cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mck SembWever (JIRA)" <j...@apache.org>
Subject [jira] [Issue Comment Edited] (CASSANDRA-2388) ColumnFamilyRecordReader fails for a given split because a host is down, even if records could reasonably be read from other replica.
Date Tue, 30 Aug 2011 21:23:13 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-2388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13094097#comment-13094097
] 

Mck SembWever edited comment on CASSANDRA-2388 at 8/30/11 9:22 PM:
-------------------------------------------------------------------

This approach isn't really working for me and was committed too quickly i believe.

bq. Although the documentation on inputSplit.getLocations() is a little thin as to whether
this restricts which trackers it should run on or whether is just a preference

Tasks are still being evenly distributed around the ring regardless of what the ColumnFamilySplit.locations
is.

The chance of a task actually working is RF/N. Therefore the chances of a blacklisted node
are high. Worse is that the whole ring can quickly become blacklisted.

http://abel-perez.com/hadoop-task-assignment has an interesting section in it explaining how
the task assignment is supposed to work (and that data locality is preferred but not a requirement).
Could ColumnFamilySplit.locations be in the wrong format? (eg they should ip not hostname?).

      was (Author: michaelsembwever):
    This approach isn't really working for me and was committed too quickly i believe.

bq. Although the documentation on inputSplit.getLocations() is a little thin as to whether
this restricts which trackers it should run on or whether is just a preference

Tasks are still being evenly distributed around the ring regardless of what the ColumnFamilySplit.locations
is.

The chance of a task actually working is RF/N. Therefore the chances of a blacklisted node
are high. Worse is that the whole ring can quickly become blacklisted.
  
> ColumnFamilyRecordReader fails for a given split because a host is down, even if records
could reasonably be read from other replica.
> -------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-2388
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-2388
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Hadoop
>    Affects Versions: 0.6
>            Reporter: Eldon Stegall
>            Assignee: Mck SembWever
>              Labels: hadoop, inputformat
>             Fix For: 0.7.9, 0.8.5
>
>         Attachments: 0002_On_TException_try_next_split.patch, CASSANDRA-2388-addition1.patch,
CASSANDRA-2388-extended.patch, CASSANDRA-2388.patch, CASSANDRA-2388.patch, CASSANDRA-2388.patch,
CASSANDRA-2388.patch
>
>
> ColumnFamilyRecordReader only tries the first location for a given split. We should try
multiple locations for a given split.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message