You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by Ted Yu <yu...@gmail.com> on 2011/09/28 18:29:21 UTC
backporting HBASE-3777 to 0.90
Hi,
Bright Fulton has volunteered to backport HBASE-3777 to 0.90
I endorse his effort.
If you have comment(s), please share.
I will open a new JIRA for this effort if this motion passes.
Thanks
Re: backporting HBASE-3777 to 0.90
Posted by Andrew Purtell <ap...@apache.org>.
+1
I volunteer to test the changes when a patch is ready.
- Andy
----- Original Message -----
> From: Ted Yu <yu...@gmail.com>
> To: dev@hbase.apache.org
> Cc:
> Sent: Wednesday, September 28, 2011 9:55 AM
> Subject: Re: backporting HBASE-3777 to 0.90
>
> One reason for my endorsement is that it would take 0.92 quite some time to
> reach the level of stability of 0.90.4
> I really think HBASE-3777 would benefit HBase users a lot, and reducing
> potential future inquiry about connection-related issues.
>
> Of course, backporting increases the amount of work for validation of 0.90.5
> But I think it is worth it.
>
> My two cents.
>
> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans
> <jd...@apache.org>wrote:
>
>> I'm -0 at the moment, it's a big patch to include in a point
> release.
>>
>> I'm glad the work was done tho because it means those interested (like
>> me) can directly patch it in and test it (at my own risk).
>>
>> J-D
>>
>> On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com> wrote:
>> > Hi,
>> > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
>> > I endorse his effort.
>> >
>> > If you have comment(s), please share.
>> >
>> > I will open a new JIRA for this effort if this motion passes.
>> >
>> > Thanks
>> >
>>
>
Re: backporting HBASE-3777 to 0.90
Posted by Shrijeet Paliwal <sh...@rocketfuel.com>.
Ted,
>>Please elaborate more on your cluster setup
We have 10 RS nodes , 1 Master and 1 Zookeeper
>>usage pattern and whether your
Live writes and reads but super heavy on reads. Cache hit is pretty high.
>>Application needed to be twisted after the new build went in.
No we did not change anything in application layer.
On Wed, Sep 28, 2011 at 2:46 PM, Ted Yu <yu...@gmail.com> wrote:
> Shrijeet:
> >> I dont have power to vote.
> I don't think so.
> The fact that you have been using 3777 is the best vote.
>
> Please elaborate more on your cluster setup, usage pattern and whether your
> application needed to be twisted after the new build went in.
>
> Thanks
>
> On Wed, Sep 28, 2011 at 2:41 PM, Shrijeet Paliwal
> <sh...@rocketfuel.com>wrote:
>
> > I dont have power to vote. But if it helps, we are running with
> HBASE-3777
> > on
> > top of 0.90.3 from the day it was committed. The qps on one of our data
> > centers is 50K.
> >
> > On Wed, Sep 28, 2011 at 2:30 PM, Ted Yu <yu...@gmail.com> wrote:
> >
> > > Thanks Andy for your support.
> > > Appreciate it.
> > >
> > > On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org>
> > > wrote:
> > >
> > > > > I'd switch from -1 to +1 if we can get +1s from people who have
> tried
> > > > > it on clusters with several different real existing apps written by
> > > > > several different teams.
> > > >
> > > >
> > > > This makes sense. My +1 was partly an agreement that I'd try it.
> > > >
> > > >
> > > > Best regards,
> > > >
> > > >
> > > > - Andy
> > > >
> > > > Problems worthy of attack prove their worth by hitting back. - Piet
> > Hein
> > > > (via Tom White)
> > > >
> > > >
> > > > ----- Original Message -----
> > > > > From: Todd Lipcon <to...@cloudera.com>
> > > > > To: dev@hbase.apache.org
> > > > > Cc: Andrew Purtell <ap...@apache.org>
> > > > > Sent: Wednesday, September 28, 2011 12:40 PM
> > > > > Subject: Re: backporting HBASE-3777 to 0.90
> > > > >
> > > > > I'd switch from -1 to +1 if we can get +1s from people who have
> tried
> > > > > it on clusters with several different real existing apps written by
> > > > > several different teams. EG if we can verify that the CIQ workload,
> > > > > the SU workload, and the TM workload all work with this patch with
> no
> > > > > adverse effects, seems reasonable to commit. But just passing unit
> > > > > tests doesn't seem like enough to me since it changes behavior in a
> > > > > way that is difficult to predict.
> > > > >
> > > > > -Todd
> > > > >
> > > > > On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com>
> wrote:
> > > > >> One option is to publish the backported patch which passes all
> unit
> > > > tests
> > > > >> and 'certified' by people who play trial on it.
> > > > >>
> > > > >> The switch proposed by Todd is nice but difficult to implement.
> > > > >>
> > > > >> Cheers
> > > > >>
> > > > >> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <todd@cloudera.com
> >
> > > > > wrote:
> > > > >>
> > > > >>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com>
> > > > > wrote:
> > > > >>> >>> We could query user@ before considering commit.
> > > > >>> > Let's do this.
> > > > >>> >
> > > > >>> > Objections ?
> > > > >>> >
> > > > >>>
> > > > >>> I don't think most users will know whether this will break them
> > > > > until
> > > > >>> it's "too late". Hence defaulting to current behavior,
> > > > > and letting
> > > > >>> people switch it if the current behavior isn't working for them.
> > > > >>>
> > > > >>> -Todd
> > > > >>> --
> > > > >>> Todd Lipcon
> > > > >>> Software Engineer, Cloudera
> > > > >>>
> > > > >>
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Todd Lipcon
> > > > > Software Engineer, Cloudera
> > > > >
> > > >
> > >
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
Shrijeet:
>> I dont have power to vote.
I don't think so.
The fact that you have been using 3777 is the best vote.
Please elaborate more on your cluster setup, usage pattern and whether your
application needed to be twisted after the new build went in.
Thanks
On Wed, Sep 28, 2011 at 2:41 PM, Shrijeet Paliwal
<sh...@rocketfuel.com>wrote:
> I dont have power to vote. But if it helps, we are running with HBASE-3777
> on
> top of 0.90.3 from the day it was committed. The qps on one of our data
> centers is 50K.
>
> On Wed, Sep 28, 2011 at 2:30 PM, Ted Yu <yu...@gmail.com> wrote:
>
> > Thanks Andy for your support.
> > Appreciate it.
> >
> > On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org>
> > wrote:
> >
> > > > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > > > it on clusters with several different real existing apps written by
> > > > several different teams.
> > >
> > >
> > > This makes sense. My +1 was partly an agreement that I'd try it.
> > >
> > >
> > > Best regards,
> > >
> > >
> > > - Andy
> > >
> > > Problems worthy of attack prove their worth by hitting back. - Piet
> Hein
> > > (via Tom White)
> > >
> > >
> > > ----- Original Message -----
> > > > From: Todd Lipcon <to...@cloudera.com>
> > > > To: dev@hbase.apache.org
> > > > Cc: Andrew Purtell <ap...@apache.org>
> > > > Sent: Wednesday, September 28, 2011 12:40 PM
> > > > Subject: Re: backporting HBASE-3777 to 0.90
> > > >
> > > > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > > > it on clusters with several different real existing apps written by
> > > > several different teams. EG if we can verify that the CIQ workload,
> > > > the SU workload, and the TM workload all work with this patch with no
> > > > adverse effects, seems reasonable to commit. But just passing unit
> > > > tests doesn't seem like enough to me since it changes behavior in a
> > > > way that is difficult to predict.
> > > >
> > > > -Todd
> > > >
> > > > On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
> > > >> One option is to publish the backported patch which passes all unit
> > > tests
> > > >> and 'certified' by people who play trial on it.
> > > >>
> > > >> The switch proposed by Todd is nice but difficult to implement.
> > > >>
> > > >> Cheers
> > > >>
> > > >> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com>
> > > > wrote:
> > > >>
> > > >>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com>
> > > > wrote:
> > > >>> >>> We could query user@ before considering commit.
> > > >>> > Let's do this.
> > > >>> >
> > > >>> > Objections ?
> > > >>> >
> > > >>>
> > > >>> I don't think most users will know whether this will break them
> > > > until
> > > >>> it's "too late". Hence defaulting to current behavior,
> > > > and letting
> > > >>> people switch it if the current behavior isn't working for them.
> > > >>>
> > > >>> -Todd
> > > >>> --
> > > >>> Todd Lipcon
> > > >>> Software Engineer, Cloudera
> > > >>>
> > > >>
> > > >
> > > >
> > > >
> > > > --
> > > > Todd Lipcon
> > > > Software Engineer, Cloudera
> > > >
> > >
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Shrijeet Paliwal <sh...@rocketfuel.com>.
I dont have power to vote. But if it helps, we are running with HBASE-3777 on
top of 0.90.3 from the day it was committed. The qps on one of our data
centers is 50K.
On Wed, Sep 28, 2011 at 2:30 PM, Ted Yu <yu...@gmail.com> wrote:
> Thanks Andy for your support.
> Appreciate it.
>
> On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org>
> wrote:
>
> > > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > > it on clusters with several different real existing apps written by
> > > several different teams.
> >
> >
> > This makes sense. My +1 was partly an agreement that I'd try it.
