hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stack <st...@duboce.net>
Subject Re: (无主题)
Date Tue, 10 Nov 2015 17:30:44 GMT
On Mon, Nov 9, 2015 at 10:52 PM, WangYQ <wangyongqiang0617@163.com> wrote:

> guys, i have some problems on seq_id
> 1. what is the motivation of adding seq_id as part of the bulkload files
Bulk loaded files either get no sequence id -- and so they are considered
AFTER all current edits in the Store -- or the bulk loaded file gets the
highest current sequence id so the edits are ordered to appear BEFORE any
current edit.  The file is assigned a sequence id. It applies to all
constituent edits.

> 2. why introduce the conf hbase.mapreduce.bulkload.assign.sequenceNumbers
> in class loadIncrementalHFiles, if this is true, every time we flush,
> should first do a flush
This is the switch for whether bulk files are ordered first or last.

> 3. now if we compact two files, seq_id are 3 and 7, the file after compact
> will get seq_id from region, may be 12
> if i make the file after compacted seq_id=7, the largerest seq_id among
> compacted files, will cause any serious problems?
IIUC, there is no problem. The sequence id assigned is always the highest
beyond any edit that may exist in current files (and therefore beyond the
highest possible sequence id a compaction could make).


> the version is hbase0.98.10
> thanks
> 发自 网易邮箱大师

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message