You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Patrick McFadin (JIRA)" <ji...@apache.org> on 2012/11/13 15:42:12 UTC

[jira] [Created] (CASSANDRA-4954) Store trigger jar file dependencies in system keyspace

Patrick McFadin created CASSANDRA-4954:
------------------------------------------

             Summary: Store trigger jar file dependencies in system keyspace
                 Key: CASSANDRA-4954
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-4954
             Project: Cassandra
          Issue Type: New Feature
          Components: Core
            Reporter: Patrick McFadin
             Fix For: 1.3


The implementation of triggers requires the use of jar files to be stored in the local filesystem, per node. See ([https://issues.apache.org/jira/browse/CASSANDRA-1311]) 

I'm proposing that instead of using the filesystem, jars would be stored in the system keyspace. This would greatly reduce the operational complexity of implementing triggers on multiple node clusters. Some benefits would include:

* Triggers and dependencies would be part of a snapshot for backup and restore
* Every node would have a consistent version
* Everything needed for the trigger would be a part of bootstrap

Some details to start the conversation:
* Have a list of jars for each trigger be made available via JMX
* Expose the jar upload using CQL when trigger is created (USING FILE x) 
* Store the version number or date for each jar
* Implement a classloader trigger when jars are updated
* Provide some sort of guarantee that triggers aren't active until all nodes have the same version of jar(s)




--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (CASSANDRA-4954) Store trigger jar file dependencies in system keyspace

Posted by "Edward Capriolo (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/CASSANDRA-4954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13496360#comment-13496360 ] 

Edward Capriolo commented on CASSANDRA-4954:
--------------------------------------------

One way to handle the dynamic bits are "add trigger [source] [sourcespec]"

source spec could be "jar" in which case the source could be "jar:///a/b/c.jar or "http://server/a.jar" 
source spec could be "groovy" in which case the source is actually an embedded class "class triggerhandler { handleTrigger(event){} }"
...scala...
...clojure...
etc
                
> Store trigger jar file dependencies in system keyspace
> ------------------------------------------------------
>
>                 Key: CASSANDRA-4954
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-4954
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Core
>            Reporter: Patrick McFadin
>             Fix For: 1.3
>
>
> The implementation of triggers requires the use of jar files to be stored in the local filesystem, per node. See ([https://issues.apache.org/jira/browse/CASSANDRA-1311]) 
> I'm proposing that instead of using the filesystem, jars would be stored in the system keyspace. This would greatly reduce the operational complexity of implementing triggers on multiple node clusters. Some benefits would include:
> * Triggers and dependencies would be part of a snapshot for backup and restore
> * Every node would have a consistent version
> * Everything needed for the trigger would be a part of bootstrap
> Some details to start the conversation:
> * Have a list of jars for each trigger be made available via JMX
> * Expose the jar upload using CQL when trigger is created (USING FILE x) 
> * Store the version number or date for each jar
> * Implement a classloader trigger when jars are updated
> * Provide some sort of guarantee that triggers aren't active until all nodes have the same version of jar(s)

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (CASSANDRA-4954) Store trigger jar file dependencies in system keyspace

Posted by "Tupshin Harper (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/CASSANDRA-4954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13496343#comment-13496343 ] 

Tupshin Harper commented on CASSANDRA-4954:
-------------------------------------------

I envision a custom classloader that can read directly from a cf, and a "nodetool deploy trigger <jar>" command
As well as "nodetool [de]activate trigger" and
"nodetool configure trigger <triggername> [paramlist]"
Or some such. Each of these could also be performed by modifying the triggers cf directly if desired.

Enhancements could include deploying uncompiled jvm scripts in groovy, scala, etc with the same approach.
                
> Store trigger jar file dependencies in system keyspace
> ------------------------------------------------------
>
>                 Key: CASSANDRA-4954
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-4954
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Core
>            Reporter: Patrick McFadin
>             Fix For: 1.3
>
>
> The implementation of triggers requires the use of jar files to be stored in the local filesystem, per node. See ([https://issues.apache.org/jira/browse/CASSANDRA-1311]) 
> I'm proposing that instead of using the filesystem, jars would be stored in the system keyspace. This would greatly reduce the operational complexity of implementing triggers on multiple node clusters. Some benefits would include:
> * Triggers and dependencies would be part of a snapshot for backup and restore
> * Every node would have a consistent version
> * Everything needed for the trigger would be a part of bootstrap
> Some details to start the conversation:
> * Have a list of jars for each trigger be made available via JMX
> * Expose the jar upload using CQL when trigger is created (USING FILE x) 
> * Store the version number or date for each jar
> * Implement a classloader trigger when jars are updated
> * Provide some sort of guarantee that triggers aren't active until all nodes have the same version of jar(s)

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira