[ 
https://issues.apache.org/jira/browse/FLUME-962?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13229836#comment-13229836
 ] 

Hari Shreedharan commented on FLUME-962:
----------------------------------------

I just wanted to clarify the requirements for this:

A) Allow the client code to specify more than one agents to connect to, perhaps 
in priority order.
B) Try to connect to highest priority host - if it works send data, else try 
next priority and so on.
C) In case of failure and the host no longer responds, then connect to highest 
priority host available.

The above implementation raises the following questions: If client is currently 
connected to priority A, and priority B comes online, where B has higher 
priority, should we simply continue on A unitl A fails or switch to B - This 
can cause high churn in connections and also will require us to have a thread 
keep track of which hosts are available(at least hosts with higher priority). 
Do we not have priorities at all? Take a set of hosts from the client, as 
backup to the one agent they want to connect to and connect to any one of those 
if available? In this case, we can have a thread which checks for our original 
agent, and sleeps for n seconds and checks again. The 2nd approach is simpler, 
we can even just allow the user to specify one failover agent, though the first 
one is more in line with our failover sink implmentation.
                
> Failover capability for Client SDK
> ----------------------------------
>
>                 Key: FLUME-962
>                 URL: https://issues.apache.org/jira/browse/FLUME-962
>             Project: Flume
>          Issue Type: Sub-task
>    Affects Versions: v1.0.0
>            Reporter: Kathleen Ting
>             Fix For: v1.2.0
>
>
> Need a client SDK for Flume that will allow clients to be able to failover 
> from one source to another in case the first agent is not available. This 
> will help in keeping client implementations developed outside of the project 
> decoupled from internal details of HA implementation within Flume.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to