Apache
/
hbase
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HBASE-20403. Fix race between prefetch task and non-pread HFile reads 

With prefetch-on-open enabled, the task doing the prefetching was using
non-positional (i.e. streaming) reads. If the main (non-prefetch) thread
was also using non-positional reads, these two would conflict, because
inputstreams are not thread-safe for non-positional reads.

In the case of an encrypted filesystem, this could cause JVM crashes,
etc, as underlying cipher buffers were freed underneath the racing
threads. In the case of a non-encrypted filesystem, less severe errors
would be thrown. The included unit test reproduces the latter case.

(cherry picked from commit 025ddce868)
Signed-off-by: Todd Lipcon <todd@cloudera.com>
This commit is contained in:
Todd Lipcon 2018-06-20 15:23:31 -07:00
parent d6cea08efe
commit 3673bfc241
2 changed files with 30 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
hbase-server/src/main/java/org/apache/hadoop/hbase/io/hfile/HFileReaderImpl.java
View File

@ -275,8 +275,8 @@ public class HFileReaderImpl implements HFile.Reader, Configurable {
// next header, will not have happened...so, pass in the onDiskSize gotten from the
// cached block. This 'optimization' triggers extremely rarely I'd say.
long onDiskSize = prevBlock != null? prevBlock.getNextBlockOnDiskSize(): -1;
HFileBlock block = readBlock(offset, onDiskSize, true, false, false, false,
null, null);
HFileBlock block = readBlock(offset, onDiskSize, /*cacheBlock=*/true,
/*pread=*/true, false, false, null, null);
// Need not update the current block. Ideally here the readBlock won't find the
// block in cache. We call this readBlock so that block data is read from FS and
// cached in BC. So there is no reference count increment that happens here.

31
hbase-server/src/test/java/org/apache/hadoop/hbase/io/hfile/TestPrefetch.java
View File

@ -79,10 +79,35 @@ public class TestPrefetch {
@Test
public void testPrefetch() throws Exception {
Path storeFile = writeStoreFile();
Path storeFile = writeStoreFile("TestPrefetch");
readStoreFile(storeFile);
}
@Test
public void testPrefetchRace() throws Exception {
for (int i = 0; i < 10; i++) {
Path storeFile = writeStoreFile("TestPrefetchRace-" + i);
readStoreFileLikeScanner(storeFile);
}
}
/**
* Read a storefile in the same manner as a scanner -- using non-positional reads and
* without waiting for prefetch to complete.
*/
private void readStoreFileLikeScanner(Path storeFilePath) throws Exception {
// Open the file
HFile.Reader reader = HFile.createReader(fs, storeFilePath, cacheConf, true, conf);
do {
long offset = 0;
while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
HFileBlock block = reader.readBlock(offset, -1, false, /*pread=*/false,
false, true, null, null);
offset += block.getOnDiskSizeWithHeader();
}
} while (!reader.prefetchComplete());
}
private void readStoreFile(Path storeFilePath) throws Exception {
// Open the file
HFile.Reader reader = HFile.createReader(fs, storeFilePath, cacheConf, true, conf);
@ -108,8 +133,8 @@ public class TestPrefetch {
}
}
private Path writeStoreFile() throws IOException {
Path storeFileParentDir = new Path(TEST_UTIL.getDataTestDir(), "TestPrefetch");
private Path writeStoreFile(String fname) throws IOException {
Path storeFileParentDir = new Path(TEST_UTIL.getDataTestDir(), fname);
HFileContext meta = new HFileContextBuilder()
.withBlockSize(DATA_BLOCK_SIZE)
.build();