[jira] [Commented] (ACCUMULO-3092) BatchWriter does not notice when tserver fails, continues to send mutations to it

["Eric Newton (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ACCUMULO-3092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14151879#comment-14151879 ] 

Eric Newton commented on ACCUMULO-3092:
---------------------------------------

I attempted to reproduce this:

# run 2 tservers
# create a new table
# add an impossible constraint to the metadata table
# attempt to insert a single mutation to the table
# observe hung request
# kill -STOP the tserver hosting the table
# after the tablet server loses its lock, remove the constraint on the metadata table

The tablet moves to the other server, and the mutation is written.

Anyone have a better idea?


> BatchWriter does not notice when tserver fails, continues to send mutations to it
> ---------------------------------------------------------------------------------
>
>                 Key: ACCUMULO-3092
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-3092
>             Project: Accumulo
>          Issue Type: Bug
>          Components: client
>    Affects Versions: 1.6.1
>         Environment: Accumulo 1.6.1-SNAPSHOT
>            Reporter: Dave Marion
>            Assignee: Eric Newton
>            Priority: Critical
>
> Client was writing mutations to a Tserver
> Tserver was busy and Master removed its lock
> Tserver went down
> Client continued to attempt to send mutations for 10+ hours failing each time with a MutationsRejectedException with a # server errors count of 1.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)