Brutex
/
cavis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #9208 from eclipse/ag_aeron_fixes 

Mark aeron tests not thread safe for more consistent execution, update aeron version
master
Adam Gibson 2021-03-06 09:19:55 +09:00 committed by GitHub
parent c088d4117f be25b08a29
commit 4852cb975f
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
11 changed files with 53 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libnd4j/tests_cpu/run_tests.sh
View File

@ -56,7 +56,7 @@ if [ -n "$BUILD_PATH" ]; then
export PATH="$PATH:$BUILD_PATH" export PATH="$PATH:$BUILD_PATH"
fi fi
../blasbuild/${CHIP}/tests_cpu/layers_tests/runtests ../blasbuild/${CHIP}/te NdArrayIpcTeststs_cpu/layers_tests/runtests
# Workaround to fix posix path conversion problem on Windows (http://mingw.org/wiki/Posix_path_conversion) # Workaround to fix posix path conversion problem on Windows (http://mingw.org/wiki/Posix_path_conversion)
[ -f "${GTEST_OUTPUT#*:}" ] && cp -a surefire-reports/ ../target && rm -rf surefire-reports/ [ -f "${GTEST_OUTPUT#*:}" ] && cp -a surefire-reports/ ../target && rm -rf surefire-reports/

5
nd4j/nd4j-serde/nd4j-aeron/pom.xml
View File

@ -39,7 +39,7 @@
<maven.compiler.source>1.8</maven.compiler.source> <maven.compiler.source>1.8</maven.compiler.source>
<maven.compiler.target>1.8</maven.compiler.target> <maven.compiler.target>1.8</maven.compiler.target>
<sbe.version>1.5.4</sbe.version> <sbe.version>1.5.4</sbe.version>
<aeron.version>1.4.0</aeron.version> <aeron.version>1.32.0</aeron.version>
</properties> </properties>
<dependencies> <dependencies>
@ -93,7 +93,7 @@
<configuration> <configuration>
<environmentVariables> <environmentVariables>
<LD_LIBRARY_PATH> <LD_LIBRARY_PATH>
${env.LD_LIBRARY_PATH}:${user.dir}:${libnd4jhome}/blasbuild/cpu/blas/ ${env.LD_LIBRARY_PATH}${path.separator}${user.dir}${path.separator}${libnd4jhome}/blasbuild/cpu/blas/${path.separator}${libnd4jhome}/../nd4j/nd4j-backends/nd4j-backend-impls/nd4j-native/target/classes
</LD_LIBRARY_PATH> </LD_LIBRARY_PATH>
</environmentVariables> </environmentVariables>
<testSourceDirectory>src/test/java</testSourceDirectory> <testSourceDirectory>src/test/java</testSourceDirectory>
@ -119,7 +119,6 @@
For testing large zoo models, this may not be enough (so comment it out). For testing large zoo models, this may not be enough (so comment it out).
--> -->
<argLine>-Ddtype=float -Dfile.encoding=UTF-8 -Xmx8g</argLine>
</configuration> </configuration>
</plugin> </plugin>
</plugins> </plugins>

4
nd4j/nd4j-serde/nd4j-aeron/src/main/java/org/nd4j/aeron/ipc/AeronUtil.java
View File

@ -59,7 +59,9 @@ public class AeronUtil {
ipcLength += 2; ipcLength += 2;
// System.setProperty("aeron.term.buffer.size",String.valueOf(ipcLength)); // System.setProperty("aeron.term.buffer.size",String.valueOf(ipcLength));
final MediaDriver.Context ctx = final MediaDriver.Context ctx =
new MediaDriver.Context().threadingMode(ThreadingMode.SHARED).dirsDeleteOnStart(true) new MediaDriver.Context().threadingMode(ThreadingMode.SHARED)
.dirDeleteOnStart(true)
.dirDeleteOnShutdown(true)
/* .ipcTermBufferLength(ipcLength) /* .ipcTermBufferLength(ipcLength)
.publicationTermBufferLength(ipcLength) .publicationTermBufferLength(ipcLength)
.maxTermBufferLength(ipcLength)*/ .maxTermBufferLength(ipcLength)*/

