[
https://issues.jboss.org/browse/ISPN-8912?page=com.atlassian.jira.plugin....
]
Radim Vansa commented on ISPN-8912:
-----------------------------------
The node size is limited to 32767 as its length is stored in 2 bytes (OK, we could make
that 64k or maybe a bit more, but ByteBuffer does not have unsigned integers methods), so
the actual bug is that this is not failing gracefully and that we don't document that
(documentation mentions only the limitation for single key).
May I ask why are you increasing the node size? Have you found increasing performance with
larger nodes? (what disk do you use?)
The SoftIndexFileStore runs into a deadlock at startup if more than
~20000 values have been stored and max node size is > 32777 bytes
-------------------------------------------------------------------------------------------------------------------------------------
Key: ISPN-8912
URL:
https://issues.jboss.org/browse/ISPN-8912
Project: Infinispan
Issue Type: Bug
Components: Loaders and Stores
Affects Versions: 9.1.6.Final
Reporter: Andreas Pabst
Assignee: Radim Vansa
The following unit test just puts lots of entries into the cache and restarts the cache
repeatedly.
It succeeds if maxNodeSize is set to anything up to 32777, but runs into a deadlock from
32778 upwards. If going even higher it doesn't run into a deadlock anymore, but throws
an IllegalArgumentException instead.
As a workaround one can delete the index. But the problem resurfaces after the next
restart.
{code:java}
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
import org.infinispan.Cache;
import org.infinispan.configuration.cache.Configuration;
import org.infinispan.configuration.cache.ConfigurationBuilder;
import org.infinispan.manager.DefaultCacheManager;
import org.infinispan.manager.EmbeddedCacheManager;
import
org.infinispan.persistence.sifs.configuration.SoftIndexFileStoreConfigurationBuilder;
import org.junit.After;
import org.junit.Before;
import org.junit.Rule;
import org.junit.Test;
import org.junit.rules.Timeout;
public class SifsLargeNodeTest {
private EmbeddedCacheManager cacheManager;
@Rule
public Timeout timeout = new Timeout(15, TimeUnit.SECONDS);
private static Configuration createConfig(String location, int maxNodeSizeInBytes) {
return new ConfigurationBuilder()
.persistence().addStore(SoftIndexFileStoreConfigurationBuilder.class)
.dataLocation(location + "/data").indexLocation(location +
"/index")
.maxNodeSize(maxNodeSizeInBytes)
.preload(false).purgeOnStartup(false)
.build();
}
@Before
public void setUp() {
cacheManager = new DefaultCacheManager();
}
@After
public void tearDown() {
cacheManager.stop();
}
@Test
public void testLoad() {
int maxNodeSizeInBytes = 32780; // Note: Anything up to 32777 works fine
String cacheName = "sifslargenodetestcache";
Configuration cfg = createConfig(cacheName, maxNodeSizeInBytes);
int numberOfRuns = 10;
int valuesToPutPerRun = 10000;
for (int run = 0; run < numberOfRuns; run++) {
System.out.println("RUN " + run);
cacheManager.defineConfiguration(cacheName, cfg);
Cache<Integer, String> testCacheBeforeReduction =
cacheManager.getCache(cacheName);
IntStream.range(run * valuesToPutPerRun, (run + 1) * valuesToPutPerRun)
.forEach(i -> testCacheBeforeReduction.put(i,
Integer.toString(i)));
cacheManager.stop();
cacheManager = new DefaultCacheManager();
}
}
}
{code}
--
This message was sent by Atlassian JIRA
(v7.5.0#75005)