[ 
https://issues.apache.org/jira/browse/HDFS-13373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16446358#comment-16446358
 ] 

Hanisha Koneru edited comment on HDFS-13373 at 4/20/18 9:00 PM:
----------------------------------------------------------------

Thanks for the patch [~bharatviswa].
# In {{PBHelper#convert(ReplicaTrashCommandProto)}}, the switch case should 
have a default case which throws AssertionError (please refer to 
{{BlockCommandProto convert(BlockCommand cmd)}}).
# In {{BPOfferService#processCommandFromActive}}, can we move the check whether 
the command is ReplicaTrashCommand inside the switch case so as to avoid making 
this check for every Datanode command.
{code}
final ReplicaTrashCommand rcmd = cmd instanceof ReplicaTrashCommand ? 
(ReplicaTrashCommand)cmd : null;
{code}
# We should handle the case where blocks are deleted and moved to Replica Trash 
after the expunge command is issued. These new blocks should not be removed 
from the Replica Trash. One option is to note down the timestamp when expunge 
command was received. All invalidated blocks moved to replica trash after this 
timestamp will not be expunged.


was (Author: hanishakoneru):
Thanks for the patch [~bharatviswa].
# In {{PBHelper#convert(ReplicaTrashCommandProto)}}, the switch case should 
have a default case which throws AssertionError (please refer to 
{{BlockCommandProto convert(BlockCommand cmd)}}).
# In {{BPOfferService#processCommandFromActive}}, can we move the check whether 
the command is ReplicaTrashCommand inside the switch case so as to avoid making 
this check for every Datanode command.
{code}
final ReplicaTrashCommand rcmd = cmd instanceof ReplicaTrashCommand ? 
(ReplicaTrashCommand)cmd : null;
{code}
# We should handle the case where blocks are deleted and moved to Replica Trash 
after the expunge command is issued. These new blocks should not be removed 
from the Replica Trash. One option is to note down the timestamp when expunge 
command was received. All invalidated blocks moved to replica trash after this 
timestamp should not be expunged.

> Handle expunge command on NN and DN
> -----------------------------------
>
>                 Key: HDFS-13373
>                 URL: https://issues.apache.org/jira/browse/HDFS-13373
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>            Reporter: Bharat Viswanadham
>            Assignee: Bharat Viswanadham
>            Priority: Major
>         Attachments: HDFS-13373-HDFS-12996.00.patch
>
>
> When DataNodes receive the DN_EXPUNGE command from Namenode, they will
> purge all the block replicas in replica-trash



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org

Reply via email to