You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by Dave Latham <la...@davelink.net> on 2013/07/06 23:49:05 UTC
removing lock ids from RegionObserver
In HBASE-8877 I'm proposing to change row locks to be reentrant and thread
based rather than tracking lock ids. Part of that change would mean
altering the RegionObserver coprocessor. Specifically, the methods
preBatchMutate and postBatchMutate would have the parameter
MiniBatchOperationInProgress<Pair<Mutation,
Integer>> miniBatchOp become
MiniBatchOperationInProgress<Mutation> miniBatchOp
Does anyone rely on lock ids for this coprocessor or object to the change?
Is it safe to make this incompatible change for 0.96?
Thanks, Andrew, for already chiming in on the issue.
For more background, see that issue and discussion on HBASE-8806.
Dave
Re: removing lock ids from RegionObserver
Posted by Stack <st...@duboce.net>.
On Sat, Jul 6, 2013 at 2:49 PM, Dave Latham <la...@davelink.net> wrote:
> In HBASE-8877 I'm proposing to change row locks to be reentrant and thread
> based rather than tracking lock ids. Part of that change would mean
> altering the RegionObserver coprocessor. Specifically, the methods
> preBatchMutate and postBatchMutate would have the parameter
> MiniBatchOperationInProgress<Pair<Mutation,
> Integer>> miniBatchOp become
> MiniBatchOperationInProgress<Mutation> miniBatchOp
>
> Does anyone rely on lock ids for this coprocessor or object to the change?
> Is it safe to make this incompatible change for 0.96?
>
>
+1 for 0.96. (Why would an observer want the lockid anyways?)
St.Ack
Re: removing lock ids from RegionObserver
Posted by Andrew Purtell <ap...@apache.org>.
That part isnt changed.
Note lockRow/unlockRow is removed in 0.95+ and deprecated in 0.94.
On Saturday, July 6, 2013, Anoop John wrote:
> +1 from me for changing the CP signature by removing lockids.. CP will be
> able to acquire lock on rows? (lockRow will be still public?)
>
> -Anoop-
>
> On Sun, Jul 7, 2013 at 3:34 AM, Ted Yu <yuzhihong@gmail.com <javascript:;>>
> wrote:
>
> > +1 on the removal of lock Ids.
> >
> > On Sat, Jul 6, 2013 at 2:49 PM, Dave Latham <latham@davelink.net<javascript:;>>
> wrote:
> >
> > > In HBASE-8877 I'm proposing to change row locks to be reentrant and
> > thread
> > > based rather than tracking lock ids. Part of that change would mean
> > > altering the RegionObserver coprocessor. Specifically, the methods
> > > preBatchMutate and postBatchMutate would have the parameter
> > > MiniBatchOperationInProgress<Pair<Mutation,
> > > Integer>> miniBatchOp become
> > > MiniBatchOperationInProgress<Mutation> miniBatchOp
> > >
> > > Does anyone rely on lock ids for this coprocessor or object to the
> > change?
> > > Is it safe to make this incompatible change for 0.96?
> > >
> > > Thanks, Andrew, for already chiming in on the issue.
> > > For more background, see that issue and discussion on HBASE-8806.
> > >
> > > Dave
> > >
> >
>
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: removing lock ids from RegionObserver
Posted by Ted Yu <yu...@gmail.com>.
By lockRow did you mean this (in HRegion.java):
+ @VisibleForTesting void getRowLock(byte[] row) throws IOException {
Cheers
On Sat, Jul 6, 2013 at 10:04 PM, Anoop John <an...@gmail.com> wrote:
> +1 from me for changing the CP signature by removing lockids.. CP will be
> able to acquire lock on rows? (lockRow will be still public?)
>
> -Anoop-
>
> On Sun, Jul 7, 2013 at 3:34 AM, Ted Yu <yu...@gmail.com> wrote:
>
> > +1 on the removal of lock Ids.
> >
> > On Sat, Jul 6, 2013 at 2:49 PM, Dave Latham <la...@davelink.net> wrote:
> >
> > > In HBASE-8877 I'm proposing to change row locks to be reentrant and
> > thread
> > > based rather than tracking lock ids. Part of that change would mean
> > > altering the RegionObserver coprocessor. Specifically, the methods
> > > preBatchMutate and postBatchMutate would have the parameter
> > > MiniBatchOperationInProgress<Pair<Mutation,
> > > Integer>> miniBatchOp become
> > > MiniBatchOperationInProgress<Mutation> miniBatchOp
> > >
> > > Does anyone rely on lock ids for this coprocessor or object to the
> > change?
> > > Is it safe to make this incompatible change for 0.96?
> > >
> > > Thanks, Andrew, for already chiming in on the issue.
> > > For more background, see that issue and discussion on HBASE-8806.
> > >
> > > Dave
> > >
> >
>
Re: removing lock ids from RegionObserver
Posted by Anoop John <an...@gmail.com>.
+1 from me for changing the CP signature by removing lockids.. CP will be
able to acquire lock on rows? (lockRow will be still public?)
-Anoop-
On Sun, Jul 7, 2013 at 3:34 AM, Ted Yu <yu...@gmail.com> wrote:
> +1 on the removal of lock Ids.
>
> On Sat, Jul 6, 2013 at 2:49 PM, Dave Latham <la...@davelink.net> wrote:
>
> > In HBASE-8877 I'm proposing to change row locks to be reentrant and
> thread
> > based rather than tracking lock ids. Part of that change would mean
> > altering the RegionObserver coprocessor. Specifically, the methods
> > preBatchMutate and postBatchMutate would have the parameter
> > MiniBatchOperationInProgress<Pair<Mutation,
> > Integer>> miniBatchOp become
> > MiniBatchOperationInProgress<Mutation> miniBatchOp
> >
> > Does anyone rely on lock ids for this coprocessor or object to the
> change?
> > Is it safe to make this incompatible change for 0.96?
> >
> > Thanks, Andrew, for already chiming in on the issue.
> > For more background, see that issue and discussion on HBASE-8806.
> >
> > Dave
> >
>
Re: removing lock ids from RegionObserver
Posted by Ted Yu <yu...@gmail.com>.
+1 on the removal of lock Ids.
On Sat, Jul 6, 2013 at 2:49 PM, Dave Latham <la...@davelink.net> wrote:
> In HBASE-8877 I'm proposing to change row locks to be reentrant and thread
> based rather than tracking lock ids. Part of that change would mean
> altering the RegionObserver coprocessor. Specifically, the methods
> preBatchMutate and postBatchMutate would have the parameter
> MiniBatchOperationInProgress<Pair<Mutation,
> Integer>> miniBatchOp become
> MiniBatchOperationInProgress<Mutation> miniBatchOp
>
> Does anyone rely on lock ids for this coprocessor or object to the change?
> Is it safe to make this incompatible change for 0.96?
>
> Thanks, Andrew, for already chiming in on the issue.
> For more background, see that issue and discussion on HBASE-8806.
>
> Dave
>