Agree Mike. Right this would help to recover the pdx data.
From: Michael Stolz <mst...@pivotal.io>
To: dev@geode.apache.org; Hitesh Khamesra <hitesh...@yahoo.com>
Cc: Hitesh Khamesra <hkhame...@pivotal.io>; Udo Kohlmeyer
<ukohlme...@gmail.com>; Bruce Schuchardt <bschucha...@pivotal.io>; geode
<d...@geode.incubator.apache.org>
Sent: Monday, November 21, 2016 6:07 AM
Subject: Re: Review Request 53867: GEODE-2129 make pdxtype id random(not
sequential)
I think we need to be more forceful about requiring different distributed
system id's. This idea of using hashcode seems like a debugging problem
waiting to happen.
--
Mike Stolz
Principal Engineer, GemFire Product Manager
Mobile: 631-835-4771
On Fri, Nov 18, 2016 at 6:39 PM, Hitesh Khamesra <
hitesh...@yahoo.com.invalid> wrote:
> Before using hashcode for pdxType creation, we will check whether it is
> available or not. If it's not available then, we will use next available
> id. We have introduced this mechanism to harden the cases where the user
> ends up having two clusters with the same dsid. From: Vishal Rao <
> vishal...@gmail.com>
> To: dev@geode.apache.org; Hitesh Khamesra <hkhame...@pivotal.io>
> Cc: Udo Kohlmeyer <ukohlme...@gmail.com>; Bruce Schuchardt <
> bschucha...@pivotal.io>; geode <d...@geode.incubator.apache.org>
> Sent: Thursday, November 17, 2016 11:51 PM
> Subject: Re: Review Request 53867: GEODE-2129 make pdxtype id random(not
> sequential)
>
> Apologies if this is noise, but wouldn't using the hash code actually
> increase the likelihood of ID collisions?
>
> On Fri, Nov 18, 2016 at 6:11 AM, Hitesh Khamesra <hkhame...@pivotal.io>
> wrote:
>
> >
> > -----------------------------------------------------------
> > This is an automatically generated e-mail. To reply, visit:
> > https://reviews.apache.org/r/53867/
> > -----------------------------------------------------------
> >
> > Review request for geode, Bruce Schuchardt and Udo Kohlmeyer.
> >
> >
> > Repository: geode
> >
> >
> > Description
> > -------
> >
> > Right now pdxtype id has 4 bytes. Out of those 4 bytes, one byte reserved
> > for distributed-system-id, this make sure type id generated from
> different
> > cluster has different id. For rest of the three bytes we just increment
> > counter to create new pdxtype id. In the field, we have observed that
> > sometimes this pdxType Id collides. One reason could be they end up
> having
> > same distributed-system-id for the different cluster.
> > Thus to avoid a collision, we will be using hashcode of pdxType for three
> > bytes of pdxType id. That will reduce the possibility of collision.
> >
> >
> > Diffs
> > -----
> >
> > geode-core/src/main/java/org/apache/geode/pdx/internal/EnumId.java
> > 5d399eb
> > geode-core/src/main/java/org/apache/geode/pdx/internal/PdxType.java
> > b586f64
> > geode-core/src/main/java/org/apache/geode/pdx/internal/
> PeerTypeRegistration.java
> > 0226cca
> > geode-core/src/test/java/org/apache/geode/pdx/
> PdxAttributesJUnitTest.java
> > c45abce
> > geode-core/src/test/java/org/apache/geode/pdx/
> PdxSerializableJUnitTest.java
> > 5cd822c
> > geode-wan/src/test/java/org/apache/geode/internal/cache/
> wan/WANTestBase.java
> > f9c18ec
> > geode-wan/src/test/java/org/apache/geode/internal/cache/
> > wan/misc/PDXNewWanDUnitTest.java f1e8f42
> >
> > Diff: https://reviews.apache.org/r/53867/diff/
> >
> >
> > Testing
> > -------
> >
> >
> > Thanks,
> >
> > Hitesh Khamesra
> >
> >
>
>
> --
> "The World is a book, and those who do not travel read only a page." - St.
> Augustine.
>
>
>
