Done! https://issues.apache.org/jira/browse/GEODE-8019
-Alberto G. ________________________________ From: Dave Barnes <dav...@vmware.com> Sent: Friday, April 24, 2020 5:13 PM To: user@geode.apache.org <user@geode.apache.org> Subject: Re: Delta propagation and transactions Good suggestion, Alberto. Please open a JIRA ticket so documenters can follow up. Thanks! From: Alberto Gomez <alberto.go...@est.tech> Reply-To: "user@geode.apache.org" <user@geode.apache.org> Date: Friday, April 24, 2020 at 6:58 AM To: "user@geode.apache.org" <user@geode.apache.org> Subject: Re: Delta propagation and transactions Thanks a lot for your explanations. Now it is clear. It think it will be worth to update the documentation with some more information to make crystal clear what is used from delta propagation in transactions. Regards, -Alberto G. ________________________________ From: Eric Shu <e...@pivotal.io> Sent: Thursday, April 23, 2020 8:51 PM To: user@geode.apache.org <user@geode.apache.org> Subject: Re: Delta propagation and transactions Not sure if there is a particular reason or not. However, currently tx implementation does not track delta once applied to TXEntryState in a transaction. Maybe a transaction tries to guarantee the entries in the transaction will be applied to all replicated copies when there is a concurrent non-transactional operation as well? Delta would only update parts of an entry, but the other non-transactional operation could have changed the other parts. Regards, Eric On Thu, Apr 23, 2020 at 1:07 AM Alberto Gomez <alberto.go...@est.tech> wrote: Thanks to both for the clarifications. I still have one remaining question regarding this topic. Is there a technical reason why that part of the transaction cannot use delta propagation? Best regards, -Alberto G. ________________________________ From: Eric Shu <e...@pivotal.io<mailto:e...@pivotal.io>> Sent: Thursday, April 23, 2020 1:36 AM To: user@geode.apache.org<mailto:user@geode.apache.org> <user@geode.apache.org<mailto:user@geode.apache.org>> Subject: Re: Delta propagation and transactions Anil, The tx data propagation between accessor and data store (part of the transactions) could use delta as well; and if the transaction originator does not host the primary copy of the data, it can also use delta to update within a transaction. Your understanding of client/server is correct. Regards, Eric On Wed, Apr 22, 2020 at 4:05 PM Anilkumar Gingade <aging...@pivotal.io<mailto:aging...@pivotal.io>> wrote: Eric, Is that, the tx data propagation between peers/nodes (part of the transactions) doesn't use delta? Is it used when tx is originated in client (between client and server), but not between peers....I am trying to see where delta is used. -Anil. On Wed, Apr 22, 2020 at 11:58 AM Eric Shu <e...@pivotal.io<mailto:e...@pivotal.io>> wrote: The transaction commits on a transaction host node (primary copy for partitioned region, or one of the copies for replicate region), the entry operations within this commit needs to be distributed to partitioned region's redundant copies, or other replicates for replicate region. This distribution can not use delta propagation. Regards, Eric On Wed, Apr 22, 2020 at 11:50 AM Alberto Gomez <alberto.go...@est.tech> wrote: Hi Eric, What do you mean by "the transactional commit to the replicas can not use delta propagation". Thanks, Alberto ________________________________ De: Eric Shu <e...@pivotal.io<mailto:e...@pivotal.io>> Enviado: miƩrcoles, 22 de abril de 2020 19:53 Para: user@geode.apache.org<mailto:user@geode.apache.org> Asunto: Re: Delta propagation and transactions Hi, Delta propagation is supported inside a transaction. I think the document is to indicate that the distribution of the transactional commit to the replicas can not use delta propagation. Regards, Eric On Wed, Apr 22, 2020 at 9:55 AM Alberto Gomez <alberto.go...@est.tech> wrote: Hi, According to the Geode documentation "Geode does not propagate deltas for Transactional commit". Is there a reason why delta propagation cannot be supported for data updated inside transactions? Thanks in advance, -Alberto G.