Re: Re: Re: - Mailing list pgsql-admin
From | Simon Riggs |
---|---|
Subject | Re: Re: Re: |
Date | |
Msg-id | CANbhV-HsLrE+hpgFJXHG08k5h5HM+7g+4wMtkk_wR9GnbzMNKg@mail.gmail.com Whole thread Raw |
In response to | Re: Re: Re: (MichaelDBA <MichaelDBA@sqlexec.com>) |
Responses |
Re:
|
List | pgsql-admin |
On Wed, 24 Nov 2021 at 18:18, MichaelDBA <MichaelDBA@sqlexec.com> wrote: > > Thanks, Simon, for your continued feedback. > > Simon Riggs wrote on 11/24/2021 12:49 PM: > > Anyone interested to know more can start here: > https://www.enterprisedb.com/products/bidirectional-replication-bdr-postgresql-database You clearly have an interest in BDR, and some knowledge of BDR1, so I thank you for that. The limitations of BDR1, written in 2014, are not limitations of the architecture in general, merely things it didn't do. That has led to some misunderstandings about what is possible and regrettably some incorrect points have been made that I've attempted to rebut. Some of those points relate to how Postgres-XL worked, but are not relevant to BDR. 7 years later, BDR3 has a significant number of features not present in BDR1. That may not be as well known, since as you say, BDR3 is not fully open source. I regret that I was not able to fund further development of BDR without charging users. Having said that, probably more than 50% of BDR features are actually open source and part of PostgreSQL - the contribution of new features has continued with each new release. Specific to this conversation, BDR3 supports multiple transaction modes - with various kinds of consistency. I would point out that those modes are slower - which is why multiple options are present. Some modes allow conflicts, some do not. This has nothing to do with "Loosely-coupled", which does not present a limitation. Should a conflict occur, a conflict doesn't *inevitably* cause a rollback/suspension requiring manual intervention. That doesn't seem to be a fair characterisation of the current behavior, which again differs significantly from BDR1. Just like any form of replication, various actions can cause breakage or difficulties. Thank you for pointing out deficiencies in the docs. I wrote a large part of it myself over many years and it seems we haven't yet captured all of the possible options there. I'll work on improving the information available to help those interested in BDR and/or related tech. -- Simon Riggs http://www.EnterpriseDB.com/
pgsql-admin by date: