hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "zhangduo (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-13070) Fix TestCacheOnWrite
Date Fri, 20 Feb 2015 06:13:12 GMT

    [ https://issues.apache.org/jira/browse/HBASE-13070?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14328605#comment-14328605

zhangduo commented on HBASE-13070:

Oh I think I found the problem.

We turned on hfile prefetching in this test after HBASE-12270(maybe a mistake).
See https://builds.apache.org/job/HBase-TRUNK/6110/artifact/hbase-server/target/surefire-reports/org.apache.hadoop.hbase.io.hfile.TestCacheOnWrite-output.txt
2015-02-10 06:48:00,128 DEBUG [main] hfile.PrefetchExecutor(102): Prefetch requested for /home/jenkins/jenkins-slave/workspace/HBase-TRUNK/hbase-server/target/test-data/170a3172-9a2e-4269-8085-8433f46141c6/test_cache_on_write/b7d5cfcd4b57411eaac3ebbb83626249,
delay=905 ms
And see https://builds.apache.org/job/HBase-TRUNK/5796/artifact/hbase-server/target/surefire-reports/org.apache.hadoop.hbase.io.hfile.TestCacheOnWrite-output.txt
Grep "Prefetch requested for" and get nothing.

And we can see that, the prefetch operation has a delay, usually nearly 1 sec in tests, so
if the test run fast enough then there is no problem. But if we run the test on a slow machine
then BlockCache maybe ruined before we finish checking cached blocks and make the test fail.

Thanks [~tedyu] to let me add a log when clearing BlockCache multiple times then I found the
actual issue.

> Fix TestCacheOnWrite
> --------------------
>                 Key: HBASE-13070
>                 URL: https://issues.apache.org/jira/browse/HBASE-13070
>             Project: HBase
>          Issue Type: Bug
>          Components: test
>    Affects Versions: 2.0.0
>            Reporter: zhangduo
>            Assignee: zhangduo
>         Attachments: HBASE-13070.patch
> TestCacheOnWrite uses TestHFileWriterV2.randomOrderedKey to generate a random byte array,
then use first 32 bytes as row and remaining part as family and qualifier. But TestHFileWriterV2.randomOrderedKey
may return a byte array only contains 32 bytes, so there will be zero length family and qualifier.

This message was sent by Atlassian JIRA

View raw message