6
nd4j/nd4j-serde/nd4j-aeron/src/main/java/org/nd4j/aeron/ipc/LowLatencyMediaDriver.java
View File

@ -34,8 +34,7 @@ public class LowLatencyMediaDriver {
@SuppressWarnings("checkstyle:UncommentedMain") @SuppressWarnings("checkstyle:UncommentedMain")
public static void main(final String... args) { public static void main(final String... args) {
MediaDriver.loadPropertiesFiles(args); MediaDriver.main(args);
setProperty(DISABLE_BOUNDS_CHECKS_PROP_NAME, "true"); setProperty(DISABLE_BOUNDS_CHECKS_PROP_NAME, "true");
setProperty("aeron.mtu.length", "16384"); setProperty("aeron.mtu.length", "16384");
setProperty("aeron.socket.so_sndbuf", "2097152"); setProperty("aeron.socket.so_sndbuf", "2097152");
@ -43,7 +42,8 @@ public class LowLatencyMediaDriver {
setProperty("aeron.rcv.initial.window.length", "2097152"); setProperty("aeron.rcv.initial.window.length", "2097152");
final MediaDriver.Context ctx = final MediaDriver.Context ctx =
new MediaDriver.Context().threadingMode(ThreadingMode.DEDICATED).dirsDeleteOnStart(true) new MediaDriver.Context().threadingMode(ThreadingMode.DEDICATED).dirDeleteOnStart(true)
.dirDeleteOnShutdown(true)
.termBufferSparseFile(false).conductorIdleStrategy(new BusySpinIdleStrategy()) .termBufferSparseFile(false).conductorIdleStrategy(new BusySpinIdleStrategy())
.receiverIdleStrategy(new BusySpinIdleStrategy()) .receiverIdleStrategy(new BusySpinIdleStrategy())
.senderIdleStrategy(new BusySpinIdleStrategy()); .senderIdleStrategy(new BusySpinIdleStrategy());

3
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/AeronNDArraySerdeTest.java
View File

@ -28,13 +28,14 @@ import org.nd4j.common.tests.BaseND4JTest;
import org.nd4j.linalg.api.ndarray.INDArray; import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import java.io.BufferedOutputStream; import java.io.BufferedOutputStream;
import java.io.ByteArrayOutputStream; import java.io.ByteArrayOutputStream;
import java.io.DataOutputStream; import java.io.DataOutputStream;
import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertTrue; import static org.junit.Assert.assertTrue;
@NotThreadSafe
public class AeronNDArraySerdeTest extends BaseND4JTest { public class AeronNDArraySerdeTest extends BaseND4JTest {
@Test @Test

11
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/LargeNdArrayIpcTest.java
View File

