hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "HBase Review Board (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-3276) delete followed by a put with the same timestamp
Date Mon, 29 Nov 2010 19:05:22 GMT

    [ https://issues.apache.org/jira/browse/HBASE-3276?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12964880#action_12964880
] 

HBase Review Board commented on HBASE-3276:
-------------------------------------------

Message from: stack@duboce.net


bq.  On 2010-11-26 14:54:45, Ryan Rawson wrote:
bq.  > trunk/src/main/java/org/apache/hadoop/hbase/KeyValue.java, line 1373
bq.  > <http://review.cloudera.org/r/1252/diff/1/?file=17712#file17712line1373>
bq.  >
bq.  >     what are all the consequences for not sorting by type when using KVComparator?
 Does this mean we might create HFiles that not sorted properly, because the HFile comparator
uses the KeyComparator directly with ignoreType = false. 
bq.  >     
bq.  >     While in memstore we can rely on memstoreTS to roughly order by insertion time,
and the Put/Delete should probably work in that situation, you are talking about modifiying
a pretty core and important concept in how we sort things.
bq.  >     
bq.  >     There are other ways to reconcile bugs like this, one of them is to extend the
memstoreTS concept into the HFile and use that to reconcile during reads.  There is another
JIRA where I proposed this.  
bq.  >     
bq.  >     If we are talking about 0.92 and beyond I'd prefer building a solid base rather
than dangerous hacks like this.  Our unit tests are not extremely extensive, so while they
might pass, that doesnt guarantee lack of bad behaviour later on.
bq.  >
bq.  
bq.  Pranav Khaitan wrote:
bq.      Agree. As I mentioned, this is a major change and more thought needs to be given
to it.
bq.      
bq.      However, to resolve issues like HBASE-3276, we need either such a change or extend
the memstoreTS concept to HFile as you mentioned.
bq.      
bq.      About consequences, I don't see anything negative here. This change only affects
the sorting of keys having same row, col, timestamp. After this change, all keys with the
same row, col, ts will be sorted purely based on the order in which they were inserted. When
a memstore is flushed to HFile, the memstoreTS takes care of ordering. During compactions,
the KeyValueHeap breaks ties by using the sequence ids of storefiles.
bq.  
bq.  Ryan Rawson wrote:
bq.      the problem is you are now changing how things are ordered sometimes but not all
the time.  HFile directly uses the rawcomparator, instantiating it directly rather than getting
it via the code path you changed.  So now you create a memstore in this order:
bq.      
bq.      row,col,100,Put  (memstoreTS=1)
bq.      row,col,100,Delete (memstoreTS=2)
bq.      row,col,100,Put (memstoreTS=3)
bq.      
bq.      But the HFile comparator will consider this out of order since it doesnt know about
memstoreTS and it still expects things to be in a certain order.
bq.      
bq.      I'm a little wary of having implicit ordering in the HFiles... in your new scheme,
Put,Delete,Put are in that order 'just because they are', and the comparator cannot put them
back in order, and must rely on scanner order.  During compactions we would place keys in
order based on which files they came from, but they wouldn't themselves have an order.  Basically
we should get rid of 'type sorting' and use memstoreTS sorting in memory and implicit sorting
in the HFiles.  
bq.      
bq. 
bq.  
bq.  Pranav Khaitan wrote:
bq.      Right. I see that HFile does an extra check and throws an IOException when it gets
data out of the order. So if we go forward with this change, we will have to ensure that the
comparator used by HFile knows about this change. This change be achieved in two ways: Firstly,
by setting the default value of ignoreType = true. Alternately, the HFile can explicitly set
ignoreType = true.

@Ryan, you say "Basically we should get rid of 'type sorting' and use memstoreTS sorting in
memory and implicit sorting in the HFiles."  You think this a receipe we should adopt going
forward?   Giving it cursory thought, it would seem like it should work.   What about migrating
data that was sorted using current KV comparator?  Do we need to migrate files made using
old sort order?  Should we mark files that have this new ordering type 2 files?


- stack


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://review.cloudera.org/r/1252/#review1993
-----------------------------------------------------------





> delete followed by a put with the same timestamp
> ------------------------------------------------
>
>                 Key: HBASE-3276
>                 URL: https://issues.apache.org/jira/browse/HBASE-3276
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Kannan Muthukkaruppan
>            Assignee: Kannan Muthukkaruppan
>
> [Note: This issue is relevant only for cases that don't use the default "time" based
versions, but provide/manage versions explicitly.]
> The fix for HBASE-1485 ensures that if there are multiple puts with the same timestamp
the later one wins.
> However, if there is a delete for a specific timestamp, then the later put doesn't win.

> Say for example the following is the sequence of operations:
> put                         row/col/v1 - value1
> deleteColumn     row/col/v1
> put                         row/col/v1 - value2
> Without the deleteColumn(), HBASE-1485 ensures that "value2" is the winner.
> However, with the deleteColumn() thrown into the mix, the delete wins, and one cannot
insert a new value at that version. [The only, unsatisfactory, workaround at this point seems
to be trigger a major compaction. The major compact would clear the delete marker, and allow
new cells to be created with that version again.] 
> ---
> Seems like it might not be too complicated to extend the fix for HBASE-1485 to also respect
ordering between delete/put operations. I'll look into this further.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message