i would like to be able to do something like
CREATE TABLESPACE bob location 'C:\a\b\c\d\e\f\bob'
without needing to create all the directory tree beforehand.
this is because i have java code that creates tablespaces on the fly and i would like to be able to run it on a separate machine (so it couldnt mkdir() or anything).
is there any sort of postgres configuration that would allow me to make postgres create the appropriate directory tree by itself?
You could try do mkdir directly in postgres stored procedure using PL/sh or any of your favorite PL/* languages that are available for PostgreSQL
Related
I'm planning to move some tables to different tablespaces (folders) on my PROD Linux box.
Overnight DB backups are done using pg_dumpall
I have also DEV environment working under Windows OS Im usually restoring sql dump (made on Linux).
Im worrying now how to restore such sql dumps, having pointers to Linux partition, in Linux notation.
I read on various webpages that same folder structure has to be created in order to restore non-standard tablespaces. But folder paths in Windows and Linux looks totally different (c:\... vs /opt/...)
Is there any command line switch allowing remap tablespace to other (Windows-like location) during restore? If not how you guys manage that scenario ?
I guess I shoud be able to archieve that by editing this SQL dump file - but it's huge, few hundred gigs file, also it is a bit problematic to automate
You can retrieve the actual tablespace definitions with a separate pg_dumpall command. You still need to do some editing, but the output is not that large. (similar for users)
pg_dumpall --tablespaces-only mydatabasename >stuff.out
There is no option to remap tablespace names during import, so you will need to create them in your Windows installation with the same name - the actual location physical location ("folder structure") is irrelevant as the SQL dump only references them by name.
If the script contains the create tablespace command you need to change that command to use a directory/path name that exists on your system before you can run the SQL script. But you only need to change that, all other places will refer to the tablespace name, not the folder path.
Typically pg_dump is easier than pg_dumpall for moving databases around (e.g. because of tablespaces).
Whenever I create a new database from pgAdmin or using the command line (using CREATE DATABASE database_name), it's not empty.
It contains some tables that are part of a previous project I worked on.
I'm not yet very familiar with Psql so I don't know what I'm doing wrong.
You probably have created objects in the database template1.
Quote from the manual:
By default, the new database will be created by cloning the standard system database template1. A different template can be specified by writing TEMPLATE name. In particular, by writing TEMPLATE template0, you can create a virgin database containing only the standard objects predefined by your version of PostgreSQL. This is useful if you wish to avoid copying any installation-local objects that might have been added to template1.
So, anything that is in the template1 database will be copied over to the new database when you run create database.
Connect to the template1 database and drop all objects you don't want.
Is it possible to create "DIRECTORY" object in Postgres?
If not can some help me with a solution how implement it on PostgreSQL.
Not the best option, but you could use:
COPY (select 1) TO PROGRAM 'mkdir --mode=777 -p /path/to/your/directory/'
Note that only the last part of directory get the permissions set in mode.
There is no equivalent concept to an "Oracle directory" in Postgres.
The alternatives depend on why the "Oracle directory" is needed.
If the directory is needed to read and write files on the database server, then this can be done through Generic File Access Functions. Access to those functions is restricted to superusers (details in the linked section of the manual). If regular users should be able to use them, the best thing would be to create wrapper functions and then grant execute on those functions to the users in question.
For security reasons, only directories inside the database cluster can be accessed.
But it's possible to create symlinks inside the data directory that point to directories outside the data directory. Access privileges on those directories need to be properly setup for the postgres operating system user (the one under which the postgres process is started)
If the directory is needed to access e.g. CSV files through Oracle's external tables, then there is no need for a "directory". The file FDW foreign data wrapper, can access files outside the data directory (provided access privileges have been setup correctly on the file system level).
The question doesn't even make sense really. PostgreSQL is a database management system. It doesn't have files and directories.
The closest parallel I can think of is schemas - see CREATE SCHEMA.
Now, if you want to use COPY to write output to the server's disk and want to create a directory to put that output in... then no, there's nothing like that. But you can use PL/Perlu or PL/Pythonu to do it easily enough.
Whenever I create a new database from pgAdmin or using the command line (using CREATE DATABASE database_name), it's not empty.
It contains some tables that are part of a previous project I worked on.
I'm not yet very familiar with Psql so I don't know what I'm doing wrong.
You probably have created objects in the database template1.
Quote from the manual:
By default, the new database will be created by cloning the standard system database template1. A different template can be specified by writing TEMPLATE name. In particular, by writing TEMPLATE template0, you can create a virgin database containing only the standard objects predefined by your version of PostgreSQL. This is useful if you wish to avoid copying any installation-local objects that might have been added to template1.
So, anything that is in the template1 database will be copied over to the new database when you run create database.
Connect to the template1 database and drop all objects you don't want.
Currently, I have an application that uses Firebird in embedded mode to connect to a relatively simple database stored as a file on my hard drive. I want to switch to using PostgreSQL to do the same thing (Yes, I know it's overkill). I know that PostgreSQL cannot operate in embedded mode and that is fine - I can leave the server process running and that's OK with me.
I'm trying to figure out a connection string that will achieve this, but have been unsuccessful. I've tried variations on the following:
jdbc:postgresql:C:\myDB.fdb
jdbc:postgresql://C:\myDB.fdb
jdbc:postgresql://localhost:[port]/C:\myDB.fdb
but nothing seems to work. PostgreSQL's directions don't include an example for this case. Is this even possible?
You can trick it. If you are running PostGRESQL on a UNIXlike system, then you should be able to create a RAMDISK and use that for the database storage. Here's a pretty good step by step guide for RAMdisks on Linux.
In general though, I would suggest using SQLITE for an SQL db in RAM type of application.
Postgres databases are not a single file. There will be one file for each table and each index in the data directory, inside a directory for the database. All files will be named with the object ID (OID) of db / table / index.
The JDBC urls point to the database name, not any specific file:
jdbc:postgresql:foodb (localhost is implied)
If by "disk that behaves like memory", you mean that the db only exists for the lifetime of your program, there's no reason why you can't create a db at program start and drop it at program exit. Note that this is just DDL to create the DB, not creating the data dir via the init-db program. You could connect to the default 'postgres' db, create your db then connect to it.
Firebird 2.1 onwards supports global temporary tables, which only exist for the duration of the database connection.
Syntax goes something like CREATE GLOBAL TEMPORARY TABLE ... ON COMMIT PRESERVE ROWS