Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Not A Problem
-
0.98.4
-
None
-
None
-
Microsoft Azure HDInsight 3.2 cluster with eight hosts
- Ubuntu 12.04.5
- HDP 2.2
- Hadoop 2.6.0
- HBase 0.98.4
Description
In a YARN job, I am creating HFiles with code that has been cribbed from the TableOutputFormat class and bulkloading them with LoadIncrementalHFiles.doBulkLoad.
On other clusters, where fs.defaultFS is set to an hdfs: URI, and my HFiles are placed in an hdfs: URI, the bulkload works as intended.
On this particular cluster, where fs.defaultFS is set to a wasb: URI and my HFiles are placed in a wasb: URI, the bulkload also works as intended.
However, on this same cluster, whenever I place the HFiles in an hdfs: URI, I get the following logs in my application from the HBase client logging repeatedly:
[02 Jun 23:23:26.002](20259/140062246807296) Info2:org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles: Trying to load hfile=hdfs://[my cluster]/[my path] first=\x00\x00\x11\x06 last=;\x8B\x85\x18
[02 Jun 23:23:26.002](20259/140062245754624) Info3:org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles: Going to connect to server region=[my namespace]:[my table],,1464909723920.00eafdb73989312bd8864f0913255f50., hostname=10.0.1.6,16020,1464698786237, seqNum=2 for row with hfile group [
]
[02 Jun 23:23:26.012](20259/140062245754624) Info1:org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles: Attempt to bulk load region containing into table [my namespace]:[my table] with files [family:[my family] path:hdfs://[my cluster]/[my path]] failed. This is recoverable and they will be retried.
[02 Jun 23:23:26.019](20259/140061634982912) Info2:org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles: Split occured while grouping HFiles, retry attempt 2 with 1 files remaining to group or split
And when I look at the appropriate region server's log, I find the following exception repeatedly:
2016-06-02 20:22:50,771 ERROR [B.DefaultRpcServer.handler=22,queue=2,port=16020] access.SecureBulkLoadEndpoint: Failed to complete bulk load
java.io.FileNotFoundException: File doesn't exist: hdfs://[my cluster]/[my path] at org.apache.hadoop.fs.azure.NativeAzureFileSystem.setPermission(NativeAzureFileSystem.java:2192)
at org.apache.hadoop.hbase.security.access.SecureBulkLoadEndpoint$1.run(SecureBulkLoadEndpoint.java:280)
at org.apache.hadoop.hbase.security.access.SecureBulkLoadEndpoint$1.run(SecureBulkLoadEndpoint.java:270)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:356)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1651)
at org.apache.hadoop.hbase.security.access.SecureBulkLoadEndpoint.secureBulkLoadHFiles(SecureBulkLoadEndpoint.java:270)
at org.apache.hadoop.hbase.protobuf.generated.SecureBulkLoadProtos$SecureBulkLoadService.callMethod(SecureBulkLoadProtos.java:4631)
at org.apache.hadoop.hbase.regionserver.HRegion.execService(HRegion.java:6986)
at org.apache.hadoop.hbase.regionserver.HRegionServer.execServiceOnRegion(HRegionServer.java:3456)
at org.apache.hadoop.hbase.regionserver.HRegionServer.execService(HRegionServer.java:3438)
at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:29998)
at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2080)
at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:114)
at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:94)
at java.lang.Thread.run(Thread.java:745)
Looking at the appropriate code in SecureBulkLoadEndpoint.java, I'm finding the following:
public Boolean run() {
FileSystem fs = null;
try {
Configuration conf = env.getConfiguration();
fs = FileSystem.get(conf);
for(Pair<byte[], String> el: familyPaths) {
Path p = new Path(el.getSecond());
Path stageFamily = new Path(bulkToken, Bytes.toString(el.getFirst()));
if(!fs.exists(stageFamily))
}
The call to FileSystem.get is obviously the culprit, since it gets the FileSystem object based on fs.defaultFS, which is suboptimal in this case and other cases where the HFiles are located on a different type of filesystem than the defaultFS.