> >
> >
> > Best regards,
> >
> >
> > - Andy
> >
> > Problems worthy of attack prove their worth by hitting back. - Piet Hein
> > (via Tom White)
> >
> >
> > ----- Original Message -----
> > > From: Todd Lipcon <to...@cloudera.com>
> > > To: dev@hbase.apache.org
> > > Cc: Andrew Purtell <ap...@apache.org>
> > > Sent: Wednesday, September 28, 2011 12:40 PM
> > > Subject: Re: backporting HBASE-3777 to 0.90
> > >
> > > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > > it on clusters with several different real existing apps written by
> > > several different teams. EG if we can verify that the CIQ workload,
> > > the SU workload, and the TM workload all work with this patch with no
> > > adverse effects, seems reasonable to commit. But just passing unit
> > > tests doesn't seem like enough to me since it changes behavior in a
> > > way that is difficult to predict.
> > >
> > > -Todd
> > >
> > > On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
> > >> One option is to publish the backported patch which passes all unit
> > tests
> > >> and 'certified' by people who play trial on it.
> > >>
> > >> The switch proposed by Todd is nice but difficult to implement.
> > >>
> > >> Cheers
> > >>
> > >> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com>
> > > wrote:
> > >>
> > >>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com>
> > > wrote:
> > >>> >>> We could query user@ before considering commit.
> > >>> > Let's do this.
> > >>> >
> > >>> > Objections ?
> > >>> >
> > >>>
> > >>> I don't think most users will know whether this will break them
> > > until
> > >>> it's "too late". Hence defaulting to current behavior,
> > > and letting
> > >>> people switch it if the current behavior isn't working for them.
> > >>>
> > >>> -Todd
> > >>> --
> > >>> Todd Lipcon
> > >>> Software Engineer, Cloudera
> > >>>
> > >>
> > >
> > >
> > >
> > > --
> > > Todd Lipcon
> > > Software Engineer, Cloudera
> > >
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
Thanks Andy for your support.
Appreciate it.
On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org> wrote:
> > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > it on clusters with several different real existing apps written by
> > several different teams.
>
>
> This makes sense. My +1 was partly an agreement that I'd try it.
>
>
> Best regards,
>
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>
>
> ----- Original Message -----
> > From: Todd Lipcon <to...@cloudera.com>
> > To: dev@hbase.apache.org
> > Cc: Andrew Purtell <ap...@apache.org>
> > Sent: Wednesday, September 28, 2011 12:40 PM
> > Subject: Re: backporting HBASE-3777 to 0.90
> >
> > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > it on clusters with several different real existing apps written by
> > several different teams. EG if we can verify that the CIQ workload,
> > the SU workload, and the TM workload all work with this patch with no
> > adverse effects, seems reasonable to commit. But just passing unit
> > tests doesn't seem like enough to me since it changes behavior in a
> > way that is difficult to predict.
> >
> > -Todd
> >
> > On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
> >> One option is to publish the backported patch which passes all unit
> tests
> >> and 'certified' by people who play trial on it.
> >>
> >> The switch proposed by Todd is nice but difficult to implement.
> >>
> >> Cheers
> >>
> >> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com>
> > wrote:
> >>
> >>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com>
> > wrote:
> >>> >>> We could query user@ before considering commit.
> >>> > Let's do this.
> >>> >
> >>> > Objections ?
> >>> >
> >>>
> >>> I don't think most users will know whether this will break them
> > until
> >>> it's "too late". Hence defaulting to current behavior,
> > and letting
> >>> people switch it if the current behavior isn't working for them.
> >>>
> >>> -Todd
> >>> --
> >>> Todd Lipcon
> >>> Software Engineer, Cloudera
> >>>
> >>
> >
> >
> >
> > --
> > Todd Lipcon
> > Software Engineer, Cloudera
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
Unit test suite passed for patch v4.
Cheers
On Tue, Oct 4, 2011 at 8:44 PM, Ted Yu <yu...@gmail.com> wrote:
> Bright will upload patch v4 which would disable
> TestHCM.testManyNewConnectionsDoesnotOOME (disabled in TRUNK)
>
> Otherwise patch v3 is ready for wider validation effort.
>
> Thanks in advance for people who put this backport in their clusters.
>
>
> On Wed, Sep 28, 2011 at 3:45 PM, Gary Helmling <gh...@gmail.com>wrote:
>
>> On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org>
>> wrote:
>>
>> > > I'd switch from -1 to +1 if we can get +1s from people who have tried
>> > > it on clusters with several different real existing apps written by
>> > > several different teams.
>> >
>> >
>> > This makes sense. My +1 was partly an agreement that I'd try it.
>> >
>> >
>> I think I can agree to this as well. Despite my previous messages, my
>> bigger concern is subtler side effects from scope of the change, which has
>> had longer to bake in 0.92/trunk. If the patch can be verified in enough
>> real workloads, then I can support it with suitable messaging in the
>> release.
>>
>> Ultimately I think it would be good to more directly expose a cluster
>> "connection" as an entry point to the client APIs. Then client code could
>> share connections or not, however it chose. The current dependence on
>> Configuration and hiding under HTable leaves us in an odd gray area where
>> it's not clear if the current behavior is expected/needs to be supported,
>> or
>> just an implementation detail. But that's a whole separate discussion
>> that
>> may or may not be worth it. :)
>>
>
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
Bright will upload patch v4 which would disable
TestHCM.testManyNewConnectionsDoesnotOOME (disabled in TRUNK)
Otherwise patch v3 is ready for wider validation effort.
