[ 
https://issues.apache.org/jira/browse/BOOKKEEPER-758?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13995965#comment-13995965
 ] 

Sijie Guo commented on BOOKKEEPER-758:
--------------------------------------

{quote}
 It sounds like you want to have a weaker version of ReadLastAddConfirmed that 
simply returns some valid value, not necessarily the last.
{quote}

Correct. The TryReadLastAddconfirmed API will callback with the first larger 
LAC that it received, which is most of readers want as they want a new LAC to 
move on.

{quote}
It looks like the callback can be called multiple times
{quote}

Yup. that is the intention behavior in TryReadLastAddConfirmedOp, so we could 
get the LAC updated when we received callbacks, as we already issued read lac 
to multiple bookies, we don't want to waste bandwidth.
But for the api that exposed to user, it would be only callback once. 

{quote}
In the test cases, don't we have to wait until the bookie is down when we kill 
a bookie?
{quote}

killBookie is a synchronous call. so the bookie should be down after it 
returned.

{quote}
Also, in the last test case, it shouldn't be numEntries here:
{quote}

Good catch. Will fix it in new patch.


> Add TryReadLastAddConfirmed API
> -------------------------------
>
>                 Key: BOOKKEEPER-758
>                 URL: https://issues.apache.org/jira/browse/BOOKKEEPER-758
>             Project: Bookkeeper
>          Issue Type: Improvement
>          Components: bookkeeper-client
>            Reporter: Sijie Guo
>            Assignee: Sijie Guo
>             Fix For: 4.3.0
>
>         Attachments: BOOKKEEPER-758.diff
>
>
> add TryReadLastConfirmed to read last confirmed without coverage checking, as 
> for readers which polls LAC, they just need LAC.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to