Skip to content
Browse files
HBASE-27042. Remove S3Guard awareness from HBoss
* Cut HBoss test code which would explicitly disable S3Guard
  -obsolete and breaks the build on forthcoming hadoop releases.
* Also cut all references in docs and XML.
* EmbeddedS3 to implement deleteObject() and so support
  AWS SDKs 1.11.1026+, which expects this (hadoop 3.3.2+).
* Maven test runs to provide full stack traces on failures,
  so runtime failures can be debugged and fixed.

Change-Id: Iab10cfc6780cea064ea116efc1c177332d780514
Signed-off-by: Josh Elser <>
  • Loading branch information
steveloughran authored and joshelser committed May 23, 2022
1 parent 773be13 commit 8ba3a9138d0d9fee8cba7a21c8ee542eb2ffe490
Showing 5 changed files with 12 additions and 24 deletions.
@@ -23,6 +23,7 @@
@@ -145,6 +146,11 @@ public void deleteObject(String bucketName, String key) {

public void deleteObject(DeleteObjectRequest request) {
LOG.debug("deleteObject: {}", request);

public DeleteObjectsResult deleteObjects(DeleteObjectsRequest request) {
for (DeleteObjectsRequest.KeyVersion keyVersion : request.getKeys()) {
String key = keyVersion.getKey();
@@ -74,19 +74,17 @@ You may also want to configure:

An implementation based on Amazon's DynamoDB lock library was considered but
was not completed due to the lack of an efficient way to traverse the tree and
discover locks on child nodes. The benefit is that S3Guard is required for s3a
use and as such there's a dependency on DynamoDB anyway.
discover locks on child nodes.

## Storage Implementations

Currently HBOSS is primarily designed for and exclusively tested with Hadoop's
s3a client against Amazon S3. *S3Guard must be enabled, which is available in
Hadoop 2.9.0, 3.0.0, and higher*.
s3a client against Amazon S3.

Both the use of S3Guard and Zookeeper for locking (i.e. Zookeeper) have
The use of Zookeeper for locking (i.e. Zookeeper) have
implications for other clients that are not configured to share the same
metadata store and Zookeeper ensemble. Ideally, all clients should be have the
same configuration in these respects. Read-only clients may not share these
Zookeeper ensemble. Ideally, all clients should have the
same configuration. Read-only clients may not share these
resources with the HBase processes, but they will not have the added safety
provided by these features. Clients that do not share these resources and modify
data can compromise the correctness of HBase.
@@ -19,7 +19,6 @@
package org.apache.hadoop.hbase.oss;

import static org.apache.hadoop.fs.s3a.Constants.S3_CLIENT_FACTORY_IMPL;
import static org.apache.hadoop.fs.s3a.Constants.S3_METADATA_STORE_IMPL;
import static org.apache.hadoop.hbase.oss.Constants.DATA_URI;

@@ -29,7 +28,6 @@
import org.apache.commons.lang3.StringUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.s3a.s3guard.LocalMetadataStore;
import org.apache.hadoop.hbase.oss.sync.EmbeddedZK;
import org.apache.hadoop.metrics2.lib.DefaultMetricsSystem;
import org.apache.yetus.audience.InterfaceAudience;
@@ -69,9 +67,6 @@ public String getIdentifier() {
public static boolean usingEmbeddedS3 = false;

public static void conditionalStart(Configuration conf) {
if (StringUtils.isEmpty(conf.get(S3_METADATA_STORE_IMPL))) {
conf.set(S3_METADATA_STORE_IMPL, LocalMetadataStore.class.getName());

boolean notConfigured = StringUtils.isEmpty(conf.get(DATA_URI));
if (notConfigured) {
@@ -42,18 +42,6 @@
S3Guard will also be configured to the LocalMetadataStore implementation
by default. You may wish to configure DynamoDBMetadataStore for integration
or if S3Guard is not needed (S3-compatible appliance, for instance), you
can disable it by configuring NullMetadataStore explicitly (see below).

<description>Uncomment to disable S3Guard entirely</description>

<include xmlns="" href="auth-keys.xml">
@@ -106,6 +106,7 @@

0 comments on commit 8ba3a91

Please sign in to comment.