[YSQL] LC_CTYPE = 'C' support

[philtreF]

Jira Link: DB-3424

Description

Hi,

I wanted to deploy a Matrix Synapse server on Yugabyte.
As explained on this page this software needs LC_CTYPE = 'C' like on this example :

postgresserver=#  \l
                                 List of databases
   Name    |  Owner   | Encoding |  Collate   |   Ctype    |   Access privileges   
-----------+----------+----------+------------+------------+-----------------------
 synapse   | synapse  | UTF8     | C          | C          | 

However when I try to create this kind of database I got this error :

yugabyte=# CREATE DATABASE synapse2 WITH OWNER = synapse TEMPLATE='template0' ENCODING='UTF-8' LC_CTYPE='C';
ERROR:  Value other than 'en_US.UTF-8' for lc_ctype option is not yet supported
LINE 1: ...R = synapse TEMPLATE='template0' ENCODING='UTF-8' LC_CTYPE='...

Devs are specific on the fact that this is mandatory to avoid database corruption source
Quote :

allow_unsafe_locale is an option specific to Postgres. Under the default behavior, Synapse will refuse to start if the postgres db is set to a non-C locale. You can override this behavior (which is not recommended) by setting allow_unsafe_locale to true. Note that doing so may corrupt your database.

I am not sure about the difference on the backend for you. Is that a planned feature to implement ? Does this make sense ?

Fred

[ddorian]

allow_unsafe_locale is an option specific to Postgres. Under the default behavior, Synapse will refuse to start if the postgres db is set to a non-C locale. You can override this behavior (which is not recommended) by setting allow_unsafe_locale to true. Note that doing so may corrupt your database.

Note that this should work in our case, there's no risk of db corruption.

[philtreF]

Thanks @ddorian for your answer. I understood this should not lead me to corrupt the database. Can you explain me why you think this (link, source code, etc) ?
Do you know if this LC_CTYPE = 'C' will be implemented soon in Yugabyte?

[ddorian]

I understood this should not lead me to corrupt the database. Can you explain me why you think this (link, source code, etc) ?

I think they should explain why it may lead to corruption (the exact scenario).

And we can see how/if it applies to us.

Steps to reproduce

have DB use wrong locale
update libc with locale change
notice corrupt DB

I don't think this will apply to us. cc @m-iancu

Do you know if this LC_CTYPE = 'C' will be implemented soon in Yugabyte?

It's not on the near-term roadmap https://github.com/yugabyte/yugabyte-db#current-roadmap