[
https://issues.apache.org/jira/browse/HDDS-245?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16560000#comment-16560000
]
Ajay Kumar commented on HDDS-245:
---------------------------------
[~elek] thanks for working on this. Patch 4 looks good. Few questions:
# ContainerReportHandler#emitReplicationRequestEvent: Do we need to delay
replication of new containers, assuming some of there replicas are in process
of being reported?
# ReplicationActivityStatus: I like the idea of being it event driven and then
publishing its live status via metrics. However i don't see any class
publishing event to enable replication, Shall we enable it by default?
Few minor comments
# Node2ContainerMap#updateDatanodeMap: We seems to be replacing current
datanode containers with new argument, shall we rename the function accordingly?
# ContainerReportHandler: L85 : We dont have any formal recommandations around
use of streams but it has been flagged for performance in java 8 in some
jiras. Do we have any rough estimates for containers per node? May be we can
use traditional for-each if its smaller range or think about parallel streams
if it can benefit from parallelization.
> Handle ContainerReports in the SCM
> ----------------------------------
>
> Key: HDDS-245
> URL: https://issues.apache.org/jira/browse/HDDS-245
> Project: Hadoop Distributed Data Store
> Issue Type: Improvement
> Components: SCM
> Reporter: Elek, Marton
> Assignee: Elek, Marton
> Priority: Major
> Fix For: 0.2.1
>
> Attachments: HDDS-245.001.patch, HDDS-245.002.patch,
> HDDS-245.003.patch, HDDS-245.004.patch
>
>
> HDDS-242 provides a new class ContainerReportHandler which could handle the
> ContainerReports from the SCMHeartbeatDispatchere.
> HDDS-228 introduces a new map to store the container -> datanode[] mapping
> HDDS-199 implements the ReplicationManager which could send commands to the
> datanodes to copy the datanode.
> To wire all these components, we need to put implementation to the
> ContainerReportHandler (created in HDDS-242).
> The ContainerReportHandler should process the new ContainerReportForDatanode
> events, update the containerStateMap and node2ContainerMap and calculate the
> missing/duplicate containers and send the ReplicateCommand to the
> ReplicateManager.
--
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