Thanks in advance for people who put this backport in their clusters.
On Wed, Sep 28, 2011 at 3:45 PM, Gary Helmling <gh...@gmail.com> wrote:
> On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org>
> wrote:
>
> > > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > > it on clusters with several different real existing apps written by
> > > several different teams.
> >
> >
> > This makes sense. My +1 was partly an agreement that I'd try it.
> >
> >
> I think I can agree to this as well. Despite my previous messages, my
> bigger concern is subtler side effects from scope of the change, which has
> had longer to bake in 0.92/trunk. If the patch can be verified in enough
> real workloads, then I can support it with suitable messaging in the
> release.
>
> Ultimately I think it would be good to more directly expose a cluster
> "connection" as an entry point to the client APIs. Then client code could
> share connections or not, however it chose. The current dependence on
> Configuration and hiding under HTable leaves us in an odd gray area where
> it's not clear if the current behavior is expected/needs to be supported,
> or
> just an implementation detail. But that's a whole separate discussion that
> may or may not be worth it. :)
>
Re: backporting HBASE-3777 to 0.90
Posted by Gary Helmling <gh...@gmail.com>.
On Wed, Sep 28, 2011 at 1:39 PM, Andrew Purtell <ap...@apache.org> wrote:
> > I'd switch from -1 to +1 if we can get +1s from people who have tried
> > it on clusters with several different real existing apps written by
> > several different teams.
>
>
> This makes sense. My +1 was partly an agreement that I'd try it.
>
>
I think I can agree to this as well. Despite my previous messages, my
bigger concern is subtler side effects from scope of the change, which has
had longer to bake in 0.92/trunk. If the patch can be verified in enough
real workloads, then I can support it with suitable messaging in the
release.
Ultimately I think it would be good to more directly expose a cluster
"connection" as an entry point to the client APIs. Then client code could
share connections or not, however it chose. The current dependence on
Configuration and hiding under HTable leaves us in an odd gray area where
it's not clear if the current behavior is expected/needs to be supported, or
just an implementation detail. But that's a whole separate discussion that
may or may not be worth it. :)
Re: backporting HBASE-3777 to 0.90
Posted by Andrew Purtell <ap...@apache.org>.
> I'd switch from -1 to +1 if we can get +1s from people who have tried
> it on clusters with several different real existing apps written by
> several different teams.
This makes sense. My +1 was partly an agreement that I'd try it.
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
----- Original Message -----
> From: Todd Lipcon <to...@cloudera.com>
> To: dev@hbase.apache.org
> Cc: Andrew Purtell <ap...@apache.org>
> Sent: Wednesday, September 28, 2011 12:40 PM
> Subject: Re: backporting HBASE-3777 to 0.90
>
> I'd switch from -1 to +1 if we can get +1s from people who have tried
> it on clusters with several different real existing apps written by
> several different teams. EG if we can verify that the CIQ workload,
> the SU workload, and the TM workload all work with this patch with no
> adverse effects, seems reasonable to commit. But just passing unit
> tests doesn't seem like enough to me since it changes behavior in a
> way that is difficult to predict.
>
> -Todd
>
> On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
>> One option is to publish the backported patch which passes all unit tests
>> and 'certified' by people who play trial on it.
>>
>> The switch proposed by Todd is nice but difficult to implement.
>>
>> Cheers
>>
>> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com>
> wrote:
>>
>>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com>
> wrote:
>>> >>> We could query user@ before considering commit.
>>> > Let's do this.
>>> >
>>> > Objections ?
>>> >
>>>
>>> I don't think most users will know whether this will break them
> until
>>> it's "too late". Hence defaulting to current behavior,
> and letting
>>> people switch it if the current behavior isn't working for them.
>>>
>>> -Todd
>>> --
>>> Todd Lipcon
>>> Software Engineer, Cloudera
>>>
>>
>
>
>
> --
> Todd Lipcon
> Software Engineer, Cloudera
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
Bright wasn't aware of the discussion so far.
Looks like we have two -1's, two +1's and one -0
On Wed, Sep 28, 2011 at 12:40 PM, Todd Lipcon <to...@cloudera.com> wrote:
> I'd switch from -1 to +1 if we can get +1s from people who have tried
> it on clusters with several different real existing apps written by
> several different teams. EG if we can verify that the CIQ workload,
> the SU workload, and the TM workload all work with this patch with no
> adverse effects, seems reasonable to commit. But just passing unit
> tests doesn't seem like enough to me since it changes behavior in a
> way that is difficult to predict.
