[
https://issues.apache.org/jira/browse/HBASE-20668?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Duo Zhang updated HBASE-20668:
------------------------------
Fix Version/s: 3.0.0-alpha-1
> Avoid permission change if ExportSnapshot's copy fails
> ------------------------------------------------------
>
> Key: HBASE-20668
> URL: https://issues.apache.org/jira/browse/HBASE-20668
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Priority: Major
> Fix For: 3.0.0-alpha-1, 2.1.0
>
> Attachments: 20668.v1.txt, 20668.v2.txt
>
>
> The exception from FileSystem operation in the finally block of
> ExportSnapshot#doWork may hide the real exception from FileUtil.copy call.
> I was debugging the following error [~romil.choksi] saw during testing
> ExportSnapshot :
> {code:java}
> 2018-06-01 02:40:52,363|INFO|MainThread|machine.py:167 -
> run()||GUID=1cacb7bc-f7cc-4710-82e0-4a4513f0c1f9|2018-06-01 02:40:52,358
> ERROR [main] util.AbstractHBaseTool: Error running command-line tool
> 2018-06-01 02:40:52,363|INFO|MainThread|machine.py:167 -
> run()||GUID=1cacb7bc-f7cc-4710-82e0-4a4513f0c1f9|java.io.FileNotFoundException:
> Directory/File does not exist /apps/
> hbase/data/.hbase-snapshot/.tmp/snapshot_table_334546
> 2018-06-01 02:40:52,364|INFO|MainThread|machine.py:167 -
> run()||GUID=1cacb7bc-f7cc-4710-82e0-4a4513f0c1f9|at
> org.apache.hadoop.hdfs.server.namenode.FSDirectory.
> checkOwner(FSDirectory.java:1777)
> 2018-06-01 02:40:52,364|INFO|MainThread|machine.py:167 -
> run()||GUID=1cacb7bc-f7cc-4710-82e0-4a4513f0c1f9|at
> org.apache.hadoop.hdfs.server.namenode.FSDirAttrOp.
> setOwner(FSDirAttrOp.java:82)
> {code}
> Here is corresponding code (with extra log added):
> {code:java}
> try {
> LOG.info("Copy Snapshot Manifest from " + snapshotDir + " to " +
> initialOutputSnapshotDir);
> boolean ret = FileUtil.copy(inputFs, snapshotDir, outputFs,
> initialOutputSnapshotDir, false,
> false, conf);
> LOG.info("return val = " + ret);
> } catch (IOException e) {
> LOG.warn("Failed to copy the snapshot directory: from=" +
> snapshotDir + " to=" + initialOutputSnapshotDir, e);
> throw new ExportSnapshotException("Failed to copy the snapshot
> directory: from=" +
> snapshotDir + " to=" + initialOutputSnapshotDir, e);
> } finally {
> if (filesUser != null || filesGroup != null) {
> LOG.warn((filesUser == null ? "" : "Change the owner of " +
> needSetOwnerDir + " to "
> + filesUser)
> + (filesGroup == null ? "" : ", Change the group of " +
> needSetOwnerDir + " to "
> + filesGroup));
> setOwner(outputFs, needSetOwnerDir, filesUser, filesGroup, true);
> }
> {code}
> "return val = " was not seen in rerun of the test.
> This is what the additional log revealed:
> {code:java}
> 2018-06-01 09:22:54,247|INFO|MainThread|machine.py:167 -
> run()||GUID=3961d249-9981-429d-81a8-39c7df53cf58|2018-06-01 09:22:54,241 WARN
> [main] snapshot.ExportSnapshot: Failed to copy the snapshot directory:
> from=hdfs://ns1/apps/hbase/data/.hbase-snapshot/snapshot_table_157842
> to=hdfs://ns3/apps/hbase/data/.hbase-snapshot/.tmp/snapshot_table_157842
> 2018-06-01 09:22:54,248|INFO|MainThread|machine.py:167 -
> run()||GUID=3961d249-9981-429d-81a8-39c7df53cf58|org.apache.hadoop.security.AccessControlException:
> Permission denied: user=hbase, access=WRITE,
> inode="/apps/hbase/data/.hbase-snapshot/.tmp":hrt_qa:hadoop:dr----x-wT
> 2018-06-01 09:22:54,248|INFO|MainThread|machine.py:167 -
> run()||GUID=3961d249-9981-429d-81a8-39c7df53cf58|at
> org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.
> check(FSPermissionChecker.java:399)
> 2018-06-01 09:22:54,249|INFO|MainThread|machine.py:167 -
> run()||GUID=3961d249-9981-429d-81a8-39c7df53cf58|at
> org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.
> checkPermission(FSPermissionChecker.java:255)
> {code}
> It turned out that the exception from {{setOwner}} call in the finally block
> eclipsed the real exception.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
