[
https://issues.apache.org/jira/browse/BOOKKEEPER-422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13473232#comment-13473232
]
Flavio Junqueira commented on BOOKKEEPER-422:
---------------------------------------------
I don't mean to complicate this patch, but I was wondering if it makes sense to
split the subscription state into local and non-local. This way
#hasLocalSubscriptions can be implemented efficiently and it would be
independent of assumptions about the distribution of subscriptions. One way
would be to split top2sub2seq into two data structures, one for local and one
for non-local.
Even though Stu's argument makes sense, we may end up hitting corner cases and
if we can avoid them altogether, better for us.
Let me know if you guys think it makes sense.
> Simplify AbstractSubscriptionManager
> ------------------------------------
>
> Key: BOOKKEEPER-422
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-422
> Project: Bookkeeper
> Issue Type: Improvement
> Components: hedwig-server
> Reporter: Stu Hood
> Assignee: Stu Hood
> Priority: Minor
> Attachments: bk-422.diff, bk-422.diff
>
>
> It's difficult to maintain a duplicated/cached count of local subscribers,
> and we've experienced a few issues due to it getting out of sync with the
> actual set of subscribers. Since a count of local subscribers can be
> calculated from the top2sub2seq map, let's do that instead.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
