This is an automated email from the ASF dual-hosted git repository.
NSAmelchev pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/ignite.git
The following commit(s) were added to refs/heads/master by this push:
new bd3c3bab661 IGNITE-29175 Copy compressed snapshot partitions sparsely
on restore instead of punching holes (#13710)
bd3c3bab661 is described below
commit bd3c3bab66175d43fc1d5d5fec59ad6f2ff0fa7b
Author: Nikita Amelchev <[email protected]>
AuthorDate: Tue Nov 3 18:31:19 2026 +0300
IGNITE-29175 Copy compressed snapshot partitions sparsely on restore
instead of punching holes (#13710)
---
.../snapshot/SnapshotCompressionBasicTest.java | 40 +++++++++++
.../snapshot/SnapshotRestoreProcess.java | 84 ++++++++++++++++------
2 files changed, 104 insertions(+), 20 deletions(-)
diff --git
a/modules/compress/src/test/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotCompressionBasicTest.java
b/modules/compress/src/test/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotCompressionBasicTest.java
index ae0c2da94d2..62a3e205864 100644
---
a/modules/compress/src/test/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotCompressionBasicTest.java
+++
b/modules/compress/src/test/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotCompressionBasicTest.java
@@ -43,6 +43,7 @@ import org.apache.ignite.IgniteCheckedException;
import org.apache.ignite.IgniteException;
import org.apache.ignite.cache.QueryEntity;
import org.apache.ignite.cache.QueryIndex;
+import org.apache.ignite.cache.affinity.rendezvous.RendezvousAffinityFunction;
import org.apache.ignite.cluster.ClusterState;
import org.apache.ignite.configuration.CacheConfiguration;
import org.apache.ignite.configuration.DiskPageCompression;
@@ -57,6 +58,8 @@ import
org.apache.ignite.internal.processors.compress.CompressionProcessor;
import org.apache.ignite.internal.util.typedef.F;
import org.apache.ignite.internal.util.typedef.G;
import org.apache.ignite.internal.util.typedef.internal.U;
+import org.apache.ignite.metric.MetricRegistry;
+import org.apache.ignite.spi.metric.LongMetric;
import org.apache.ignite.testframework.GridTestUtils;
import org.junit.After;
import org.junit.Before;
@@ -67,6 +70,9 @@ import org.mockito.Mockito;
import static org.apache.ignite.events.EventType.EVTS_CLUSTER_SNAPSHOT;
import static
org.apache.ignite.events.EventType.EVT_CLUSTER_SNAPSHOT_RESTORE_FINISHED;
import static
org.apache.ignite.events.EventType.EVT_CLUSTER_SNAPSHOT_RESTORE_STARTED;
+import static
org.apache.ignite.internal.processors.cache.CacheGroupMetricsImpl.CACHE_GROUP_METRICS_PREFIX;
+import static
org.apache.ignite.internal.processors.metric.impl.MetricUtils.metricName;
+import static org.junit.Assume.assumeFalse;
/** */
public class SnapshotCompressionBasicTest extends AbstractSnapshotSelfTest {
@@ -265,6 +271,40 @@ public class SnapshotCompressionBasicTest extends
AbstractSnapshotSelfTest {
}
}
+ /** */
+ @Test
+ public void testRestoredPartitionsAreSparse() throws Exception {
+ // Doesn't matter for a single node.
+ assumeFalse(onlyPrimary);
+
+ IgniteEx ignite = startGrid(0);
+
+ ignite.cluster().state(ClusterState.ACTIVE);
+
+ IgniteCache<Integer, byte[]> cache = ignite.createCache(new
CacheConfiguration<Integer, byte[]>(DEFAULT_CACHE_NAME)
+ .setAffinity(new RendezvousAffinityFunction(false, 4))
+ .setDiskPageCompression(DISK_PAGE_COMPRESSION));
+
+ // Partition files must be well above 64 KiB: the end-of-file
preallocation that can make the copy dense (XFS does it)
+ // starts at that size.
+ for (int i = 0; i < 4_000; i++)
+ cache.put(i, new byte[1024]);
+
+ ignite.snapshot().createSnapshot(SNAPSHOT_NAME).get(TIMEOUT);
+
+ cache.destroy();
+
+ ignite.snapshot().restoreSnapshot(SNAPSHOT_NAME, null).get(TIMEOUT);
+
+ MetricRegistry mreg =
ignite.context().metric().registry(metricName(CACHE_GROUP_METRICS_PREFIX,
DEFAULT_CACHE_NAME));
+
+ long size = mreg.<LongMetric>findMetric("StorageSize").value();
+ long sparseSize =
mreg.<LongMetric>findMetric("SparseStorageSize").value();
+
+ // Zero-filled values compress every 8 KiB page into one 4 KiB block,
so a sparse group takes about half of its size.