>
> -Todd
>
> On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
> > One option is to publish the backported patch which passes all unit tests
> > and 'certified' by people who play trial on it.
> >
> > The switch proposed by Todd is nice but difficult to implement.
> >
> > Cheers
> >
> > On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com> wrote:
> >
> >> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com> wrote:
> >> >>> We could query user@ before considering commit.
> >> > Let's do this.
> >> >
> >> > Objections ?
> >> >
> >>
> >> I don't think most users will know whether this will break them until
> >> it's "too late". Hence defaulting to current behavior, and letting
> >> people switch it if the current behavior isn't working for them.
> >>
> >> -Todd
> >> --
> >> Todd Lipcon
> >> Software Engineer, Cloudera
> >>
> >
>
>
>
> --
> Todd Lipcon
> Software Engineer, Cloudera
>
Re: backporting HBASE-3777 to 0.90
Posted by Todd Lipcon <to...@cloudera.com>.
I'd switch from -1 to +1 if we can get +1s from people who have tried
it on clusters with several different real existing apps written by
several different teams. EG if we can verify that the CIQ workload,
the SU workload, and the TM workload all work with this patch with no
adverse effects, seems reasonable to commit. But just passing unit
tests doesn't seem like enough to me since it changes behavior in a
way that is difficult to predict.
-Todd
On Thu, Sep 29, 2011 at 4:33 AM, Ted Yu <yu...@gmail.com> wrote:
> One option is to publish the backported patch which passes all unit tests
> and 'certified' by people who play trial on it.
>
> The switch proposed by Todd is nice but difficult to implement.
>
> Cheers
>
> On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com> wrote:
>
>> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com> wrote:
>> >>> We could query user@ before considering commit.
>> > Let's do this.
>> >
>> > Objections ?
>> >
>>
>> I don't think most users will know whether this will break them until
>> it's "too late". Hence defaulting to current behavior, and letting
>> people switch it if the current behavior isn't working for them.
>>
>> -Todd
>> --
>> Todd Lipcon
>> Software Engineer, Cloudera
>>
>
--
Todd Lipcon
Software Engineer, Cloudera
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
One option is to publish the backported patch which passes all unit tests
and 'certified' by people who play trial on it.
The switch proposed by Todd is nice but difficult to implement.
Cheers
On Wed, Sep 28, 2011 at 12:29 PM, Todd Lipcon <to...@cloudera.com> wrote:
> On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com> wrote:
> >>> We could query user@ before considering commit.
> > Let's do this.
> >
> > Objections ?
> >
>
> I don't think most users will know whether this will break them until
> it's "too late". Hence defaulting to current behavior, and letting
> people switch it if the current behavior isn't working for them.
>
> -Todd
> --
> Todd Lipcon
> Software Engineer, Cloudera
>
Re: backporting HBASE-3777 to 0.90
Posted by Todd Lipcon <to...@cloudera.com>.
On Thu, Sep 29, 2011 at 4:27 AM, Ted Yu <yu...@gmail.com> wrote:
>>> We could query user@ before considering commit.
> Let's do this.
>
> Objections ?
>
I don't think most users will know whether this will break them until
it's "too late". Hence defaulting to current behavior, and letting
people switch it if the current behavior isn't working for them.
-Todd
--
Todd Lipcon
Software Engineer, Cloudera
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
>> We could query user@ before considering commit.
Let's do this.
Objections ?
On Wed, Sep 28, 2011 at 12:17 PM, Andrew Purtell <ap...@apache.org>wrote:
> > If anyone running 0.90 relies on the current behavior to
> > enforce separate connections (for whatever reason), using separate
> > Configuration instances, this would break that behavior and appear as a
> > regression right?
>
> Does anyone do this? We could query user@ before considering commit.
>
> > I agree it's nice to have the backport for those interested in
> > trying it.
>
> Another option to consider is putting it into a branch that Ted could
> maintain, if he's agreeable to that and someone is going to -1 putting this
> into 0.90.
>
> - Andy
>
>
>
>
>
> ----- Original Message -----
> > From: Gary Helmling <gh...@gmail.com>
> > To: dev@hbase.apache.org
> > Cc:
> > Sent: Wednesday, September 28, 2011 12:06 PM
> > Subject: Re: backporting HBASE-3777 to 0.90
> >
> > Changing the connection identity behavior in the middle of a release
> series
> > seems like a bad idea.
> >
> > The 0.20 releases did connection identity based on Configuration
> contents,
> > 0.90 changed this to Configuration instance identity, then 0.90.5 would
> be
> > going back to contents again (acknowledged with a smarter subset and
> guards
> > against changes)? If anyone running 0.90 relies on the current behavior
> to
> > enforce separate connections (for whatever reason), using separate
> > Configuration instances, this would break that behavior and appear as a
> > regression right?
