hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [hadoop] ayushtkn commented on a change in pull request #2305: HDFS-15578: Fix the rename issues with fallback fs enabled
Date Wed, 16 Sep 2020 17:18:03 GMT

ayushtkn commented on a change in pull request #2305:
URL: https://github.com/apache/hadoop/pull/2305#discussion_r489589949

File path: hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestViewDistributedFileSystemWithMountLinks.java
@@ -61,4 +64,55 @@ public void testCreateOnRoot() throws Exception {
   public void testMountLinkWithNonExistentLink() throws Exception {
+  @Test
+  public void testRenameOnInternalDirWithFallback() throws Exception {
+    Configuration conf = getConf();
+    URI defaultFSURI =
+        URI.create(conf.get(CommonConfigurationKeys.FS_DEFAULT_NAME_KEY));
+    final Path hdfsTargetPath1 = new Path(defaultFSURI + "/HDFSUser");
+    final Path hdfsTargetPath2 = new Path(defaultFSURI + "/NewHDFSUser/next");
+    ViewFsTestSetup.addMountLinksToConf(defaultFSURI.getAuthority(),
+        new String[] {"/HDFSUser", "/NewHDFSUser/next"},
+        new String[] {hdfsTargetPath1.toUri().toString(),
+            hdfsTargetPath2.toUri().toString()}, conf);
+    //Making sure parent dir structure as mount points available in fallback.
+    try (DistributedFileSystem dfs = new DistributedFileSystem()) {
+      dfs.initialize(defaultFSURI, conf);
+      dfs.mkdirs(hdfsTargetPath1);
+      dfs.mkdirs(hdfsTargetPath2);
+    }
+    try (FileSystem fs = FileSystem.get(conf)) {
+      Path src = new Path("/newFileOnRoot");
+      Path dst = new Path("/newFileOnRoot1");
+      fs.create(src).close();
+      verifyRename(fs, src, dst);
+      src = new Path("/newFileOnRoot1");
+      dst = new Path("/NewHDFSUser/newFileOnRoot");
+      fs.mkdirs(dst.getParent());
+      verifyRename(fs, src, dst);
+      src = new Path("/NewHDFSUser/newFileOnRoot");
+      dst = new Path("/NewHDFSUser/newFileOnRoot1");
+      verifyRename(fs, src, dst);
+      src = new Path("/NewHDFSUser/newFileOnRoot1");
+      dst = new Path("/newFileOnRoot");
+      verifyRename(fs, src, dst);
+      src = new Path("/HDFSUser/newFileOnRoot1");
+      dst = new Path("/HDFSUser/newFileOnRoot");
+      fs.create(src).close();
+      verifyRename(fs, src, dst);
+    }
+  }
+  private void verifyRename(FileSystem fs, Path src, Path dst)
+      throws IOException {
+    fs.rename(src, dst);
+    Assert.assertFalse(fs.exists(src));
+    Assert.assertTrue(fs.exists(dst));
+  }

Review comment:
       Thanx @umamaheswararao for the update.
   Regarding the Case 2: When the same directory structure isn't available in the fallback.
   In `ViewFs` I think this was handled and `createParent` was explicitly made `true` always.
It would be just for rename this compulsion would be there. 
   Considering a mount entry like -- `/mount/sub1/sub2` --> `/nsPath`
   if someone calls rename with `dst` as `/mount/sub1/renameFile` will fail, but if he calls
create `/mount/sub1/createFile` without `createParent` it would pass and this `create` call
will create the internal directory structure as well. So, now again the user calls the same
rename command, it would succeed. Same for `mkdir` with `createParent` as `false`
   This would be little intermittent behavior for the end user, one API behaving differently.
   Secondly creating the same directory structure at `fallback` just for `rename` to work
doesn't seems feasible, It would be too many empty directories, increasing the number of inodes
at NN. IIRC something like this, to create empty directories for mount entries in case of
RBF was discussed for some issue recently, and UBER folks had concerns with inode numbers
going high due to empty directories.
   I think we should explicitly take care of this in `rename` as well, May be in non-atomic
way only? Later we might find a better way, Maybe adding one more flag to `rename2` and argument
to `rename` for `createParent` in a follow up.

This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:

To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org

View raw message