+ assertTrue("sparseSize=" + sparseSize + ", size=" + size, sparseSize <
0.75 * size);
+ }
+
/** {@inheritDoc} */
@Override protected Function<Integer, Object> valueBuilder() {
return i -> new Value("name_" + i);
diff --git
a/modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotRestoreProcess.java
b/modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotRestoreProcess.java
index 2643a5575f6..f36bd768665 100644
---
a/modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotRestoreProcess.java
+++
b/modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotRestoreProcess.java
@@ -19,6 +19,7 @@ package
org.apache.ignite.internal.processors.cache.persistence.snapshot;
import java.io.File;
import java.io.IOException;
+import java.io.RandomAccessFile;
import java.nio.ByteBuffer;
import java.nio.file.Files;
import java.nio.file.Path;
@@ -74,6 +75,8 @@ import
org.apache.ignite.internal.processors.cache.distributed.dht.topology.Grid
import
org.apache.ignite.internal.processors.cache.persistence.CacheStripedExecutor;
import
org.apache.ignite.internal.processors.cache.persistence.GridCacheDatabaseSharedManager;
import
org.apache.ignite.internal.processors.cache.persistence.checkpoint.CheckpointProgress;
+import org.apache.ignite.internal.processors.cache.persistence.file.FileIO;
+import
org.apache.ignite.internal.processors.cache.persistence.file.FileIOFactory;
import
org.apache.ignite.internal.processors.cache.persistence.file.FilePageStore;
import
org.apache.ignite.internal.processors.cache.persistence.file.FilePageStoreManager;
import
org.apache.ignite.internal.processors.cache.persistence.file.FileVersionCheckingFactory;
@@ -99,6 +102,9 @@ import org.apache.ignite.metric.MetricRegistry;
import org.apache.ignite.plugin.extensions.communication.Message;
import org.jetbrains.annotations.Nullable;
+import static java.nio.file.StandardOpenOption.CREATE;
+import static java.nio.file.StandardOpenOption.READ;
+import static java.nio.file.StandardOpenOption.WRITE;
import static java.util.Optional.ofNullable;
import static
org.apache.ignite.internal.pagemem.PageIdAllocator.INDEX_PARTITION;
import static
org.apache.ignite.internal.processors.cache.persistence.filename.NodeFileTree.partId;
@@ -1764,30 +1770,14 @@ public class SnapshotRestoreProcess {
", snpDir=" + snpFile.getAbsolutePath() + ", name=" +
snpFile.getName() + ']');
}
- IgniteSnapshotManager.copy(snapMgr.ioFactory(), snpFile,
tmpPartFile, snpFile.length());
+ if (opCtx.isGroupCompressed(grpId))
+ copyCompressed(opCtx, grpId, snpFile, tmpPartFile);
+ else
+ IgniteSnapshotManager.copy(snapMgr.ioFactory(), snpFile,
tmpPartFile, snpFile.length());
return tmpPartFile.toPath();
}, snapMgr.snapshotExecutorService());
- if (opCtx.isGroupCompressed(grpId)) {
- copyPartFut = copyPartFut.thenComposeAsync(
- p -> {
- IgniteCompletableFuture<Path> result = new
IgniteCompletableFuture<>();
-
- try {
- punchHole(grpId, partFut.partId, tmpPartFile);
-
- result.complete(tmpPartFile.toPath());
- }
- catch (Throwable t) {
- result.completeExceptionally(t);
- }
- return result;
- },
- snapMgr.snapshotExecutorService()
- );
- }
-
copyPartFut.whenComplete((r, t) -> opCtx.errHnd.accept(t))
.whenComplete((r, t) -> {
if (t == null)
@@ -1797,6 +1787,60 @@ public class SnapshotRestoreProcess {
});
}
+ /**
+ * Copies a partition of a compressed cache group page by page writing
only the compressed bytes of each page, so the
+ * rest of the page is left unallocated by the file system and the copy is
sparse without punching holes in it.
+ */
+ private void copyCompressed(SnapshotRestoreContext opCtx, int grpId, File
from, File to) {
+ FileIOFactory ioFactory =
ctx.cache().context().snapshotMgr().ioFactory();
+ FilePageStoreManager storeMgr =
(FilePageStoreManager)ctx.cache().context().pageStore();
+ int pageSize = ctx.cache().context().database().pageSize();
+
+ try (FileIO src = ioFactory.create(from, READ); FileIO dest =
ioFactory.create(to, CREATE, WRITE)) {
+ long len = src.size();
+
+ // The file gets its final size first: XFS preallocates blocks on
extending writes, and those between the pages
+ // would stay allocated, see "speculative preallocation" in
https://xfs.org/index.php/XFS_FAQ.
+ try (RandomAccessFile file = new RandomAccessFile(to, "rw")) {
+ file.setLength(len);
+ }
+
+ ByteBuffer buf = locBuff.get();
+
+ buf.clear();
+ src.readFully(buf, 0);
+ buf.flip();
+
+ buf.getLong(); // Signature.
+
+ int hdrSize = storeMgr.getPageStoreFactory(grpId,
null).headerSize(buf.getInt());
+
+ buf.rewind();
+ buf.limit(hdrSize);
+ dest.writeFully(buf, 0);
+
+ for (long off = hdrSize; off < len; off += pageSize) {
+ if (opCtx.stopChecker.getAsBoolean())
+ throw new IgniteInterruptedException("The operation has
been stopped on copy file: " + from.getAbsolutePath());
+
+ if (Thread.interrupted())
+ throw new IgniteInterruptedException("Thread has been
interrupted: " + Thread.currentThread().getName());
+
+ buf.clear();
+ src.readFully(buf, off);
+ buf.flip();
+
+ if (PageIO.getCompressionType(buf) !=
CompressionProcessor.UNCOMPRESSED_PAGE)
+ buf.limit(PageIO.getCompressedSize(buf));
+
+ dest.writeFully(buf, off);
+ }
+ }
+ catch (IOException e) {
+ throw new IgniteException(e);
+ }
+ }
+
/** */
private void punchHole(int grpId, int partId, File partFile) throws
Exception {
FilePageStoreManager storeMgr =
(FilePageStoreManager)ctx.cache().context().pageStore();