accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Keith Turner (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-2801) define tablet syncs walog for each tablet in a batch
Date Mon, 30 Jun 2014 16:24:25 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-2801?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14047809#comment-14047809
] 

Keith Turner commented on ACCUMULO-2801:
----------------------------------------

If not syncing after defining a tablet, the the following cases will need to be handled in
recovery.

One case is that the tablet is referenced in metadata table, but not defined in walog.  What
will happen in recovery code?

Another case is that an old instance of the tablet is defined in walog.  This could happen
in the following way.

 # tablet TAB1 is assigned to TSERVER1
 # TAB1 is defined in walog as DEF1
 # TAB1 is unassigned from TSERVER1
 # TAB1 is  assigned to TSERVER1
 # TAB1 is defined in walog as DEF2
 # TSERVER1 dies before walog is synced

At recovery time DEF1 will be seen instead of DEF2.  I have no idea if this is a problem,
I don't remember enough about the how the code works.  But I think this situation can occur
and we would need to determine if its a problem or not.

> define tablet syncs walog for each tablet in a batch
> ----------------------------------------------------
>
>                 Key: ACCUMULO-2801
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2801
>             Project: Accumulo
>          Issue Type: Bug
>    Affects Versions: 1.5.0, 1.5.1, 1.6.0
>            Reporter: Keith Turner
>
> When the batch writer sends a batch of mutations for N tablets that were not currently
using a walog, then define tablet will be called for each tablet.  Define tablet will sync
the walog.   In hadoop 2 hsync is used, which is much slower than hadoop1 sync calls.  If
hsync takes 50ms and there are 100 tablets, then this operation would take 5 secs.  The calls
to define tablet do not occur frequently, just when walogs switch or tablets are loaded so
the cost will be amortized.  Ideally there could be one walog sync call for all of the tablets
in a batch of mutations.



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

Mime
View raw message