Make the default S3 buffer size depend on the available memory. (#21299)

Currently the default S3 buffer size is 100MB, which can be a lot for small
heaps. This pull request updates the default to be 100MB for heaps that are
greater than 2GB and 5% of the heap size otherwise.
This commit is contained in:
Adrien Grand 2016-11-03 16:07:52 +01:00 committed by GitHub
parent 548c12d60f
commit 7ec51d628d
3 changed files with 29 additions and 4 deletions

View File

@ -217,7 +217,8 @@ The following settings are supported:
to split the chunk into several parts, each of `buffer_size` length, and to split the chunk into several parts, each of `buffer_size` length, and
to upload each part in its own request. Note that setting a buffer to upload each part in its own request. Note that setting a buffer
size lower than `5mb` is not allowed since it will prevents the use of the size lower than `5mb` is not allowed since it will prevents the use of the
Multipart API and may result in upload errors. Defaults to `100mb`. Multipart API and may result in upload errors. Defaults to the minimum
between `100mb` and `5%` of the heap size.
`max_retries`:: `max_retries`::

View File

@ -33,6 +33,7 @@ import org.elasticsearch.common.settings.Setting.Property;
import org.elasticsearch.common.settings.Settings; import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.unit.ByteSizeUnit; import org.elasticsearch.common.unit.ByteSizeUnit;
import org.elasticsearch.common.unit.ByteSizeValue; import org.elasticsearch.common.unit.ByteSizeValue;
import org.elasticsearch.monitor.jvm.JvmInfo;
import org.elasticsearch.repositories.RepositoryException; import org.elasticsearch.repositories.RepositoryException;
import org.elasticsearch.repositories.blobstore.BlobStoreRepository; import org.elasticsearch.repositories.blobstore.BlobStoreRepository;
@ -101,14 +102,27 @@ public class S3Repository extends BlobStoreRepository {
*/ */
Setting<Boolean> SERVER_SIDE_ENCRYPTION_SETTING = Setting<Boolean> SERVER_SIDE_ENCRYPTION_SETTING =
Setting.boolSetting("repositories.s3.server_side_encryption", false, Property.NodeScope); Setting.boolSetting("repositories.s3.server_side_encryption", false, Property.NodeScope);
/**
* Default is to use 100MB (S3 defaults) for heaps above 2GB and 5% of
* the available memory for smaller heaps.
*/
ByteSizeValue DEFAULT_BUFFER_SIZE = new ByteSizeValue(
Math.max(
ByteSizeUnit.MB.toBytes(5), // minimum value
Math.min(
ByteSizeUnit.MB.toBytes(100),
JvmInfo.jvmInfo().getMem().getHeapMax().getBytes() / 20)),
ByteSizeUnit.BYTES);
/** /**
* repositories.s3.buffer_size: Minimum threshold below which the chunk is uploaded using a single request. Beyond this threshold, * repositories.s3.buffer_size: Minimum threshold below which the chunk is uploaded using a single request. Beyond this threshold,
* the S3 repository will use the AWS Multipart Upload API to split the chunk into several parts, each of buffer_size length, and * the S3 repository will use the AWS Multipart Upload API to split the chunk into several parts, each of buffer_size length, and
* to upload each part in its own request. Note that setting a buffer size lower than 5mb is not allowed since it will prevents the * to upload each part in its own request. Note that setting a buffer size lower than 5mb is not allowed since it will prevents the
* use of the Multipart API and may result in upload errors. Defaults to 100m. * use of the Multipart API and may result in upload errors. Defaults to the minimum between 100MB and 5% of the heap size.
*/ */
Setting<ByteSizeValue> BUFFER_SIZE_SETTING = Setting<ByteSizeValue> BUFFER_SIZE_SETTING =
Setting.byteSizeSetting("repositories.s3.buffer_size", new ByteSizeValue(100, ByteSizeUnit.MB), Setting.byteSizeSetting("repositories.s3.buffer_size", DEFAULT_BUFFER_SIZE,
new ByteSizeValue(5, ByteSizeUnit.MB), new ByteSizeValue(5, ByteSizeUnit.TB), Property.NodeScope); new ByteSizeValue(5, ByteSizeUnit.MB), new ByteSizeValue(5, ByteSizeUnit.TB), Property.NodeScope);
/** /**
* repositories.s3.max_retries: Number of retries in case of S3 errors. Defaults to 3. * repositories.s3.max_retries: Number of retries in case of S3 errors. Defaults to 3.
@ -195,12 +209,13 @@ public class S3Repository extends BlobStoreRepository {
* @see Repositories#SERVER_SIDE_ENCRYPTION_SETTING * @see Repositories#SERVER_SIDE_ENCRYPTION_SETTING
*/ */
Setting<Boolean> SERVER_SIDE_ENCRYPTION_SETTING = Setting.boolSetting("server_side_encryption", false); Setting<Boolean> SERVER_SIDE_ENCRYPTION_SETTING = Setting.boolSetting("server_side_encryption", false);
/** /**
* buffer_size * buffer_size
* @see Repositories#BUFFER_SIZE_SETTING * @see Repositories#BUFFER_SIZE_SETTING
*/ */
Setting<ByteSizeValue> BUFFER_SIZE_SETTING = Setting<ByteSizeValue> BUFFER_SIZE_SETTING =
Setting.byteSizeSetting("buffer_size", new ByteSizeValue(100, ByteSizeUnit.MB), Setting.byteSizeSetting("buffer_size", Repositories.DEFAULT_BUFFER_SIZE,
new ByteSizeValue(5, ByteSizeUnit.MB), new ByteSizeValue(5, ByteSizeUnit.TB)); new ByteSizeValue(5, ByteSizeUnit.MB), new ByteSizeValue(5, ByteSizeUnit.TB));
/** /**
* max_retries * max_retries

View File

@ -32,6 +32,7 @@ import org.elasticsearch.common.unit.ByteSizeUnit;
import org.elasticsearch.common.unit.ByteSizeValue; import org.elasticsearch.common.unit.ByteSizeValue;
import org.elasticsearch.repositories.RepositoryException; import org.elasticsearch.repositories.RepositoryException;
import org.elasticsearch.test.ESTestCase; import org.elasticsearch.test.ESTestCase;
import org.hamcrest.Matchers;
import static org.elasticsearch.repositories.s3.S3Repository.Repositories; import static org.elasticsearch.repositories.s3.S3Repository.Repositories;
import static org.elasticsearch.repositories.s3.S3Repository.Repository; import static org.elasticsearch.repositories.s3.S3Repository.Repository;
@ -117,4 +118,12 @@ public class S3RepositoryTests extends ESTestCase {
assertEquals("foo/bar/", s3repo.basePath().buildAsString()); // make sure leading `/` is removed and trailing is added assertEquals("foo/bar/", s3repo.basePath().buildAsString()); // make sure leading `/` is removed and trailing is added
} }
public void testDefaultBufferSize() {
ByteSizeValue defaultBufferSize = S3Repository.Repository.BUFFER_SIZE_SETTING.get(Settings.EMPTY);
assertThat(defaultBufferSize, Matchers.lessThanOrEqualTo(new ByteSizeValue(100, ByteSizeUnit.MB)));
assertThat(defaultBufferSize, Matchers.greaterThanOrEqualTo(new ByteSizeValue(5, ByteSizeUnit.MB)));
ByteSizeValue defaultNodeBufferSize = S3Repository.Repositories.BUFFER_SIZE_SETTING.get(Settings.EMPTY);
assertEquals(defaultBufferSize, defaultNodeBufferSize);
}
} }