[HBASE-21755] RS aborts while performing replication with wal dir on hdfs, root dir on s3

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Duplicate
• Affects Version/s: 1.5.0, 2.1.3
• Component/s: Filesystem Integration, Replication, wal
• Labels: s3
• Assignee: Unassigned
• Reporter: Nihal Jain
• Created: 22/Jan/19 10:02
• Updated: 24/Jun/22 17:52
• Resolved: 29/Jan/19 05:10

Description

Environment/Configuration

• hbase.wal.dir : Configured to be on hdfs
• hbase.rootdir : Configured to be on s3

In replication scenario, while trying to get archived log dir (using method WALEntryStream.java#L314) we get the following exception:

2019-01-21 17:43:55,440 ERROR [RS_REFRESH_PEER-regionserver/host2:22222-1.replicationSource,2.replicationSource.wal-reader.host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1,2] regionserver.ReplicationSource: Unexpected exception in RS_REFRESH_PEER-regionserver/host2:22222-1.replicationSource,2.replicationSource.wal-reader.host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1,2 currentPath=hdfs://dummy_path/hbase/WALs/host2,22222,1548063439555/host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1.1548063492594
java.lang.IllegalArgumentException: Wrong FS: s3a://xxxxxx/hbase128/oldWALs/host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1.1548063492594, expected: hdfs://dummy_path
	at org.apache.hadoop.fs.FileSystem.checkPath(FileSystem.java:781)
	at org.apache.hadoop.hdfs.DistributedFileSystem.getPathName(DistributedFileSystem.java:246)
	at org.apache.hadoop.hdfs.DistributedFileSystem$29.doCall(DistributedFileSystem.java:1622)
	at org.apache.hadoop.hdfs.DistributedFileSystem$29.doCall(DistributedFileSystem.java:1619)
	at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
	at org.apache.hadoop.hdfs.DistributedFileSystem.getFileStatus(DistributedFileSystem.java:1634)
	at org.apache.hadoop.fs.FilterFileSystem.getFileStatus(FilterFileSystem.java:465)
	at org.apache.hadoop.fs.FileSystem.exists(FileSystem.java:1742)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.getArchivedLog(WALEntryStream.java:319)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.resetReader(WALEntryStream.java:404)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.reset(WALEntryStream.java:161)
	at org.apache.hadoop.hbase.replication.regionserver.ReplicationSourceWALReader.run(ReplicationSourceWALReader.java:148)
2019-01-21 17:43:55,444 ERROR [RS_REFRESH_PEER-regionserver/host2:22222-1.replicationSource,2.replicationSource.wal-reader.host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1,2] regionserver.HRegionServer: ***** ABORTING region server host2,22222,1548063439555: Unexpected exception in RS_REFRESH_PEER-regionserver/host2:22222-1.replicationSource,2.replicationSource.wal-reader.host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1,2 *****
java.lang.IllegalArgumentException: Wrong FS: s3a://xxxxxx/hbase128/oldWALs/host2%2C22222%2C1548063439555.host2%2C22222%2C1548063439555.regiongroup-1.1548063492594, expected: hdfs://dummy_path
	at org.apache.hadoop.fs.FileSystem.checkPath(FileSystem.java:781)
	at org.apache.hadoop.hdfs.DistributedFileSystem.getPathName(DistributedFileSystem.java:246)
	at org.apache.hadoop.hdfs.DistributedFileSystem$29.doCall(DistributedFileSystem.java:1622)
	at org.apache.hadoop.hdfs.DistributedFileSystem$29.doCall(DistributedFileSystem.java:1619)
	at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
	at org.apache.hadoop.hdfs.DistributedFileSystem.getFileStatus(DistributedFileSystem.java:1634)
	at org.apache.hadoop.fs.FilterFileSystem.getFileStatus(FilterFileSystem.java:465)
	at org.apache.hadoop.fs.FileSystem.exists(FileSystem.java:1742)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.getArchivedLog(WALEntryStream.java:319)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.resetReader(WALEntryStream.java:404)
	at org.apache.hadoop.hbase.replication.regionserver.WALEntryStream.reset(WALEntryStream.java:161)
	at org.apache.hadoop.hbase.replication.regionserver.ReplicationSourceWALReader.run(ReplicationSourceWALReader.java:148)

 
Current code is:

  private Path getArchivedLog(Path path) throws IOException {
    Path rootDir = FSUtils.getRootDir(conf);

    // Try found the log in old dir
    Path oldLogDir = new Path(rootDir, HConstants.HREGION_OLDLOGDIR_NAME);
    Path archivedLogLocation = new Path(oldLogDir, path.getName());
    if (fs.exists(archivedLogLocation)) {
      LOG.info("Log " + path + " was moved to " + archivedLogLocation);
      return archivedLogLocation;
    }
    .
    .
    .
    return path;
  }

It considers root dir while we should use wal dir.

Issue Links

• duplicates: HBASE-21688 Address WAL filesystem issues
• relates to: HBASE-17437 Support specifying a WAL directory outside of the root directory