hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Suresh Srinivas (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-2191) Move datanodeMap from FSNamesystem to DatanodeManager
Date Tue, 26 Jul 2011 15:19:09 GMT

    [ https://issues.apache.org/jira/browse/HDFS-2191?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13071149#comment-13071149
] 

Suresh Srinivas commented on HDFS-2191:
---------------------------------------

> Is this correct? 10000 * heartbeatIntervalSeconds Why 10000? Should you be multiplying
it by 1000 when getting it from conf and not 10000 *?
Sorry I was not clear in my comments. The newly added code is:
{noformat}
+    
+    final long heartbeatIntervalSeconds = conf.getLong(
+        DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_KEY,
+        DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_DEFAULT);
+    final int heartbeatRecheckInterval = conf.getInt(
+        DFSConfigKeys.DFS_NAMENODE_HEARTBEAT_RECHECK_INTERVAL_KEY, 
+        DFSConfigKeys.DFS_NAMENODE_HEARTBEAT_RECHECK_INTERVAL_DEFAULT); // 5 minutes
+    this.heartbeatExpireInterval = 2 * heartbeatRecheckInterval
+        + 10000 * heartbeatIntervalSeconds;
{noformat}

{{heartbeatIntervalSeconds}} is actually in milliseconds. So can you call it {{heartbeatIntervalMilliSeconds}}.
Rest of the code looks correct.

> FSNamesystem currently requires quite a few methods in DatanodeMnaager, e.g. getNetworkTopology(),
registerDatanode(..), etc. The getDatanode(..) methods are only two of those methods. Let's
think about how to change the APIs after the code separation is done. Okay?
Sounds good

> I think you are talking about FSNamesystem.heartbeatCheck(). I incorrectly removed the
synchronization. Fixed it.
DatanodeManager#getDatanode() - how is the map access synchronized in this?


> It should accquire the write lock of DatanodeManager when we have read-write lock in
DatanodeManager in the future.
We should track this with a jira, since the synchronization that is in the current code is
not correct.



> Move datanodeMap from FSNamesystem to DatanodeManager
> -----------------------------------------------------
>
>                 Key: HDFS-2191
>                 URL: https://issues.apache.org/jira/browse/HDFS-2191
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: name-node
>            Reporter: Tsz Wo (Nicholas), SZE
>            Assignee: Tsz Wo (Nicholas), SZE
>         Attachments: h2191_20110723.patch, h2191_20110723b.patch, h2191_20110726.patch
>
>


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

        

Mime
View raw message