Обсуждение: how to create multiple databases running in different dirs

Hi, I'm new to postgresql and sql in general. I desired to write a program in C that used an sql data base for IPC
andbecause multiple copies of my program might run on the same machine I wanted a way to ensure that only one copy of
eachmultithreaded program got one database but I'm uncertain how to go about this. As the program might be run
severaldifferent times and each time it should only get the data base from it's predisesor I can't use the pid
asa unique data base name. I thought that I might start multiple pgsql instances but somehow that seems wrong. Any
ideas?

 

Thanks, David

 

Hi, I'm new to postgresql and sql in general. I desired to write a program in C that used an sql data base for IPC and because multiple copies of my program might run on the same machine I wanted a way to ensure that only one copy of each multithreaded program got one database but I'm uncertain how to go about this. As the program might be run several different times and each time it should only get the data base from it's predisesor I can't use the pid as a unique data base name. I thought that I might start multiple pgsql instances but somehow that seems wrong. Any ideas?

 

Thanks, David

 

On 6/30/2014 4:58 PM, frank ernest wrote:
> Hi, I'm new to postgresql and sql in general. I desired to write a
> program in C that used an sql data base for IPC and because multiple
> copies of my program might run on the same machine I wanted a way to
> ensure that only one copy of each multithreaded program got one
> database but I'm uncertain how to go about this. As the program might
> be run several different times and each time it should only get the
> data base from it's predisesor I can't use the pid as a unique data
> base name. I thought that I might start multiple pgsql instances but
> somehow that seems wrong. Any ideas?

how would an instance of your program know what to connect to, or which
previous instance its 'predecessor' was ?

normally, you have ONE database for a given set of applications, and all
the applications share the same database tables and such.


--
john r pierce                                      37N 122W
somewhere on the middle of the left coast

On 7/1/2014 7:19 AM, Rémi Cura wrote:
>
> If it is so your desire, you could also have multiple server on the
> same machine (althought on different port).
> This way each server would have its own repository.

which still doesn't answer the question, how would an instance of his
program know which database (port) to connect to?



--
john r pierce                                      37N 122W
somewhere on the middle of the left coast

> how would an instance of your program know what to connect to, or which
> previous instance its 'predecessor' was ?

> normally, you have ONE database for a given set of applications, and all
> the applications share the same database tables and such.


That's the problem, is there some way to tell pgsql "Go to dir X, open your data base Y and prepare for
connectionsat adress Z and port P"? And could pgsql accept multiple connections on the same address and port? I
wasthinking of using my pID but that would change and if I used a user created string then if I started only on a
singleinstace of pgsql and pointed it to it's databases the user might get the strings duplicated and that would be
nogood. I also thought of naming each database with a name generated by using the uuid library but I'm not sure
whichwould be best. 

 

If I choose to have multiple servers running on the same machine how would my program start them?

 

> Is each runtime instance of your application to have a new, empty
> database? No data from any other run of the application?

 

No, the dynamically generated content is to be dropped (drop table dynamic_content;) but the rest is to be preserved.
Theidea is to create a database of a file system but the files contain no data, I only want their metadata and I will
adda few additional metadta values to each file. 

 

Thanks, David

 

> > how would an instance of your program know what to connect to, or which
> > previous instance its 'predecessor' was ?
> > normally, you have ONE database for a given set of applications, and all
> > the applications share the same database tables and such.
>
> That's the problem, is there some way to tell pgsql "Go to dir X, open
> your data base Y and prepare for connections at adress Z and port P"?
> And could pgsql accept multiple connections on the same address and
> port? I was thinking of using my pID but that would change and if I
> used a user created string then if I started only on a single instace
> of pgsql and pointed it to it's databases the user might get the
> strings duplicated and that would be no good. I also thought of naming
> each database with a name generated by using the uuid library but I'm
> not sure which would be best.

why directory, address and port?   why not just 'open database X' ?
this still doesn't answer my question, HOW do you determine which
database a given instance of your program is supposed to connect to?

yes, postgres supports many different connections to the same address
and port, either to the same database or different databases.
relational databases are totally built around concurrent transactional
operations.

there's no such thing as 'directory', all database tables are stored in
the database server's private storage, the client app has no need to
know where this is.  There /is/ a concept of 'tablespaces', these are
used when you want to put different tables on different file systems,
typically used for very large scale databases, and/or when you have
different performance tiers of storage.


> No, the dynamically generated content is to be dropped (drop table
> dynamic_content;) but the rest is to be preserved. The idea is to
> create a database of a file system but the files contain no data, I
> only want their metadata and I will add a few additional metadta
> values to each file.

using TRUNCATE SOME_TABLE  would likely make more sense, that deletes
all the data in table SOME_TABLE, without actually deleting the table
definition, so you can then proceed to insert new data into it.

I really don't understand what you mean by 'a database of a file
system'.   I'm beginning to suspect you don't understand how SQL
databases work.




--
john r pierce                                      37N 122W
somewhere on the middle of the left coast