[ https://issues.apache.org/jira/browse/HBASE-27437?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17621116#comment-17621116 ]
Duo Zhang commented on HBASE-27437: ----------------------------------- OK, the code is for testing performance... {noformat} @Test public void testAutoCalcFixedOverHead() { Class[] classList = new Class[] { HFileContext.class, HRegion.class, BlockCacheKey.class, HFileBlock.class, HStore.class, LruBlockCache.class, StoreContext.class }; for (Class cl : classList) { // do estimate in advance to ensure class is loaded ClassSize.estimateBase(cl, false); long startTime = EnvironmentEdgeManager.currentTime(); ClassSize.estimateBase(cl, false); long endTime = EnvironmentEdgeManager.currentTime(); assertTrue(endTime - startTime < 5); } } {noformat} It is introduced in HBASE-24659, which is for making sure that the calculation overhead is small. But if the machine itself is small, it is possible that we fail since the assertion is a bit strict, especially if there is a gc... [~niuyulin] [~stack] WDYT? Can we relax the threshold? Thanks. > TestHeapSize is flaky > --------------------- > > Key: HBASE-27437 > URL: https://issues.apache.org/jira/browse/HBASE-27437 > Project: HBase > Issue Type: Bug > Components: test > Reporter: Duo Zhang > Priority: Major > > I believe it is just in memory computation, so it is weird that why it can be > flaky. > Need to dig more. -- This message was sent by Atlassian Jira (v8.20.10#820010)