@ -31,11 +31,13 @@ import org.nd4j.common.tests.BaseND4JTest;
import org.nd4j.linalg.api.ndarray.INDArray; import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import java.util.concurrent.atomic.AtomicBoolean; import java.util.concurrent.atomic.AtomicBoolean;
import static org.junit.Assert.assertFalse; import static org.junit.Assert.assertFalse;
@Slf4j @Slf4j
@NotThreadSafe
public class LargeNdArrayIpcTest extends BaseND4JTest { public class LargeNdArrayIpcTest extends BaseND4JTest {
private MediaDriver mediaDriver; private MediaDriver mediaDriver;
private Aeron.Context ctx; private Aeron.Context ctx;
@ -73,9 +75,10 @@ public class LargeNdArrayIpcTest extends BaseND4JTest {
int length = (int) 1e7; int length = (int) 1e7;
INDArray arr = Nd4j.ones(length); INDArray arr = Nd4j.ones(length);
AeronNDArrayPublisher publisher; AeronNDArrayPublisher publisher;
ctx = new Aeron.Context().publicationConnectionTimeout(-1).availableImageHandler(AeronUtil::printAvailableImage) ctx = new Aeron.Context()
.driverTimeoutMs(-1).availableImageHandler(AeronUtil::printAvailableImage)
.unavailableImageHandler(AeronUtil::printUnavailableImage) .unavailableImageHandler(AeronUtil::printUnavailableImage)
.aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveInterval(10000) .aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveIntervalNs(10000)
.errorHandler(err -> err.printStackTrace()); .errorHandler(err -> err.printStackTrace());
final AtomicBoolean running = new AtomicBoolean(true); final AtomicBoolean running = new AtomicBoolean(true);
@ -149,10 +152,10 @@ public class LargeNdArrayIpcTest extends BaseND4JTest {
private Aeron.Context getContext() { private Aeron.Context getContext() {
if (ctx == null) if (ctx == null)
ctx = new Aeron.Context().publicationConnectionTimeout(-1) ctx = new Aeron.Context().driverTimeoutMs(-1)
.availableImageHandler(AeronUtil::printAvailableImage) .availableImageHandler(AeronUtil::printAvailableImage)
.unavailableImageHandler(AeronUtil::printUnavailableImage) .unavailableImageHandler(AeronUtil::printUnavailableImage)
.aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveInterval(10000) .aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveIntervalNs(10000)
.errorHandler(err -> err.printStackTrace()); .errorHandler(err -> err.printStackTrace());
return ctx; return ctx;
} }

3
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/NDArrayMessageTest.java
View File

@ -26,8 +26,11 @@ import org.nd4j.common.tests.BaseND4JTest;
import org.nd4j.linalg.api.ndarray.INDArray; import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertEquals;
@NotThreadSafe
public class NDArrayMessageTest extends BaseND4JTest { public class NDArrayMessageTest extends BaseND4JTest {
@Test @Test

6
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/NdArrayIpcTest.java
View File

@ -32,12 +32,14 @@ import org.nd4j.linalg.factory.Nd4j;
import org.slf4j.Logger; import org.slf4j.Logger;
import org.slf4j.LoggerFactory; import org.slf4j.LoggerFactory;
import javax.annotation.concurrent.NotThreadSafe;
import java.util.concurrent.ExecutorService; import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors; import java.util.concurrent.Executors;
import java.util.concurrent.atomic.AtomicBoolean; import java.util.concurrent.atomic.AtomicBoolean;
import static org.junit.Assert.assertFalse; import static org.junit.Assert.assertFalse;
@NotThreadSafe
public class NdArrayIpcTest extends BaseND4JTest { public class NdArrayIpcTest extends BaseND4JTest {
private MediaDriver mediaDriver; private MediaDriver mediaDriver;
private static Logger log = LoggerFactory.getLogger(NdArrayIpcTest.class); private static Logger log = LoggerFactory.getLogger(NdArrayIpcTest.class);
@ -223,10 +225,10 @@ public class NdArrayIpcTest extends BaseND4JTest {
private Aeron.Context getContext() { private Aeron.Context getContext() {
if (ctx == null) if (ctx == null)
ctx = new Aeron.Context().publicationConnectionTimeout(1000) ctx = new Aeron.Context().driverTimeoutMs(1000)
.availableImageHandler(image -> System.out.println(image)) .availableImageHandler(image -> System.out.println(image))
.unavailableImageHandler(AeronUtil::printUnavailableImage) .unavailableImageHandler(AeronUtil::printUnavailableImage)
.aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveInterval(1000) .aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveIntervalNs(1000)
.errorHandler(e -> log.error(e.toString(), e)); .errorHandler(e -> log.error(e.toString(), e));
return ctx; return ctx;
} }

3
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/chunk/ChunkAccumulatorTests.java
View File

