cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Avinash Lakshman (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-14) The initial arrive time should not be set to zero
Date Fri, 10 Apr 2009 15:31:15 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-14?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12697839#action_12697839
] 

Avinash Lakshman commented on CASSANDRA-14:
-------------------------------------------

Actually this will happen for the initial start maybe for a few seconds. But it will stabilize
very very soon. In any system in this nature one could think of this as a barrier synchronization
point where one waits for a few seconds for things to stabilize. It takes about less than
10 secs in a 100 node cluster for all this to kinda settle down. In my opinion a non-issue.

> The initial arrive time should not be set to zero
> -------------------------------------------------
>
>                 Key: CASSANDRA-14
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-14
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: zhu han
>            Assignee: Avinash Lakshman
>             Fix For: 0.3
>
>   Original Estimate: 0.5h
>  Remaining Estimate: 0.5h
>
> In line 253 of src.org.apache.cassandra.gms .FailureDetector, the initial arrive time
should not be zero. Otherwise, the failure detector would report the new joining node as dead
because the phi is positive infinity when the mean of arrive time is zero. 
> A proper value should be set as the initial value. I use half of the gossip period as
the initial value,  whose default configuration is 1000ms. 

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


Mime
View raw message