You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2010/01/21 04:12:54 UTC
[jira] Created: (CASSANDRA-728) refactor blocking io to use thread
(and queue) per conn pool, not thread per socket
refactor blocking io to use thread (and queue) per conn pool, not thread per socket
-----------------------------------------------------------------------------------
Key: CASSANDRA-728
URL: https://issues.apache.org/jira/browse/CASSANDRA-728
Project: Cassandra
Issue Type: Improvement
Components: Core
Reporter: Jonathan Ellis
Assignee: Jonathan Ellis
Priority: Minor
Fix For: 0.6
this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (CASSANDRA-728) refactor blocking io to use thread
(and queue) per conn pool, not thread per socket
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis updated CASSANDRA-728:
-------------------------------------
Attachment: 728.txt
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-728) refactor blocking io to use
thread (and queue) per conn pool, not thread per socket
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12803846#action_12803846 ]
Jonathan Ellis commented on CASSANDRA-728:
------------------------------------------
note that we don't have to null out old OTCs anymore from pool.reset(), since a closed OTC will automatically reconnect if/when write() is called again.
no attempt is made to clear out pool entries from decommissioned nodes (which is something we probably want to do eventually, but this does not represent a regression from the pre-CASSANDRA-705 status quo).
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-728) refactor blocking io to use
thread (and queue) per conn pool, not thread per socket
Posted by "Gary Dusbabek (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12803906#action_12803906 ]
Gary Dusbabek commented on CASSANDRA-728:
-----------------------------------------
+1
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-728) refactor blocking io to use
thread (and queue) per conn pool, not thread per socket
Posted by "Jaakko Laine (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12804468#action_12804468 ]
Jaakko Laine commented on CASSANDRA-728:
----------------------------------------
One lingering issue is what to do about connection pool to dead nodes? Connection pool is reset when node is convicted, but since gossiper tries to contact dead nodes too, connection pool will always be created again even if nobody is actively trying to write to the node. This causes connection pool to a node remain active even after the node has been removed from the ring completely. So basically we can either (1) alter the mechanism so that there will be no reconnect immediately (currently resetting the pool on convict does not make sense) and/or (2) kill connection pool for the node when it is removed or evicted from gossiper membership.
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (CASSANDRA-728) refactor blocking io to use thread
(and queue) per conn pool, not thread per socket
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis resolved CASSANDRA-728.
--------------------------------------
Resolution: Fixed
committed
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-728) refactor blocking io to use
thread (and queue) per conn pool, not thread per socket
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12804074#action_12804074 ]
Hudson commented on CASSANDRA-728:
----------------------------------
Integrated in Cassandra #332 (See [http://hudson.zones.apache.org/hudson/job/Cassandra/332/])
move connecting into OutboundTcpConnection so write does not block for connection startup
patch by jbellis; reviewed by gdusbabek for
> refactor blocking io to use thread (and queue) per conn pool, not thread per socket
> -----------------------------------------------------------------------------------
>
> Key: CASSANDRA-728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-728
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.6
>
> Attachments: 728.txt
>
>
> this means the pool can be in charge of reconnecting and no caller has to block for that part, as discussed in Jaakko's feedback to CASSANDRA-705
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.