefix);for (int i = 0; i < HBASE_TABLE_LENGTH; i++) {sb.append(ALPHA_NUM.charAt(ran.nextInt(ALPHA_NUM.length())));
diff --git a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/BulkLoadJob.java b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/BulkLoadJob.java
index 30616c5..d1caccb 100644
--- a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/BulkLoadJob.java
+++ b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/BulkLoadJob.java
@@ -47,7 +47,7 @@ public class BulkLoadJob extends AbstractHadoopJob {options.addOption(OPTION_CUBE_NAME);parseOptions(options, args);- String tableName = getOptionValue(OPTION_HTABLE_NAME).toUpperCase();
+ String tableName = getOptionValue(OPTION_HTABLE_NAME);// e.g// /tmp/kylin-3f150b00-3332-41ca-9d3d-652f67f044d7/test_kylin_cube_with_slr_ready_2_segments/hfile/// end with "/"
diff --git a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
index 1a624c4..9f9b382 100644
--- a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
+++ b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
@@ -68,7 +68,7 @@ public class CubeHFileJob extends AbstractHadoopJob {Path partitionFilePath = new Path(getOptionValue(OPTION_PARTITION_FILE_PATH));Path output = new Path(getOptionValue(OPTION_OUTPUT_PATH));
- String cubeName = getOptionValue(OPTION_CUBE_NAME).toUpperCase();
+ String cubeName = getOptionValue(OPTION_CUBE_NAME);CubeManager cubeMgr = CubeManager.getInstance(KylinConfig.getInstanceFromEnv());--
1.9.1
2、、、From 677400ee7b4cf5be0ddc9b595d80e02a0ba87c12 Mon Sep 17 00:00:00 2001
From: Liu Shaohui <liushaohui@xiaomi.com>
Date: Fri, 15 Sep 2017 14:34:18 +0800
Subject: [PATCH] Addium for KYLIN-2846---.../main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java | 2 +-1 file changed, 1 insertion(+), 1 deletion(-)diff --git a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
index 9f9b382..7a583ca 100644
--- a/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
+++ b/storage-hbase/src/main/java/org/apache/kylin/storage/hbase/steps/CubeHFileJob.java
@@ -92,7 +92,7 @@ public class CubeHFileJob extends AbstractHadoopJob {attachCubeMetadata(cube, job.getConfiguration());Configuration hbaseConf = HBaseConfiguration.create(getConf());
- HTable htable = new HTable(hbaseConf, getOptionValue(OPTION_HTABLE_NAME).toUpperCase());
+ HTable htable = new HTable(hbaseConf, getOptionValue(OPTION_HTABLE_NAME));// Automatic config !HFileOutputFormat.configureIncrementalLoad(job, htable);
--
1.9.1
3、clean htablediff --git a/server-base/src/main/java/org/apache/kylin/rest/job/StorageCleanJobHbaseUtil.java b/server-base/src/main/java/org/apache/kylin/rest/job/StorageCleanJobHbaseUtil.java
index 3728ea1..2d95542 100644
--- a/server-base/src/main/java/org/apache/kylin/rest/job/StorageCleanJobHbaseUtil.java
+++ b/server-base/src/main/java/org/apache/kylin/rest/job/StorageCleanJobHbaseUtil.java
@@ -46,15 +46,28 @@ public class StorageCleanJobHbaseUtil {public static void cleanUnusedHBaseTables(boolean delete, int deleteTimeout) throws IOException {Configuration conf = HBaseConfiguration.create();
- CubeManager cubeMgr = CubeManager.getInstance(KylinConfig.getInstanceFromEnv());
+ KylinConfig config = KylinConfig.getInstanceFromEnv();
+ CubeManager cubeMgr = CubeManager.getInstance(config);// get all kylin hbase tablestry (HBaseAdmin hbaseAdmin = new HBaseAdmin(conf)) {
- String tableNamePrefix = IRealizationConstants.SharedHbaseStorageLocationPrefix;
+ String namespace = config.getHBaseStorageNameSpace();
+ StringBuffer sb = new StringBuffer();
+ String tableNamePrefix = null;
+ if(namespace.equals("default") || namespace.equals("")){
+ tableNamePrefix = IRealizationConstants.SharedHbaseStorageLocationPrefix;
+ }else{
+ sb.append(config.getHBaseStorageNameSpace()).append(":");
+ sb.append(IRealizationConstants.SharedHbaseStorageLocationPrefix);
+ tableNamePrefix = sb.toString();
+ }
+HTableDescriptor[] tableDescriptors = hbaseAdmin.listTables(tableNamePrefix + ".*");
+List<String> allTablesNeedToBeDropped = new ArrayList<String>();for (HTableDescriptor desc : tableDescriptors) {
+String host = desc.getValue(IRealizationConstants.HTableTag);
- if (KylinConfig.getInstanceFromEnv().getMetadataUrlPrefix().equalsIgnoreCase(host)) {
+ if (config.getMetadataUrlPrefix().equalsIgnoreCase(host)) {//only take care htables that belongs to self, and created more than 2 daysallTablesNeedToBeDropped.add(desc.getTableName().getNameAsString());}
修改完后,core-common, core-cure, storage-hbase三个目录重新打jar包,覆盖原Kylin安装目录KYLIN_HOME/tomcat/webapps/kylin.war下对应的jar包,重启Kylin,问题解决。