accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ivan Bella (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-1903) Monitor creates a new ZK object to fetch gc status and doesn't wait for it to connect.
Date Sat, 16 Nov 2013 17:01:24 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-1903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13824533#comment-13824533
] 

Ivan Bella commented on ACCUMULO-1903:
--------------------------------------

http://zookeeper.apache.org/doc/r3.3.3/api/org/apache/zookeeper/ZooKeeper.html#ZooKeeper(java.lang.String,
int, org.apache.zookeeper.Watcher)

> Monitor creates a new ZK object to fetch gc status and doesn't wait for it to connect.
> --------------------------------------------------------------------------------------
>
>                 Key: ACCUMULO-1903
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-1903
>             Project: Accumulo
>          Issue Type: Bug
>          Components: monitor
>    Affects Versions: 1.5.0
>            Reporter: John Vines
>            Assignee: Eric Newton
>              Labels: newbie
>             Fix For: 1.5.1, 1.6.0
>
>
> May affect older versions, didn't check them for the same code, but it does exist in
1.5.1-SNAPSHOT and 1.6.0-SNAPSHOT currently.
> Seeing a warning in the monitor "Unable to contact the garbage collector at null", stemming
from a KeeperException KeeperErrorCode ConnectionLoss in fetchGcStatus().
> Initially I thought this was just a connection error, but I took a look at the code and
a quick google search led me to http://zookeeper-user.578899.n2.nabble.com/zookeeper-connection-loss-exception-occurs-on-new-created-ZooKeeper-instance-too-much-why-td6766831.html
> In the getchGcStatus() we create a new ZooKeeper object with every call (!) and then
immediately try to use it (!!) without waiting for it to finish establishing a connection.
Because of these, I think we should A. ensure it's connected before attempting to use and
B. try to reuse the same ZK object instead of creating a new one for each call.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message