[ 
https://issues.apache.org/jira/browse/SOLR-5649?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13897313#comment-13897313
 ] 

Mark Miller edited comment on SOLR-5649 at 2/11/14 12:06 AM:
-------------------------------------------------------------

A couple additional items:

* We don't want to call connected() after updating the SolrZooKeeper instance 
in process - the new event thread will call on the connect event. We just want 
to start the reconnected thread and then that thread will process the thread 
death event and die.

* Greg mentioned to me he also noticed the connected status variable was now in 
a non synchronized block - it should be synchronized or changed to a volatile.


was (Author: [email protected]):
A couple additional items:

* We don't want to call connected() after updating the SolrZooKeeper instance 
in process - the new event thread will call on the connect event. We just want 
to start the reconnected thread and then that thread will process the thread 
death event and die.

* Greg mentioned to me he also noticed the connected status variable was now in 
an synchronized block - it should be synchronized or changed to a volatile.

> Small ConnectionManager improvements
> ------------------------------------
>
>                 Key: SOLR-5649
>                 URL: https://issues.apache.org/jira/browse/SOLR-5649
>             Project: Solr
>          Issue Type: Bug
>          Components: SolrCloud
>    Affects Versions: 4.6.1
>            Reporter: Gregory Chanan
>            Assignee: Mark Miller
>            Priority: Minor
>             Fix For: 5.0, 4.7
>
>
> I was just looking through the ConnectionManager and want to jot these down 
> before I forget them.  I'm happy to make a patch if someone thinks it's 
> valuable as well.
> - "clientConnected" doesn't seem to be read, can be eliminated
> - "state" is a private volatile variable, but only used in one function -- 
> seems unlikely private volatile is what is wanted
> - A comment explaining why disconnected() is not called in the case of 
> Expired would be helpful (Expired means we have already waited the timeout 
> period so we want to reject updates right away)



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to