> >
> > Changing these underlying assumptions in a minor release doesn't seem
> > right. I agree it's nice to have the backport for those interested in
> > trying it. But I'd need some convincing that the current 0.90 behavior
> is
> > completely broken rather than sub-optimal to agree to include it.
> >
> > --gh
> >
> >
> > On Wed, Sep 28, 2011 at 9:55 AM, Ted Yu <yu...@gmail.com> wrote:
> >
> >> One reason for my endorsement is that it would take 0.92 quite some
> time to
> >> reach the level of stability of 0.90.4
> >> I really think HBASE-3777 would benefit HBase users a lot, and reducing
> >> potential future inquiry about connection-related issues.
> >>
> >> Of course, backporting increases the amount of work for validation of
> >> 0.90.5
> >> But I think it is worth it.
> >>
> >> My two cents.
> >>
> >> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <
> jdcryans@apache.org
> >> >wrote:
> >>
> >> > I'm -0 at the moment, it's a big patch to include in a point
> > release.
> >> >
> >> > I'm glad the work was done tho because it means those interested
> > (like
> >> > me) can directly patch it in and test it (at my own risk).
> >> >
> >> > J-D
> >> >
> >> > On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com>
> > wrote:
> >> > > Hi,
> >> > > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
> >> > > I endorse his effort.
> >> > >
> >> > > If you have comment(s), please share.
> >> > >
> >> > > I will open a new JIRA for this effort if this motion passes.
> >> > >
> >> > > Thanks
> >> > >
> >> >
> >>
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
>> default to the broken way
LOL
On Wed, Sep 28, 2011 at 12:24 PM, Todd Lipcon <to...@cloudera.com> wrote:
> IMO ideal would be to somehow duplicate the codepaths - is it
> completely impossible to do so? Or could we hack in a flag like
> hbase.connpool.by.identity=true -- default to the broken way, and let
> users switch to the "new" codepath by toggling the boolean?
>
> Sorry I don't have enough context on the patch to know if the above is
> crazy-talk.
>
> -Todd
>
> On Thu, Sep 29, 2011 at 4:17 AM, Andrew Purtell <ap...@apache.org>
> wrote:
> >> If anyone running 0.90 relies on the current behavior to
> >> enforce separate connections (for whatever reason), using separate
> >> Configuration instances, this would break that behavior and appear as a
> >> regression right?
> >
> > Does anyone do this? We could query user@ before considering commit.
> >
> >> I agree it's nice to have the backport for those interested in
> >> trying it.
> >
> > Another option to consider is putting it into a branch that Ted could
> maintain, if he's agreeable to that and someone is going to -1 putting this
> into 0.90.
> >
> > - Andy
> >
> >
> >
> >
> >
> > ----- Original Message -----
> >> From: Gary Helmling <gh...@gmail.com>
> >> To: dev@hbase.apache.org
> >> Cc:
> >> Sent: Wednesday, September 28, 2011 12:06 PM
> >> Subject: Re: backporting HBASE-3777 to 0.90
> >>
> >> Changing the connection identity behavior in the middle of a release
> series
> >> seems like a bad idea.
> >>
> >> The 0.20 releases did connection identity based on Configuration
> contents,
> >> 0.90 changed this to Configuration instance identity, then 0.90.5 would
> be
> >> going back to contents again (acknowledged with a smarter subset and
> guards
> >> against changes)? If anyone running 0.90 relies on the current behavior
> to
> >> enforce separate connections (for whatever reason), using separate
> >> Configuration instances, this would break that behavior and appear as a
> >> regression right?
> >>
> >> Changing these underlying assumptions in a minor release doesn't seem
> >> right. I agree it's nice to have the backport for those interested in
> >> trying it. But I'd need some convincing that the current 0.90 behavior
> is
> >> completely broken rather than sub-optimal to agree to include it.
> >>
> >> --gh
> >>
> >>
> >> On Wed, Sep 28, 2011 at 9:55 AM, Ted Yu <yu...@gmail.com> wrote:
> >>
> >>> One reason for my endorsement is that it would take 0.92 quite some
> time to
> >>> reach the level of stability of 0.90.4
> >>> I really think HBASE-3777 would benefit HBase users a lot, and
> reducing
> >>> potential future inquiry about connection-related issues.
> >>>
> >>> Of course, backporting increases the amount of work for validation of
> >>> 0.90.5
> >>> But I think it is worth it.
> >>>
> >>> My two cents.
> >>>
> >>> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <
> jdcryans@apache.org
> >>> >wrote:
> >>>
> >>> > I'm -0 at the moment, it's a big patch to include in a point
> >> release.
> >>> >
> >>> > I'm glad the work was done tho because it means those interested
> >> (like
> >>> > me) can directly patch it in and test it (at my own risk).
> >>> >
> >>> > J-D
> >>> >
> >>> > On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com>
> >> wrote:
> >>> > > Hi,
> >>> > > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
> >>> > > I endorse his effort.
> >>> > >
> >>> > > If you have comment(s), please share.
