Re: What is the best setup for distributed and fault-tolerant PG database?

Dear PG experts,

I am new to postgres, and I am also not a DBA. I am a solo developer who is
trying to evaluate what database to use for my hybrid multi-tenancy
sub-apps i.e. users of the application will be authorised to use part or
whole of the application based on their authorisation levels. This
delineation of user access has to also be supported by the database, if
possible. Also, for audit purposes the data is append only. And the design
is based on just two tables(vertices and edges) to emulate a
document-oriented(jsonb) graph structure.

Postgres is the database I am leaning towards for this project. But as I am
not a DBA and also a solo developer, I am trying to understand how I can
spend less time managing the DB and more time developing the application. I
would like to have a distributed and fault-tolerant DB setup with multiple
read and write nodes with little to no configuration on my part, if
possible. I am looking for a self-hosted open source solution.

Is this possible with PG? What is the best way to achieve this for a
non-DBA solo developer like me?

Thanks and kind regards

On 1/4/23 06:26, Age Apache wrote:
> Dear PG experts,
>
> I am new to postgres, and I am also not a DBA. I am a solo developer
> who is trying to evaluate what database to use for my hybrid
> multi-tenancy sub-apps i.e. users of the application will be
> authorised to use part or whole of the application based on their
> authorisation levels. This delineation of user access has to also be
> supported by the database, if possible. Also, for audit purposes the
> data is append only. And the design is based on just two
> tables(vertices and edges) to emulate a document-oriented(jsonb) graph
> structure.
>
> Postgres is the database I am leaning towards for this project. But as
> I am not a DBA and also a solo developer, I am trying to understand
> how I can spend less time managing the DB and more time developing the
> application. I would like to have a distributed and fault-tolerant DB
> setup with multiple read and write nodes with little to no
> configuration on my part, if possible. I am looking for a self-hosted
> open source solution.
>
> Is this possible with PG? What is the best way to achieve this for a
> non-DBA solo developer like me?
>
> Thanks and kind regards
None of the experts chimed in so I ante up my $0.02.

It won't be possible unless you become a serious DBA _and_ solo (full
stack) developer.  Or you pay for db support.

What Rob said... plus, I would urge you to give some more thought to "for
audit purposes the data is append only". If your application is ever
successful, non-insignificant storage costs are something you'll need to
deal with sooner or later.

Anyway, what you ask is certainly achievable, but not without sustained
effort. IMO your options are: spend the time to learn on your own with much
reading + trial and error; pay somebody to set it up for you; or, cross the
high-availability bridge after you've got something of substance developed,
app-wise, on a single local DB.

Best of luck!

On Wed, Jan 4, 2023, 6:54 p.m. Rob Sargent <robjsargent(at)gmail(dot)com> wrote:

> On 1/4/23 06:26, Age Apache wrote:
>
> Dear PG experts,
>
> I am new to postgres, and I am also not a DBA. I am a solo developer who
> is trying to evaluate what database to use for my hybrid multi-tenancy
> sub-apps i.e. users of the application will be authorised to use part or
> whole of the application based on their authorisation levels. This
> delineation of user access has to also be supported by the database, if
> possible. Also, for audit purposes the data is append only. And the design
> is based on just two tables(vertices and edges) to emulate a
> document-oriented(jsonb) graph structure.
>
> Postgres is the database I am leaning towards for this project. But as I
> am not a DBA and also a solo developer, I am trying to understand how I can
> spend less time managing the DB and more time developing the application. I
> would like to have a distributed and fault-tolerant DB setup with multiple
> read and write nodes with little to no configuration on my part, if
> possible. I am looking for a self-hosted open source solution.
>
> Is this possible with PG? What is the best way to achieve this for a
> non-DBA solo developer like me?
>
> Thanks and kind regards
>
> None of the experts chimed in so I ante up my $0.02.
>
> It won't be possible unless you become a serious DBA _and_ solo (full
> stack) developer. Or you pay for db support.
>
>
>

>>> age(dot)apache(at)gmail(dot)com wrote:
>>>
>>> I am new to postgres, and I am also not a DBA. I am a solo developer who is trying to evaluate what database to use for my hybrid multi-tenancy sub-apps i.e. users of the application will be authorised to use part or whole of the application based on their authorisation levels. This delineation of user access has to also be supported by the database, if possible. Also, for audit purposes the data is append only. And the design is based on just two tables(vertices and edges) to emulate a document-oriented(jsonb) graph structure.
>>>
>>> Postgres is the database I am leaning towards for this project. But as I am not a DBA and also a solo developer, I am trying to understand how I can spend less time managing the DB and more time developing the application. I would like to have a distributed and fault-tolerant DB setup with multiple read and write nodes with little to no configuration on my part, if possible. I am looking for a self-hosted open source solution.
>>>
>>> Is this possible with PG? What is the best way to achieve this for a non-DBA solo developer like me?
>>
>> robjsargent(at)gmail(dot)com wrote:
>>
>> None of the experts chimed in so I ante up my $0.02. It won't be possible unless you become a serious DBA _and_ solo (full stack) developer. Or you pay for db support.
>
> saul(dot)perdomo(at)gmail(dot)com wrote:
>
> What Rob said... plus, I would urge you to give some more thought to "for audit purposes the data is append only". If your application is ever successful, non-insignificant storage costs are something you'll need to deal with sooner or later.
>
> Anyway, what you ask is certainly achievable, but not without sustained effort. IMO your options are: spend the time to learn on your own with much reading + trial and error; pay somebody to set it up for you; or, cross the high-availability bridge after you've got something of substance developed, app-wise, on a single local DB.

There’s always Internet search. Mention as many key phrases as you see fit. For example:

> Fully managed cloud service for highly available, fault tolerant, Postgres-compatible distributed SQL database

Look at my email address. That outfit is among the hits. So full disclosure is done. But, as they say, “other services are available”. So I don’t think that my suggested search string is inappropriate for a list like this.