[
https://issues.apache.org/jira/browse/BOOKKEEPER-163?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13227512#comment-13227512
]
Sijie Guo commented on BOOKKEEPER-163:
--------------------------------------
I am OK for the new patch, +1. Although it seems that hadoop-common jar
introduces lot of dependencies, which are not be used actually, it seems that
we don't have a better library to support hardlink than hadoop common jar. I
would like to hear other one's opinion.
> Prevent incorrect NoSuchLedgerException for readLastConfirmed.
> --------------------------------------------------------------
>
> Key: BOOKKEEPER-163
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-163
> Project: Bookkeeper
> Issue Type: Bug
> Components: bookkeeper-client, bookkeeper-server
> Affects Versions: 4.0.0
> Reporter: Sijie Guo
> Assignee: Ivan Kelly
> Fix For: 4.1.0
>
> Attachments: BOOKKEEPER-163.diff, BOOKKEEPER-163.diff,
> BOOKKEEPER-163.diff
>
>
> bookkeeper client treats NoSuchLedgerException as valid response when reading
> last confirmed. If NoSuchLedgerException is caused due to an empty directory
> in following cases, it is an incorrect response.
> 1) A disk is replaced or ledger index is removed by a sloppy admin.
> 2) A disk is not mounted when a bookie machine is restarted.
> We need a mechanism to prevent such incorrect responses.
> Ivan suggested to generate a instance key for each bookie and write it into
> the ledger directories. If a directory doesn't have the key, and other
> directories do, then it shouldn't start. This would also resolve the issue
> that someone starting a new bookie with the same IP as a bookie which has
> previously died.
--
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
