Varun Saxena commented on YARN-2136:

[~sunilg], on closer look at the code, when we close the RMStateStore, we close 
the ZKClients as well. Hence dispatcher queue draining shouldn't matter as 
ZKClient is already closed. 
Ofcourse as more than one thread is involved, AsyncDispatcher's event handling 
thread may pick up an event and process it(store/update operation) before RM 
can close RMStateStore(while switching to standby). 

But in my view, this shouldn't be too big an impact I think to warrant adding a 
FENCED state.

> RMStateStore can explicitly handle store/update events when fenced
> ------------------------------------------------------------------
>                 Key: YARN-2136
>                 URL: https://issues.apache.org/jira/browse/YARN-2136
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Jian He
> RMStateStore can choose to handle/ignore store/update events upfront instead 
> of invoking more ZK operations if state store is at fenced state. 

This message was sent by Atlassian JIRA

Reply via email to