You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-issues@hadoop.apache.org by "Qi Liu (JIRA)" <ji...@apache.org> on 2009/12/19 02:50:18 UTC

[jira] Created: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

IPC server should set SO_KEEPALIVE to the accepted sockets
----------------------------------------------------------

                 Key: HADOOP-6459
                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
             Project: Hadoop Common
          Issue Type: Improvement
    Affects Versions: 0.20.1
            Reporter: Qi Liu


When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Hudson (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12805481#action_12805481 ] 

Hudson commented on HADOOP-6459:
--------------------------------

Integrated in Hadoop-Common-trunk #234 (See [http://hudson.zones.apache.org/hudson/job/Hadoop-Common-trunk/234/])
    Revert commit 903015 because it was tagged with the wrong jira number .


> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.0.20.patch, HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Hudson (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12804803#action_12804803 ] 

Hudson commented on HADOOP-6459:
--------------------------------

Integrated in Hadoop-Common-trunk-Commit #148 (See [http://hudson.zones.apache.org/hudson/job/Hadoop-Common-trunk-Commit/148/])
    . IPC client bug may cause rpc call hang. Contributed by Hairong Kuang.


> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.0.20.patch, HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Updated: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Qi Liu (JIRA)" <ji...@apache.org>.

     [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Qi Liu updated HADOOP-6459:
---------------------------

    Attachment: HADOOP-6459.patch

Patch to trunk.

> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Hudson (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12805278#action_12805278 ] 

Hudson commented on HADOOP-6459:
--------------------------------

Integrated in Hadoop-Common-trunk-Commit #149 (See [http://hudson.zones.apache.org/hudson/job/Hadoop-Common-trunk-Commit/149/])
    Revert commit 903015 because it was tagged with the wrong jira number .


> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.0.20.patch, HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Updated: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Qi Liu (JIRA)" <ji...@apache.org>.

     [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Qi Liu updated HADOOP-6459:
---------------------------

    Attachment: HADOOP-6459.0.20.patch

Patch for Hadoop-0.20

> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.0.20.patch, HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (HADOOP-6459) IPC server should set SO_KEEPALIVE to the accepted sockets

Posted by "Hudson (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-6459?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12804976#action_12804976 ] 

Hudson commented on HADOOP-6459:
--------------------------------

Integrated in Hadoop-Common-trunk #233 (See [http://hudson.zones.apache.org/hudson/job/Hadoop-Common-trunk/233/])
    . IPC client bug may cause rpc call hang. Contributed by Hairong Kuang.


> IPC server should set SO_KEEPALIVE to the accepted sockets
> ----------------------------------------------------------
>
>                 Key: HADOOP-6459
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6459
>             Project: Hadoop Common
>          Issue Type: Improvement
>    Affects Versions: 0.20.1
>            Reporter: Qi Liu
>         Attachments: HADOOP-6459.0.20.patch, HADOOP-6459.patch
>
>
> When there is some unexpected network outage or client-side issue, the network connection could stay alive forever, due to the lack of keep alive. We have observed such dead connection in name node. Though it has no known performance hits, but it is a waste of resource. Adding SO_KEEPALIVE to all accepted sockets should not be heavy, since the keep alive TCP packets will only be sent once every four hours.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.