Move PostgreSQL Data Directory on Ubuntu

PostgreSQL stores each cluster’s databases and system files inside its data directory. When that directory lives on a disk that also hosts the operating system, the cluster can run out of space or start competing for I/O with other processes.

This guide shows how to move the PostgreSQL data directory to a new mount point by copying the files with rsync, updating postgresql.conf, and adjusting the Ubuntu AppArmor profile so PostgreSQL can access the new path. You will also use systemctl to stop and start PostgreSQL and psql to verify the migration.

By the end, PostgreSQL should start cleanly from the new data directory, and you will have a .bak rollback path if you need to revert.

• The default PostgreSQL data directory is /var/lib/postgresql/<version>/main, and you can confirm it with SHOW data_directory; in psql.
• rsync is preferred over cp because it preserves permissions, timestamps, symlinks, and special file metadata, and it can resume by copying only differences.
• AppArmor restricts PostgreSQL to specific filesystem paths, so a non-default data directory path often needs a profile override on Ubuntu.
• After restart, you can verify the migration succeeded by running SHOW data_directory; in psql.
• Keep the original directory as /var/lib/postgresql/<version>/main.bak to rollback safely by renaming it back and reverting data_directory.

Before you move the data directory, understand what each component inside it does. An incomplete or corrupted copy of any of these files will prevent PostgreSQL from starting or cause silent data loss.

Configuration files are also part of this picture. Even though postgresql.conf is commonly a symlink into /etc/postgresql/<version>/main/postgresql.conf, PostgreSQL reads the active data_directory value from that file.

The default layout looks like this (replace 14 with your installed PostgreSQL major version):

/var/lib/postgresql/
└── 14/
    └── main/
        ├── base/
        ├── global/
        ├── pg_wal/
        ├── pg_hba.conf
        ├── postgresql.conf -> /etc/postgresql/14/main/postgresql.conf
        └── PG_VERSION

Each entry in that tree has a specific role:

• base/ holds the actual on-disk database files. Each subdirectory inside base/ corresponds to a database in the cluster, identified by its object ID (OID). This is the largest part of the data directory on most systems.
• global/ holds cluster-wide system catalog tables such as pg_database and pg_authid. These tables track which databases exist and which roles can access them.
• pg_wal/ holds write-ahead log (WAL) segments. PostgreSQL writes changes here before applying them to the main data files. If pg_wal/ is missing or incomplete after a migration, PostgreSQL cannot guarantee data consistency and will refuse to start.
• pg_hba.conf controls client authentication rules. It defines which hosts can connect, which users they can connect as, and which authentication method applies.
• PG_VERSION is a plain text file containing the PostgreSQL major version number. PostgreSQL reads this at startup to confirm the data directory format matches the running binary.

The data_directory directive in postgresql.conf tells PostgreSQL where to find those files. After the migration, the only PostgreSQL configuration change you typically need is updating data_directory, alongside the AppArmor path permissions required on Ubuntu.

Before you move files, confirm your PostgreSQL major version and verify that AppArmor is active on the host.

To complete this guide, you will need:

• An Ubuntu server with a non-root user with sudo privileges. If you need to set this up, see Initial Server Setup with Ubuntu.

• PostgreSQL installed on your server. If you need installation steps, see How To Install and Use PostgreSQL on Ubuntu.

• Your PostgreSQL major version number. You can get it with either of these commands:

psql --version

pg_lsclusters

• AppArmor enabled. Ubuntu installs AppArmor by default on many systems, and it should be active for the profile behavior described in Step 6. Verify with:

sudo aa-status

• A new mounted location for the data directory. In examples, this uses /mnt/volume_nyc1_01, but you must substitute your own mount point path.

If you need a refresher on the file copy method used here, read How to Use rsync to Sync Local and Remote Directories. If you need to prepare the disk that backs the mount point, see How to Partition and Format Storage Devices in Linux.

Confirm where PostgreSQL is currently reading data from, then check that the new mount point has enough free space for the full cluster directory.

Run psql as the postgres user and print the current data_directory:

sudo -u postgres psql

SHOW data_directory;

       data_directory
-----------------------------
 /var/lib/postgresql/14/main
(1 row)

Exit psql with:

\q

Now confirm the mount point is available and has enough space:

df -h /mnt/volume_nyc1_01

Filesystem      Size  Used Avail Use% Mounted on
/dev/sdx1       100G   10G   90G  10% /mnt/volume_nyc1_01

Also check how large your current data directory is so you can compare it to the destination free space: <$>[warning] The destination must have at least as much free space as the current data directory size. <$>

Check the current size of your data directory:

sudo du -sh /var/lib/postgresql/

12G    /var/lib/postgresql/

If the mount point is new or freshly attached, confirm it is accessible before running the copy:

ls -la /mnt/volume_nyc1_01

total 8
drwxr-xr-x  2 root root 4096 Jan 10 09:00 .
drwxr-xr-x 12 root root 4096 Jan 10 08:55 ..

