hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sergey Shelukhin (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-8700) IntegrationTestBigLinkedList can fail due to random number collision
Date Fri, 07 Jun 2013 01:30:21 GMT

     [ https://issues.apache.org/jira/browse/HBASE-8700?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Sergey Shelukhin updated HBASE-8700:
------------------------------------

    Description: 
The test can fail due to random number collision, claiming there are unreferenced elements
for obvious reasons (we rewrite some link). Original Accumulo test has one-stage generation
so it doesn't count unreferenced elements as failures, only undefined ones. With 200m longs
out of half-long range the probability of collision is approx 0.2%.

Moreover, without some way to debug, it's hard to debug what keys should be looked at in such
cases

  was:We have a failure in the above test where one element is missing yet it is not referenced-undefined,
so it looks like a head of the list with previous element disappearing; however such an eventuality
would require a link from tail to head to also disappear (otherwise there'd be a referenced-undefined
element), which is quite a coincidence. Without some way to debug, it's hard to debug what
keys should be looked at.

    
> IntegrationTestBigLinkedList can fail due to random number collision
> --------------------------------------------------------------------
>
>                 Key: HBASE-8700
>                 URL: https://issues.apache.org/jira/browse/HBASE-8700
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Sergey Shelukhin
>            Assignee: Sergey Shelukhin
>
> The test can fail due to random number collision, claiming there are unreferenced elements
for obvious reasons (we rewrite some link). Original Accumulo test has one-stage generation
so it doesn't count unreferenced elements as failures, only undefined ones. With 200m longs
out of half-long range the probability of collision is approx 0.2%.
> Moreover, without some way to debug, it's hard to debug what keys should be looked at
in such cases

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message