[
https://issues.apache.org/jira/browse/RATIS-736?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16973620#comment-16973620
]
Tsz-wo Sze commented on RATIS-736:
----------------------------------
> ... How does it help in the overflow case?
When overflow happens, it changes unit only if multiplier < 2. We reduce the
multiplier by
{code}
multiply(2).multiply(multiplier / 2);
{code}
We do not want to change unit in order to minimize the error. When
TimeDuration t = 10ms and multiplier = Long.MAX/2. If we convert t to seconds,
t becomes 0.
> Add a new RetryPolicy similar to MultipleLinearRandomRetry in HDFS
> ------------------------------------------------------------------
>
> Key: RATIS-736
> URL: https://issues.apache.org/jira/browse/RATIS-736
> Project: Ratis
> Issue Type: Improvement
> Components: client
> Reporter: Tsz-wo Sze
> Assignee: Tsz-wo Sze
> Priority: Major
> Attachments: r736_20191106.patch
>
>
> In HDFS-3504, we first tried to implement exponential backoff retry policy
> but found that the later retries will need to wait for a very long time
> (since it is exponential). We then implement a MultipleLinearRandomRetry.
> We should do the same here.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
