Distributed systems and primary keys

On 07/12/2013 07:23 AM, Melvin Call wrote:

Hello list,

Can anyone point me to some reading material on how auto-generated
sequence primary keys are handled on distributed systems? I think the
advice used to be to use GUIDs, but I thought I read somewhere that
PostgreSQL now assigns a pool of numbers to each node when a sequence is
implemented. I have searched the PostgreSQL 9.1.5 Documentation, but
apparently my search terms are not quite what it takes, or dreamt that up.

PostgreSQL itself does not support a distributed architecture. You may
be thinking of Postgres-XC?

Sequences are local to each instances and it is not a pool, it is a
64bit allocation for each sequence within the local node, generally
constrained only when called from the serial (big serial being 64 bits)
type to 32 bits.

Sincerely,

Joshua D. Drake

Thanks,
Melvin

--
Command Prompt, Inc. - http://www.commandprompt.com/ 509-416-6579
PostgreSQL Support, Training, Professional Services and Development
High Availability, Oracle Conversion, Postgres-XC, @cmdpromptinc
For my dreams of your image that blossoms
a rose in the deeps of my heart. - W.B. Yeats

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

In Postgres-XC, GTM assigns sequence value to all the transactions in its
cluster. XC is a kind of tightly-coupled distributed system. In a
loosely-coupled distributed system, where each database is autonomous, we
may need another mechanism.

I've learned that logical replication (used to be bi-directional
replication) people are doing this kind of work.

Regards;

----------
Koichi Suzuki

2013/7/12 Joshua D. Drake <jd@commandprompt.com>

On Fri, Jul 12, 2013 at 9:46 AM, Joshua D. Drake <jd@commandprompt.com>wrote:

On 07/12/2013 07:23 AM, Melvin Call wrote:

Hello list,

Can anyone point me to some reading material on how auto-generated
sequence primary keys are handled on distributed systems? I think the
advice used to be to use GUIDs, but I thought I read somewhere that
PostgreSQL now assigns a pool of numbers to each node when a sequence is
implemented. I have searched the PostgreSQL 9.1.5 Documentation, but
apparently my search terms are not quite what it takes, or dreamt that up.

PostgreSQL itself does not support a distributed architecture. You may be
thinking of Postgres-XC?

Sequences are local to each instances and it is not a pool, it is a 64bit
allocation for each sequence within the local node, generally constrained
only when called from the serial (big serial being 64 bits) type to 32 bits.

Sincerely,

Joshua D. Drake

Thank you Joshua.

Yep, if you are correct, and I believe you are, then obviously I was
thinking of something else (this happens when you work on multiple systems
I'm afraid). That would explain why I was having so much trouble finding
something that wasn't there.

This will make our planning a bit easier because I now know what the limits
currently are. I will be checking out Postgres-XC as well.

Regards,
Melvin

On Fri, Jul 12, 2013 at 10:04 AM, Koichi Suzuki <koichi.szk@gmail.com>wrote:

In Postgres-XC, GTM assigns sequence value to all the transactions in its
cluster. XC is a kind of tightly-coupled distributed system. In a
loosely-coupled distributed system, where each database is autonomous, we
may need another mechanism.

I've learned that logical replication (used to be bi-directional
replication) people are doing this kind of work.

Regards;

----------
Koichi Suzuki

Thanks Koichi. I will be looking into it shortly.

Melvin