Azure PostgreSQL 服务器服务排序规则创建错误

Question

I'm trying to import my current existing database to postgre instance that is running on Azure PostgreSQL Server

I already configured my azure postgresql server parameters to use encoding as UTF8 (I'm not sure that it's applying without restart but even if i don't have option to restart it)

I'm trying to do this action with this command:

sudo -u postgres pg_dump --encoding="UTF-8" --no-owner DBNAME | psql --host=xxx.postgres.database.azure.com --port=5432 --username=xxx@yyy --dbname=DBNAME

However it's getting an error something like this:

ERROR: could not create locale "en_US.utf8": No error

I dive in the process and try to run it with UTF8 (Instead of UTF-8) and other options but it's got same error everytime.

I created a dump file to check the contents and the line that is generating this error is:

CREATE COLLATION "en_US.utf8" (lc_collate = 'en_US.utf8', lc_ctype = 'en_US.utf8');

This is causing that error. I tried to execute it manually on Azure postgre instance and i got same error too.

By the way I already created that DATABASE from console by using this SQL:

create database "DBNAME" with owner=xxx encoding='UTF-8' template template0;

Also i tried other options to create and import but non of them work.

Can someone help me?

Answer 1

I resolved my issue with this command:

sudo -u postgres pg_dump -Fp —no-owner DBNAME |sed “/COLLATE/s/en_US.utf8/English_United States.1252/ig”|sed "/CREATE COLLATION/s/en_US.utf8/English_United States.1252/ig"|psql --host=aaa.postgres.database.azure.com --port=5432 --username=xxx@yyy --dbname=DBNAME

The issue is coming from the server OS differences. My current PostgreSQL server is using Ubuntu but Azure PostgreSQL Server is using windows so the collate names are little bit different.

Answer 2

CREATE COLLATION creates a operating system independent name that can be used to refer to OS locales (in queries etc).

Here lc_collate = 'en_US.utf8' and lc_ctype = 'en_US.utf8' refer to Linux operating system locales, which are named differently on Windows, which Azure PostgreSQL uses (and they're different on MacOS, too).

On Windows, this should work

CREATE COLLATION "en_US.utf8" (lc_collate = 'English_United States', lc_ctype = 'English_United States');

And on MacOS,

CREATE COLLATION "en_US.utf8" (lc_collate = 'en_US', lc_ctype = 'en_US');

But most of the time, you don't want to create your own collations with CREATE COLLATION , but can just use the pre-included collations in PostgreSQL. And normally a dump created with pg_dump does not include any CREATE COLLATION statements, since you haven't created any yourself.

Unless you're doing something special with collations, is it possible to remove those schema-specific collations, so they won't get included in the dump?

Answer 3

We've previously run into this same issue running Atlassian's Confluence. The application would not start due to an incorrect collation type. We were able to resolve by assigning English_United States.1252 directly.

CREATE DATABASE db 
  WITH ENCODING 'UTF8' 
  LC_COLLATE    'English_United States.1252' 
  LC_CTYPE      'English_United States.1252' 
  TEMPLATE      template0 
  OWNER         conf;