[jira] [Comment Edited] (HDFS-14649) Add suspect probe for DeadNodeDetector
[
https://issues.apache.org/jira/browse/HDFS-14649?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16982570#comment-16982570
]
Lisheng Sun edited comment on HDFS-14649 at 11/26/19 2:53 PM:
--
{quote}
We can introduce one additional variable in Detector class, like
disabledProbeThreadForTest and then reset this value in the UT. We don't need
to add new config for this.
{quote}
the v005 patch fixed this problem. To reset disabledProbeThreadForTest in the
UT controls whether to start probeThread.
was (Author: leosun08):
{quote}
We can introduce one additional variable in Detector class, like
disabledProbeThreadForTest and then reset this value in the UT. We don't need
to add new config for this.
{quote}
the v004 patch fixed this problem. To reset disabledProbeThreadForTest in the
UT controls whether to start probeThread.
> Add suspect probe for DeadNodeDetector
> --
>
> Key: HDFS-14649
> URL: https://issues.apache.org/jira/browse/HDFS-14649
> Project: Hadoop HDFS
> Issue Type: Sub-task
>Reporter: Lisheng Sun
>Assignee: Lisheng Sun
>Priority: Major
> Attachments: HDFS-14649.001.patch, HDFS-14649.002.patch,
> HDFS-14649.003.patch, HDFS-14649.004.patch, HDFS-14649.005.patch
>
>
> Add suspect probe for DeadNodeDetector.
> when some DataNode of the block is found to inaccessible, put the DataNode
> into it will be placed in the Suspicious Node list. Because when DataNode is
> not accessible, it is likely that the replica has been removed from the
> DataNode.Therefore, it needs to be confirmed by re-probing and requires a
> higher priority processing.
> And when DataNode is placed in the Suspicious Node list, it is accessed by
> other dfsinputstream.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Comment Edited] (HDFS-14649) Add suspect probe for DeadNodeDetector
[
https://issues.apache.org/jira/browse/HDFS-14649?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16981612#comment-16981612
]
Yiqun Lin edited comment on HDFS-14649 at 11/25/19 2:41 PM:
[~leosun08], we can add a corner case that the node is added in suspect list
but probed succeed, then removed from the suspect list. I mean this is the case
of misreported node. Maybe cause by the restart node behavior in some real
scenarios.
Another one, can you address following this comment?
{quote}
Can we rename dfsInputStreamNodes to suspectAndDeadNodes? It will look more
readable.
{quote}
was (Author: linyiqun):
[~leosun08], we can add a corner case that the node is added in suspect list
and then removed from the suspect list. I mean this is the case of misreported
node. Maybe cause by the restart node behavior in some real scenarios.
Another one, can you address following this comment?
{quote}
Can we rename dfsInputStreamNodes to suspectAndDeadNodes? It will look more
readable.
{quote}
> Add suspect probe for DeadNodeDetector
> --
>
> Key: HDFS-14649
> URL: https://issues.apache.org/jira/browse/HDFS-14649
> Project: Hadoop HDFS
> Issue Type: Sub-task
>Reporter: Lisheng Sun
>Assignee: Lisheng Sun
>Priority: Major
> Attachments: HDFS-14649.001.patch, HDFS-14649.002.patch
>
>
> Add suspect probe for DeadNodeDetector.
> when some DataNode of the block is found to inaccessible, put the DataNode
> into it will be placed in the Suspicious Node list. Because when DataNode is
> not accessible, it is likely that the replica has been removed from the
> DataNode.Therefore, it needs to be confirmed by re-probing and requires a
> higher priority processing.
> And when DataNode is placed in the Suspicious Node list, it is accessed by
> other dfsinputstream.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
