[jira] [Commented] (BOOKKEEPER-918) Create Ledger NameSpace

["Sijie Guo (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BOOKKEEPER-918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15230320#comment-15230320 ] 

Sijie Guo commented on BOOKKEEPER-918:
--------------------------------------

[~jujjuri] could this be done in a upper level who is using the bookkeeper core library?

> Create Ledger NameSpace
> -----------------------
>
>                 Key: BOOKKEEPER-918
>                 URL: https://issues.apache.org/jira/browse/BOOKKEEPER-918
>             Project: Bookkeeper
>          Issue Type: New Feature
>          Components: bookkeeper-client, bookkeeper-server
>            Reporter: Venkateswararao Jujjuri (JV)
>            Assignee: Venkateswararao Jujjuri (JV)
>
> Bookkeeper doesn't really have a real NameSpace, it operates by ledgerId. I have opened and worked on https://issues.apache.org/jira/browse/BOOKKEEPER-873 which allows applications to pass-in LedgerId instead of BK client API picking one for them. But that really doesn't offer much flexibility.
> Since ledgerId is a long we can't have 128-bit UUIDs as ledgerIDs and this severely restricts the range if we are generating ledgerIDs through a random generator (https://issues.apache.org/jira/browse/BOOKKEEPER-864)
> Best solution for this is to offer a real name space, and I believe it is relatively simple than offering 128-bit ledgerIDs.
> Analogous to regular filesystem, we treat our ledgerId as inode-number. Create a new name space in ZK with the user provided pretty-name and provide simple mapping at ZK between name-to-ledgerId. This mapping is static, and never changes once created. i.e name-x always points ledgerId-z. Given this most part of bookie code doesn't need to be aware of this mapping, and changes mostly will be confined client code and in create/delete/list paths.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)