You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geode.apache.org by Mario Ivanac <ma...@est.tech> on 2020/02/14 08:16:46 UTC
Creation of buckets for partitioned region
Hi geode dev,
we have observed following behavior, at creation of partitioned regions.
After partitioned region is created, initialization of bucket will take place:
* only at point when first data is inserted in region (bucket will be incrementally created for every added entry, till [max buckets]),
* or "select *" query is performed against that partitioned region (in this case all buckets [max buckets] are created at once).
Can you confirm that this is expected behavior?
Thanks,
Mario
Re: Creation of buckets for partitioned region
Posted by Xiaojian Zhou <gz...@pivotal.io>.
But these servers will be assigned buckets later by rebalance.
On Fri, Feb 14, 2020 at 9:25 AM Barry Oglesby <bo...@pivotal.io> wrote:
> Mario,
>
> Yes, a query execution causes the buckets to be created.
>
> Also, onRegion function execution causes them to be created as well.
>
> There is an API to create the buckets for a region called
> PartitionRegionHelper.assignBucketsToPartitions
>
> Be careful about when that method is called, though. Any servers that are
> started after it is called will contain no buckets.
>
> Thanks,
> Barry Oglesby
>
>
>
> On Fri, Feb 14, 2020 at 7:48 AM Udo Kohlmeyer <ud...@apache.com> wrote:
>
> > Hi there Mario,
> >
> > I can confirm the first observation. Buckets are created lazily. Upon
> > data create, buckets are created as required.
> >
> > --Udo
> >
> > On 2/14/20 12:16 AM, Mario Ivanac wrote:
> > > Hi geode dev,
> > >
> > > we have observed following behavior, at creation of partitioned
> regions.
> > >
> > > After partitioned region is created, initialization of bucket will take
> > place:
> > >
> > >
> > > * only at point when first data is inserted in region (bucket will
> > be incrementally created for every added entry, till [max buckets]),
> > > * or "select *" query is performed against that partitioned region
> > (in this case all buckets [max buckets] are created at once).
> > >
> > > Can you confirm that this is expected behavior?
> > >
> > > Thanks,
> > > Mario
> > >
> >
>
Re: Creation of buckets for partitioned region
Posted by Barry Oglesby <bo...@pivotal.io>.
Mario,
Yes, a query execution causes the buckets to be created.
Also, onRegion function execution causes them to be created as well.
There is an API to create the buckets for a region called
PartitionRegionHelper.assignBucketsToPartitions
Be careful about when that method is called, though. Any servers that are
started after it is called will contain no buckets.
Thanks,
Barry Oglesby
On Fri, Feb 14, 2020 at 7:48 AM Udo Kohlmeyer <ud...@apache.com> wrote:
> Hi there Mario,
>
> I can confirm the first observation. Buckets are created lazily. Upon
> data create, buckets are created as required.
>
> --Udo
>
> On 2/14/20 12:16 AM, Mario Ivanac wrote:
> > Hi geode dev,
> >
> > we have observed following behavior, at creation of partitioned regions.
> >
> > After partitioned region is created, initialization of bucket will take
> place:
> >
> >
> > * only at point when first data is inserted in region (bucket will
> be incrementally created for every added entry, till [max buckets]),
> > * or "select *" query is performed against that partitioned region
> (in this case all buckets [max buckets] are created at once).
> >
> > Can you confirm that this is expected behavior?
> >
> > Thanks,
> > Mario
> >
>
Re: Creation of buckets for partitioned region
Posted by Udo Kohlmeyer <ud...@apache.com>.
Hi there Mario,
I can confirm the first observation. Buckets are created lazily. Upon
data create, buckets are created as required.
--Udo
On 2/14/20 12:16 AM, Mario Ivanac wrote:
> Hi geode dev,
>
> we have observed following behavior, at creation of partitioned regions.
>
> After partitioned region is created, initialization of bucket will take place:
>
>
> * only at point when first data is inserted in region (bucket will be incrementally created for every added entry, till [max buckets]),
> * or "select *" query is performed against that partitioned region (in this case all buckets [max buckets] are created at once).
>
> Can you confirm that this is expected behavior?
>
> Thanks,
> Mario
>