> >>> > >
> >>> > > I will open a new JIRA for this effort if this motion passes.
> >>> > >
> >>> > > Thanks
> >>> > >
> >>> >
> >>>
> >>
> >
>
>
>
> --
> Todd Lipcon
> Software Engineer, Cloudera
>
Re: backporting HBASE-3777 to 0.90
Posted by Todd Lipcon <to...@cloudera.com>.
IMO ideal would be to somehow duplicate the codepaths - is it
completely impossible to do so? Or could we hack in a flag like
hbase.connpool.by.identity=true -- default to the broken way, and let
users switch to the "new" codepath by toggling the boolean?
Sorry I don't have enough context on the patch to know if the above is
crazy-talk.
-Todd
On Thu, Sep 29, 2011 at 4:17 AM, Andrew Purtell <ap...@apache.org> wrote:
>> If anyone running 0.90 relies on the current behavior to
>> enforce separate connections (for whatever reason), using separate
>> Configuration instances, this would break that behavior and appear as a
>> regression right?
>
> Does anyone do this? We could query user@ before considering commit.
>
>> I agree it's nice to have the backport for those interested in
>> trying it.
>
> Another option to consider is putting it into a branch that Ted could maintain, if he's agreeable to that and someone is going to -1 putting this into 0.90.
>
> - Andy
>
>
>
>
>
> ----- Original Message -----
>> From: Gary Helmling <gh...@gmail.com>
>> To: dev@hbase.apache.org
>> Cc:
>> Sent: Wednesday, September 28, 2011 12:06 PM
>> Subject: Re: backporting HBASE-3777 to 0.90
>>
>> Changing the connection identity behavior in the middle of a release series
>> seems like a bad idea.
>>
>> The 0.20 releases did connection identity based on Configuration contents,
>> 0.90 changed this to Configuration instance identity, then 0.90.5 would be
>> going back to contents again (acknowledged with a smarter subset and guards
>> against changes)? If anyone running 0.90 relies on the current behavior to
>> enforce separate connections (for whatever reason), using separate
>> Configuration instances, this would break that behavior and appear as a
>> regression right?
>>
>> Changing these underlying assumptions in a minor release doesn't seem
>> right. I agree it's nice to have the backport for those interested in
>> trying it. But I'd need some convincing that the current 0.90 behavior is
>> completely broken rather than sub-optimal to agree to include it.
>>
>> --gh
>>
>>
>> On Wed, Sep 28, 2011 at 9:55 AM, Ted Yu <yu...@gmail.com> wrote:
>>
>>> One reason for my endorsement is that it would take 0.92 quite some time to
>>> reach the level of stability of 0.90.4
>>> I really think HBASE-3777 would benefit HBase users a lot, and reducing
>>> potential future inquiry about connection-related issues.
>>>
>>> Of course, backporting increases the amount of work for validation of
>>> 0.90.5
>>> But I think it is worth it.
>>>
>>> My two cents.
>>>
>>> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <jdcryans@apache.org
>>> >wrote:
>>>
>>> > I'm -0 at the moment, it's a big patch to include in a point
>> release.
>>> >
>>> > I'm glad the work was done tho because it means those interested
>> (like
>>> > me) can directly patch it in and test it (at my own risk).
>>> >
>>> > J-D
>>> >
>>> > On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com>
>> wrote:
>>> > > Hi,
>>> > > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
>>> > > I endorse his effort.
>>> > >
>>> > > If you have comment(s), please share.
>>> > >
>>> > > I will open a new JIRA for this effort if this motion passes.
>>> > >
>>> > > Thanks
>>> > >
>>> >
>>>
>>
>
--
Todd Lipcon
Software Engineer, Cloudera
Re: backporting HBASE-3777 to 0.90
Posted by Andrew Purtell <ap...@apache.org>.
> If anyone running 0.90 relies on the current behavior to
> enforce separate connections (for whatever reason), using separate
> Configuration instances, this would break that behavior and appear as a
> regression right?
Does anyone do this? We could query user@ before considering commit.
> I agree it's nice to have the backport for those interested in
> trying it.
Another option to consider is putting it into a branch that Ted could maintain, if he's agreeable to that and someone is going to -1 putting this into 0.90.
- Andy
----- Original Message -----
> From: Gary Helmling <gh...@gmail.com>
> To: dev@hbase.apache.org
> Cc:
> Sent: Wednesday, September 28, 2011 12:06 PM
> Subject: Re: backporting HBASE-3777 to 0.90
>
> Changing the connection identity behavior in the middle of a release series
> seems like a bad idea.
>
> The 0.20 releases did connection identity based on Configuration contents,
> 0.90 changed this to Configuration instance identity, then 0.90.5 would be
> going back to contents again (acknowledged with a smarter subset and guards
> against changes)? If anyone running 0.90 relies on the current behavior to
> enforce separate connections (for whatever reason), using separate
> Configuration instances, this would break that behavior and appear as a
> regression right?
