hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From lars hofhansl <la...@apache.org>
Subject Re: Regressions on upgrading from 0.94 to 0.96
Date Mon, 24 Mar 2014 18:44:23 GMT
Yeah. Sorry. That's not good. I usually make sure that does not happen, but I missed that one.

I don't know what to do about issues like this. The RM's can't possibly watch all issues.
Another area where I find this a lot is with bug fixes that are committed to 0.96 or 0.98
and later, but the committer forgets about 0.94. I try to catch all these things, but some
will slip.

So here's a renewed call to all committers (including myself):
- There should never be a feature gap (i.e. a feature in 0.94, not in 0.96, but then again
0.98). All committers should check for that.
- Bugfixes should always be considered as to whether they'd useful for earlier releases. Unless
the area of code is new the likely answer is yes.

-- Lars



________________________________
 From: Stack <stack@duboce.net>
To: HBase Dev List <dev@hbase.apache.org> 
Sent: Monday, March 24, 2014 11:12 AM
Subject: Re: Regressions on upgrading from 0.94 to 0.96
 

On Sun, Mar 23, 2014 at 9:32 PM, Ishan Chhabra <ichhabra@rocketfuel.com>wrote:

> When looking at HBASE-8063<
> https://issues.apache.org/jira/browse/HBASE-8063>,
> I noticed that it was backported to 0.94
> (HBASE-8198<https://issues.apache.org/jira/browse/HBASE-8198>)
> but not to 0.96. I don't know if this was intentional or if it was just
> missed. Missing this patch could impact performance on certain kinds of
> queries.
>


We try to avoid having older versions have features that are then missing
from later versions.  The commit to 0.94 is done w/o comment/justification
in the issue.  Maybe a discussion was held elsewhere and not cited.




>
> Are there any other known features/fixes that did not get ported to 0.96
> but got ported to 0.94?
>

Hopefully none Ishan.  It is a but in our process if there is.
St.Ack
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message