hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Lars Hofhansl (Issue Comment Edited) (JIRA)" <j...@apache.org>
Subject [jira] [Issue Comment Edited] (HBASE-5058) Allow HBaseAmin to use an existing connection
Date Tue, 20 Dec 2011 18:14:30 GMT

    [ https://issues.apache.org/jira/browse/HBASE-5058?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13173364#comment-13173364
] 

Lars Hofhansl edited comment on HBASE-5058 at 12/20/11 6:12 PM:
----------------------------------------------------------------

@Stack: I think that about sums it up. The complexity of layers and timeout stories are alleviated
somewhat by parent HBASE-4805 (no per HTable threadpool, HTablePool no longer needed).
I had a brief look at the first issue, unless I am missing something this would require a
nontrivial amount of refactoring. The simplest would be to do all network IO from the Connection
thread rather than the application thread (as described in HBASE-4956). Would need allow for
the client to synchronize and retrieve exceptions on/from a Future.

Short term, should we take HBASE-4805 all the way and add a getTable(...) method to HConnection?
(Or even further and add put/get/scan/etc methods that take a table name to HConnection?)

Long term a design based on asynchhbase with a thin synchronous layer on top is probably the
best option.
                
      was (Author: lhofhansl):
    @Stack: I think that about sums it up. The complexity of layers and timeout stories are
alleviated somewhat by parent HBASE-4805 (no per HTable threadpool, HTablePool no longer needed).
I had a brief look at the first issue, unless I am missing something this would require a
nontrivial amount of refactoring. The simplest would be to do all network IO from the Connection
thread rather than the application thread (as described in HBASE-4956). Would need allow for
the client to synchronize and retrieved exceptions on/from a Future.

Short term, should we take HBASE-4805 all the way and a getTable(...) method to HConnection?
(Or even further and add put/get/scan/etc methods that take a table name to HConnection?)

Long term a design based on asynchhbase with a thin synchronous layer on top is probably the
best option.
                  
> Allow HBaseAmin to use an existing connection
> ---------------------------------------------
>
>                 Key: HBASE-5058
>                 URL: https://issues.apache.org/jira/browse/HBASE-5058
>             Project: HBase
>          Issue Type: Sub-task
>          Components: client
>    Affects Versions: 0.94.0
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>            Priority: Minor
>             Fix For: 0.94.0
>
>         Attachments: 5058-v2.txt, 5058-v3.txt, 5058-v3.txt, 5058.txt
>
>
> What HBASE-4805 does for HTables, this should do for HBaseAdmin.
> Along with this the shared error handling and retrying between HBaseAdmin and HConnectionManager
can also be improved. I'll attach a first pass patch soon.

--
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

        

Mime
View raw message