[
https://issues.apache.org/jira/browse/HADOOP-7896?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13168256#comment-13168256
]
Eli Collins commented on HADOOP-7896:
-------------------------------------
* In calculateExponentialTime this should be *, otherwise you'll get +/- 0%-150%
{code}
RAND.nextFloat() + 0.5
{code}
* In RetryInvocationHandler#invoke I'd pull the sleep loop out to a Util
method, eg sleepAtLeastIgnoreInterrupts (since it's useful elsewhere and we may
sleep 2x delayMillis)
* testFailoverBetweenMultipleStandbys needs a javadoc
Otherwise looks great.
> HA: if both NNs are in Standby mode, client needs to try failing back and
> forth several times with sleeps
> ---------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-7896
> URL: https://issues.apache.org/jira/browse/HADOOP-7896
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: ha, ipc
> Affects Versions: HA Branch (HDFS-1623)
> Reporter: Todd Lipcon
> Assignee: Aaron T. Myers
> Priority: Critical
> Attachments: HADOOP-7896-HDFS-1623.patch
>
>
> For a manual failover, there may be an intermediate state for a non-trivial
> amount of time where both NNs are in standby mode. Currently, the failover
> proxy will immediately failover on receiving this exception from the first
> NN, and when it hits the same exception on the second NN, it immediately
> fails. It should probably fail back and forth nearly indefinitely if both NNs
> are in Standby mode.
--
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
