Re: Too many postgres.exe - Mailing list pgsql-general

From Scott Mead
Subject Re: Too many postgres.exe
Date
Msg-id d3ab2ec80912150610n27c6b19sbbfb757e20a9d1f0@mail.gmail.com
Whole thread Raw
In response to Too many postgres.exe  (<A.Bhattacharya@sungard.com>)
List pgsql-general
On Tue, Dec 15, 2009 at 4:51 AM, <A.Bhattacharya@sungard.com> wrote:

Hi All,

 

I have my application UI in Java and which is communicating with Postgresql database.

However whenever my application is running I could see there are too many postgres.exe are created even though the application is not doing anything in database.

Remember, every connection gets a 'postgres.exe' (dedicated backend per connection).  And it'll stick around until you close the connection.  If you create a connection and don't close it until the DB goes away, then you'll still have the backend process there even though you're not doing anything.


Can you describe what you consider 'Too Many?' The process-per-connection architecture is very common (albeit not on windows), so I'm curious what your threshold is. 

 

 

In general the observation is  that there are too many progress.exe processes get created every time I run my application and it postgres.exe eats up the maximum memory.


You'll need to use sysinternals tools here, task manager mis-reports how much memory is actually being allocated by postgres.  This is because each postgres.exe is attaching to a shared memory area.  This means that you could be using 2 GB of memory (total) for your postgres database, when it looks like you're using 20 GB of memory. (10 backends all showing 2 GB).  The reality is... you're not using (mem * num of postgres.exe), you should use sysinternals tools to discern how much memory is shared vs. per-backend.

--Scott

 

Any help/suggestion please!!

 


pgsql-general by date:

Previous
From: Josh Kupershmidt
Date:
Subject: Re: 8.3 PL/pgSQL comparing arbitrary records
Next
From: "Peter"
Date:
Subject: PlPerl scope issue