>
> Changing these underlying assumptions in a minor release doesn't seem
> right. I agree it's nice to have the backport for those interested in
> trying it. But I'd need some convincing that the current 0.90 behavior is
> completely broken rather than sub-optimal to agree to include it.
>
> --gh
>
>
> On Wed, Sep 28, 2011 at 9:55 AM, Ted Yu <yu...@gmail.com> wrote:
>
>> One reason for my endorsement is that it would take 0.92 quite some time to
>> reach the level of stability of 0.90.4
>> I really think HBASE-3777 would benefit HBase users a lot, and reducing
>> potential future inquiry about connection-related issues.
>>
>> Of course, backporting increases the amount of work for validation of
>> 0.90.5
>> But I think it is worth it.
>>
>> My two cents.
>>
>> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <jdcryans@apache.org
>> >wrote:
>>
>> > I'm -0 at the moment, it's a big patch to include in a point
> release.
>> >
>> > I'm glad the work was done tho because it means those interested
> (like
>> > me) can directly patch it in and test it (at my own risk).
>> >
>> > J-D
>> >
>> > On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com>
> wrote:
>> > > Hi,
>> > > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
>> > > I endorse his effort.
>> > >
>> > > If you have comment(s), please share.
>> > >
>> > > I will open a new JIRA for this effort if this motion passes.
>> > >
>> > > Thanks
>> > >
>> >
>>
>
Re: backporting HBASE-3777 to 0.90
Posted by Gary Helmling <gh...@gmail.com>.
Changing the connection identity behavior in the middle of a release series
seems like a bad idea.
The 0.20 releases did connection identity based on Configuration contents,
0.90 changed this to Configuration instance identity, then 0.90.5 would be
going back to contents again (acknowledged with a smarter subset and guards
against changes)? If anyone running 0.90 relies on the current behavior to
enforce separate connections (for whatever reason), using separate
Configuration instances, this would break that behavior and appear as a
regression right?
Changing these underlying assumptions in a minor release doesn't seem
right. I agree it's nice to have the backport for those interested in
trying it. But I'd need some convincing that the current 0.90 behavior is
completely broken rather than sub-optimal to agree to include it.
--gh
On Wed, Sep 28, 2011 at 9:55 AM, Ted Yu <yu...@gmail.com> wrote:
> One reason for my endorsement is that it would take 0.92 quite some time to
> reach the level of stability of 0.90.4
> I really think HBASE-3777 would benefit HBase users a lot, and reducing
> potential future inquiry about connection-related issues.
>
> Of course, backporting increases the amount of work for validation of
> 0.90.5
> But I think it is worth it.
>
> My two cents.
>
> On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <jdcryans@apache.org
> >wrote:
>
> > I'm -0 at the moment, it's a big patch to include in a point release.
> >
> > I'm glad the work was done tho because it means those interested (like
> > me) can directly patch it in and test it (at my own risk).
> >
> > J-D
> >
> > On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com> wrote:
> > > Hi,
> > > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
> > > I endorse his effort.
> > >
> > > If you have comment(s), please share.
> > >
> > > I will open a new JIRA for this effort if this motion passes.
> > >
> > > Thanks
> > >
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Ted Yu <yu...@gmail.com>.
One reason for my endorsement is that it would take 0.92 quite some time to
reach the level of stability of 0.90.4
I really think HBASE-3777 would benefit HBase users a lot, and reducing
potential future inquiry about connection-related issues.
Of course, backporting increases the amount of work for validation of 0.90.5
But I think it is worth it.
My two cents.
On Wed, Sep 28, 2011 at 9:47 AM, Jean-Daniel Cryans <jd...@apache.org>wrote:
> I'm -0 at the moment, it's a big patch to include in a point release.
>
> I'm glad the work was done tho because it means those interested (like
> me) can directly patch it in and test it (at my own risk).
>
> J-D
>
> On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com> wrote:
> > Hi,
> > Bright Fulton has volunteered to backport HBASE-3777 to 0.90
> > I endorse his effort.
> >
> > If you have comment(s), please share.
> >
> > I will open a new JIRA for this effort if this motion passes.
> >
> > Thanks
> >
>
Re: backporting HBASE-3777 to 0.90
Posted by Jean-Daniel Cryans <jd...@apache.org>.
I'm -0 at the moment, it's a big patch to include in a point release.
I'm glad the work was done tho because it means those interested (like
me) can directly patch it in and test it (at my own risk).
J-D
On Wed, Sep 28, 2011 at 9:29 AM, Ted Yu <yu...@gmail.com> wrote:
> Hi,
> Bright Fulton has volunteered to backport HBASE-3777 to 0.90
> I endorse his effort.
>
> If you have comment(s), please share.
>
> I will open a new JIRA for this effort if this motion passes.
>
> Thanks
>