@ -25,8 +25,11 @@ import org.nd4j.common.tests.BaseND4JTest;
import org.nd4j.aeron.ipc.NDArrayMessage; import org.nd4j.aeron.ipc.NDArrayMessage;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertEquals;
@NotThreadSafe
public class ChunkAccumulatorTests extends BaseND4JTest { public class ChunkAccumulatorTests extends BaseND4JTest {
@Test @Test

2
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/chunk/NDArrayMessageChunkTests.java
View File

@ -27,11 +27,13 @@ import org.nd4j.aeron.ipc.NDArrayMessage;
import org.nd4j.aeron.util.BufferUtil; import org.nd4j.aeron.util.BufferUtil;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import java.nio.ByteBuffer; import java.nio.ByteBuffer;
import static org.junit.Assert.assertArrayEquals; import static org.junit.Assert.assertArrayEquals;
import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertEquals;
@NotThreadSafe
public class NDArrayMessageChunkTests extends BaseND4JTest { public class NDArrayMessageChunkTests extends BaseND4JTest {
@Test @Test

9
nd4j/nd4j-serde/nd4j-aeron/src/test/java/org/nd4j/aeron/ipc/response/AeronNDArrayResponseTest.java
View File

@ -33,12 +33,14 @@ import org.nd4j.aeron.ipc.*;
import org.nd4j.linalg.api.ndarray.INDArray; import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.factory.Nd4j;
import javax.annotation.concurrent.NotThreadSafe;
import java.util.concurrent.atomic.AtomicBoolean; import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicInteger; import java.util.concurrent.atomic.AtomicInteger;
import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertEquals;
@Slf4j @Slf4j
@NotThreadSafe
public class AeronNDArrayResponseTest extends BaseND4JTest { public class AeronNDArrayResponseTest extends BaseND4JTest {
private MediaDriver mediaDriver; private MediaDriver mediaDriver;
@ -51,7 +53,8 @@ public class AeronNDArrayResponseTest extends BaseND4JTest {
public void before() { public void before() {
if(isIntegrationTests()) { if(isIntegrationTests()) {
final MediaDriver.Context ctx = final MediaDriver.Context ctx =
new MediaDriver.Context().threadingMode(ThreadingMode.SHARED).dirsDeleteOnStart(true) new MediaDriver.Context().threadingMode(ThreadingMode.SHARED).dirDeleteOnShutdown(true)
.dirDeleteOnStart(true)
.termBufferSparseFile(false).conductorIdleStrategy(new BusySpinIdleStrategy()) .termBufferSparseFile(false).conductorIdleStrategy(new BusySpinIdleStrategy())
.receiverIdleStrategy(new BusySpinIdleStrategy()) .receiverIdleStrategy(new BusySpinIdleStrategy())
.senderIdleStrategy(new BusySpinIdleStrategy()); .senderIdleStrategy(new BusySpinIdleStrategy());
@ -69,10 +72,10 @@ public class AeronNDArrayResponseTest extends BaseND4JTest {
int streamId = 10; int streamId = 10;
int responderStreamId = 11; int responderStreamId = 11;
String host = "127.0.0.1"; String host = "127.0.0.1";
Aeron.Context ctx = new Aeron.Context().publicationConnectionTimeout(-1) Aeron.Context ctx = new Aeron.Context().driverTimeoutMs(-1)
.availableImageHandler(AeronUtil::printAvailableImage) .availableImageHandler(AeronUtil::printAvailableImage)
.unavailableImageHandler(AeronUtil::printUnavailableImage) .unavailableImageHandler(AeronUtil::printUnavailableImage)
.aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveInterval(1000) .aeronDirectoryName(mediaDriver.aeronDirectoryName()).keepAliveIntervalNs(1000)
.errorHandler(e -> log.error(e.toString(), e)); .errorHandler(e -> log.error(e.toString(), e));
int baseSubscriberPort = 40123 + new java.util.Random().nextInt(1000); int baseSubscriberPort = 40123 + new java.util.Random().nextInt(1000);