Tidb: Postgresql Support

[posix4e]

This is not a bug it's a feature! I am requesting postgreql suport via FDW. BTW I generated the capability to do PG FDW via Rust! https://github.com/posix4e/rpgffi. We probably don't wanna do this as a part of tidb though.
Some projects which use this code

• https://github.com/instructure/jsoncdc
• https://github.com/durch/google-bigtable-postgres-fdw

[shenli]

@posix4e Thanks for your suggestion! PG protocol is on our roadmap. But it is a long-term goal.
Your repo seems interesting. I will take a look.

[posix4e]

I’d suggest an intermediate goal of a foriegn data wrapper. As my bigtable example demonstrated you can implement a simple version easily.

[ilovejs]

Could possibly need tidb as cache layer first.

Materialized view development in postgres is hard due to aggregation, sorting (million rows).

Child view depends on parent view so any change would require parent to be drop and recreated...

Although, people can use table to replace view, sometime AWS Aurora under large stress, could fail over. Then need a manual reboot (might be smarter way, but it was read as aws core internal error).

Its read replica during master DDL, will lock up...so session number went up to max..

Would you guys think of these user cases ?

Thanks 💯

[posix4e]

I think not starting with a simple Fdw is hard.
On Mon, Jul 2, 2018 at 2:53 AM M Zhuang [e-mail] wrote:

Could possibly need tidb as cache layer first. Materialized view
development in postgres is hard due to aggregation, sorting. Child view
depends on parent view so any change would require to be changed..
Althrough, people can use table to replace view. But sometime AWS Aurora
under large stress, database won't be back online anymore.

Its read replica during master DDL, will lock up...so session number went
up to max..

Would you guys think of these user case ?

Thanks 💯

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/pingcap/tidb/issues/6459#issuecomment-401741913, or mute
the thread
https://github.com/notifications/unsubscribe-auth/AAxN27TTyDAQ0YS848BBOojQxQOYXXDcks5uCe2rgaJpZM4Txcqj
.

[codelipenghui]

@shenli

@posix4e Thanks for your suggestion! PG protocol is on our roadmap. But it is a long-term goal.
Your repo seems interesting. I will take a look.

Road map link looks like it’s out of date, can you give me a new link? I'm also care about postgresql support.

Thanks!

[eurekaka]

@codelipenghui Here is the link of the roadmap, PostgreSQL protocol compatibility is not in our short term schedule.

[codelipenghui]

@eurekaka thanks.

[ForsakenHarmony]

What made you chose mysql over postgres support?

[eurekaka]

@shenli please help explain a little bit about the protocol choice, thanks.

[morgo]

@ForsakenHarmony in the MySQL-world it is very common to shard databases. We targeted MySQL support first because we have something to offer their users.

[kaushalyap]

Any progress with this?

[scsldb]

Short to medium term focus on MySQL compatibility, no plans yet.