Postgresql - unable to drop database because of some auto connections to DB

问题

Whenever I try to drop database I get:

ERROR:  database "pilot" is being accessed by other users
DETAIL:  There is 1 other session using the database.

When I use:

SELECT pg_terminate_backend(pg_stat_activity.pid)
FROM pg_stat_activity
WHERE pg_stat_activity.datname = 'TARGET_DB';

I terminated the connection from that DB, but if I try to drop database after that somehow someone automatically connects to that database and gives this error. What could be doing that? No one uses this database, except me.

回答1:

You can prevent future connections:

REVOKE CONNECT ON DATABASE thedb FROM public;

(and possibly other users/roles; see \l+ in psql)

You can then terminate all connections to this db except your own:

SELECT pid, pg_terminate_backend(pid) 
FROM pg_stat_activity 
WHERE datname = current_database() AND pid <> pg_backend_pid();

On older versions pid was called procpid so you'll have to deal with that.

Since you've revoked CONNECT rights, whatever was trying to auto-connect should no longer be able to do so.

You'll now be able to drop the DB.

This won't work if you're using superuser connections for normal operations, but if you're doing that you need to fix that problem first.

回答2:

Whenever I try to drop database I get:

ERROR:  database "pilot" is being accessed by other users
DETAIL:  There is 1 other session using the database.

First You need to revoke

REVOKE CONNECT ON DATABASE TARGET_DB FROM public;

Then use:

SELECT pg_terminate_backend(pg_stat_activity.pid)
FROM pg_stat_activity
WHERE pg_stat_activity.datname = 'TARGET_DB';

It will surely work.

回答3:

I found a solution for this problem try to run this command in terminal

ps -ef | grep postgres

kill process by this command

sudo kill -9 PID

回答4:

Simply check what is the connection, where it's coming from. You can see all this in:

select * from pg_stat_activity where datname = 'TARGET_DB';

Perhaps it is your connection?

回答5:

It means other user is accessing the database... Simply, restart the postgresql. This command will do the trick

root@kalilinux:~#sudo service postgresql restart

Then Try dropping the database:

postgres=# drop database test_database

This will do the trick. Happy coding

回答6:

If no potential impact on other services on your machine, simply service postgresql restart

回答7:

pgAdmin 4 solution using UI

First enable show activity on dashboard if you haven't:

File > Preferences > Dashboards > Display > Show Activity > true

Now disable all the processes using the db:

1. Click the DB name
2. Click Dashboard > Sessions
3. Click refresh icon
4. Click the delete (x) icon beside each process to end them

Should now be able to delete the db.

回答8:

Simple as that

sudo service postgresql restart

回答9:

Solution:
1. Shut down Pg server

2. It will disconnect all active connection
3. Restart Pg Server
4. Try your command

回答10:

In my case, I am using AWS Redshift (based on Postgres). And it appears there are no other connections to the DB, but I am getting this same error.

ERROR:  database "XYZ" is being accessed by other users

In my case, it seems the database cluster is still doing some processing on the database, and while there are no other external/user connections, the database is still internally in use. I found this by running the following:

SELECT * FROM stv_sessions;

So my hack was to write a loop in my code, looking for rows with my database name in it. (of course the loop is not infinite, and is a sleepy loop, etc)

SELECT * FROM stv_sessions where db_name = 'XYZ';

If rows found, proceed to delete each PID, one by one.

SELECT pg_terminate_backend(PUT_PID_HERE);

If no rows found, proceed to drop the database

DROP DATABASE XYZ;

Note: In my case, I am writing Java unit/system tests, where this could be considered acceptable. This is not acceptable for production code.

---

Here is the complete hack, in Java (ignore my test/utility classes).

  int i = 0;
  while (i < 10) {
    try {
      i++;
      logStandardOut("First try to delete session PIDs, before dropping the DB");
      String getSessionPIDs = String.format("SELECT stv_sessions.process, stv_sessions.* FROM stv_sessions where db_name = '%s'", dbNameToReset);
      ResultSet resultSet = databaseConnection.execQuery(getSessionPIDs);
      while (resultSet.next()) {
        int sessionPID = resultSet.getInt(1);
        logStandardOut("killPID: %s", sessionPID);
        String killSessionPID = String.format("select pg_terminate_backend(%s)", sessionPID);
        try {
          databaseConnection.execQuery(killSessionPID);
        } catch (DatabaseException dbEx) {
          //This is most commonly when a session PID is transient, where it ended between my query and kill lines
          logStandardOut("Ignore it, you did your best: %s, %s", dbEx.getMessage(), dbEx.getCause());
        }
      }

      //Drop the DB now
      String dropDbSQL = String.format("DROP DATABASE %s", dbNameToReset);
      logStandardOut(dropDbSQL);
      databaseConnection.execStatement(dropDbSQL);
      break;
    } catch (MissingDatabaseException ex) {
      //ignore, if the DB was not there (to be dropped)
      logStandardOut(ex.getMessage());
      break;
    } catch (Exception ex) {
      logStandardOut("Something went wrong, sleeping for a bit: %s, %s", ex.getMessage(), ex.getCause());
      sleepMilliSec(1000);
    }
  }

回答11:

While I found the two top-upvoted answers useful on other occasions, today, the simplest way to resolve the issue was to realize that PyCharm might be keeping a session open, and if I clicked Stop in PyCharm, that might help. With pgAdmin4 open in the browser, I did so, and almost immediately saw the Database sessions stats drop to 0, at which point I was able to drop the database.

回答12:

In my opinion there are some idle queries running in the backgroud.

1. Try showing running queries first

SELECT pid, age(clock_timestamp(), query_start), usename, query 
FROM pg_stat_activity 
WHERE query != '<IDLE>' AND query NOT ILIKE '%pg_stat_activity%' 
ORDER BY query_start desc;

2. kill idle query ( Check if they are referencing the database in question or you can kill all of them or kill a specific using the pid from the select results )

SELECT pg_terminate_backend(procpid);

Note: Killing a select query doesnt make any bad impact

回答13:

In terminal try this command:

ps -ef | grep postgres

you will see like:

501 1445 3645 0 12:05AM 0:00.03 postgres: sasha dbname [local] idle

The third number (3645) is PID.

You can delete this

sudo kill -9 3645

And after that start your PostgreSQL connection.

Start manually:

pg_ctl -D /usr/local/var/postgres start

来源：https://stackoverflow.com/questions/17449420/postgresql-unable-to-drop-database-because-of-some-auto-connections-to-db