ISPN-8150 SingleFileStoreStressTest intermittent failure #5368
Conversation
| // Verify that file size decreases and the file shrinks | ||
| assertTrue(length2 <= length1); | ||
| assertTrue(length3 < length2); | ||
| assertTrue(String.format("Length2=%d, Length3=%d", length2, length3), length3 <= length2); |
There was a problem hiding this comment.
Yeah, I think the store is definitely not shrinking all the time. This has been failing (randomly) for a long time, see https://issues.jboss.org/browse/ISPN-5730.
There was a problem hiding this comment.
Looking at the test some more, it's sleeping for 200ms, but the SingleFileStore.purge() actually does everything on the caller thread. So the test could pass a WithinThreadExecutor and skip the wait.
I also suggest testing store.getFileSize() before file.size(), just in case there's some caching going on in File.size(). I'd go even further and use TestingUtil.extractField to print the used+free entries if the store.getFileSize() check fails.
There was a problem hiding this comment.
I thought this test was about making sure that it does shrink, changing the condition in here voids the test.
There was a problem hiding this comment.
The reason it is not shrinking every time because the merging of entries in the first call to purge results in large FileEntries being created before the end of file. So when we attempt to write more entries, the existing FileEntries before the end of file are sufficient to accommodate the new entries and so when we call purge again, there are no free files at the end of the file to be truncated. I agree with @rvansa that my original changes void the test. I am going to push some changes that will ensure that there is always at least one free FileEntry at the end of the file and hence a truncation will occur.
| @@ -323,7 +306,7 @@ private void populateStoreRandomValues(int NUM_KEYS, int MAX_VALUE_SIZE, SingleF | |||
| StreamingMarshaller marshaller, List<String> keys) { | |||
| for (int j = 0; j < NUM_KEYS; j++) { | |||
| String key = "key" + j; | |||
| String value = key + "_value_" + j + times("123456789_", new Random().nextInt(MAX_VALUE_SIZE / 10)); | |||
| String value = key + "_value_" + j + times(new Random().nextInt(MAX_VALUE_SIZE / 10)); | |||
There was a problem hiding this comment.
Please reformat the parameters to camelCase.
ryanemerson
force-pushed
the
ISPN-8150
branch
2 times, most recently
from
257d981
to
623a6c6
Compare
| @@ -202,54 +200,47 @@ public void testFileTruncation() throws ExecutionException, InterruptedException | |||
| // Do some reading/writing entries with random size | |||
| final CountDownLatch stopLatch = new CountDownLatch(1); | |||
| Future[] writeFutures = new Future[NUM_WRITER_THREADS]; | |||
| for (int i = 0; i < NUM_WRITER_THREADS; i++) { | |||
| for (int i = 0; i < NUM_WRITER_THREADS; i++) | |||
ryanemerson
force-pushed
the
ISPN-8150
branch
2 times, most recently
from
c180c3d
to
6312468
Compare
Caused by test logic depending on eviction order.
|
Integrated, thanks Ryan! |
https://issues.jboss.org/browse/ISPN-8150