The rsync command in Step 3 preserves permissions and ownership from the source directory, but the destination filesystem must support standard POSIX file attributes. FAT32 and exFAT volumes do not; use ext4 or xfs for a PostgreSQL data directory.

Stop PostgreSQL before copying files so the data directory does not change during the migration.

Run:

sudo systemctl stop postgresql

Then verify the service is stopped:

sudo systemctl status postgresql

● postgresql.service - PostgreSQL RDBMS
     Loaded: loaded
     Active: inactive (dead)

As an alternative check, use pg_lsclusters to confirm the cluster is down:

pg_lsclusters

Ver Cluster Port Status Owner    Data directory               Log file
14  main    5432 down   postgres /var/lib/postgresql/14/main  ...

Copy the stopped cluster files with rsync so permissions, timestamps, and special metadata carry over to the new mount point.

Run the copy command:

sudo rsync -av /var/lib/postgresql /mnt/volume_nyc1_01

sending incremental file list
postgresql/
postgresql/14/
postgresql/14/main/
postgresql/14/main/PG_VERSION
postgresql/14/main/base/
postgresql/14/main/global/
postgresql/14/main/pg_hba.conf
postgresql/14/main/pg_wal/
...

sent 2,873,452 bytes  received 412 bytes  819,733.47 bytes/sec
total size is 2,872,198  speedup is 1.00

The final summary line confirms rsync completed the transfer.

Optional progress for large directories:

sudo rsync -av --progress /var/lib/postgresql /mnt/volume_nyc1_01

Flag behavior:

• -a enables archive mode, which preserves permissions, timestamps, symlinks, owner, group, and recursively copies directories.
• -v prints more detail so you can follow what rsync is transferring.
• There is no trailing slash on the source path. Without a trailing slash, rsync copies the postgresql directory itself into the destination, which results in the expected directory layout under your mount point.

If you want background on what these flags mean, see How to Use rsync to Sync Local and Remote Directories.

Rename the current data directory to a .bak path so you have a fast rollback option without re-copying data.

After the copy completes, move the old directory out of the way:

sudo mv /var/lib/postgresql/14/main /var/lib/postgresql/14/main.bak

Keep this .bak directory until you confirm the new cluster starts and the application behaves as expected. This preserves a rollback path that does not require re-running rsync.

sudo mv /var/lib/postgresql/14/main.bak /var/lib/postgresql/14/main

Point PostgreSQL at the new location by updating the data_directory directive in the active postgresql.conf.

Edit the configuration file:

sudo nano /etc/postgresql/14/main/postgresql.conf

On a default PostgreSQL installation, the data_directory line is often commented out because PostgreSQL uses a compiled-in default when no explicit value is set. When you open the file, you may see one of these three states:

The line is commented out with a # prefix:

#data_directory = '/var/lib/postgresql/14/main'  # use data in another directory

The line is absent entirely, with only a comment block referencing the default.

The line is already uncommented with the default path set explicitly:

data_directory = '/var/lib/postgresql/14/main'

In the first two cases, add the data_directory line explicitly with the new path. In the third case, replace the existing value. In all cases, the result should be a single uncommented data_directory line pointing to the new mount point.

Find the existing data_directory line in postgresql.conf:

data_directory = '/var/lib/postgresql/14/main'

Replace it with the new path (substitute your major version and mount point):

data_directory = '/mnt/volume_nyc1_01/postgresql/14/main'

After saving the file, confirm the value that PostgreSQL will read:

grep -v '^#' /etc/postgresql/14/main/postgresql.conf | grep data_directory

data_directory = '/mnt/volume_nyc1_01/postgresql/14/main'

The -v '^#' flag excludes commented lines so only the active directive is returned. Without this filter, grep would also return commented-out data_directory lines and make the output ambiguous.

Update the AppArmor local override so PostgreSQL can read and write the new data directory path on Ubuntu.

On Ubuntu, AppArmor restricts which filesystem paths PostgreSQL is allowed to access. After a data directory move to a non-default path, PostgreSQL can fail to start with a permission denied error unless the AppArmor profile allows the new directory.

First check whether AppArmor is enforcing the PostgreSQL profile:

sudo aa-status | grep postgres

   /usr/lib/postgresql/14/bin/postgres

List the available AppArmor local override files to confirm the correct filename for your PostgreSQL version:

ls /etc/apparmor.d/local/

usr.lib.postgresql.14.bin.postgres

Next, edit the AppArmor local override file for the PostgreSQL binary. Use your PostgreSQL major version in the path:

sudo nano /etc/apparmor.d/local/usr.lib.postgresql.14.bin.postgres

The rules in this local file are applied to the corresponding AppArmor profile for the PostgreSQL server binary. The path in the rule should match the data_directory path you set in Step 5.

Add a rule that grants read/write and locks on the new data directory tree:

/mnt/volume_nyc1_01/postgresql/** rwk,

In AppArmor rule syntax, r grants read access, w grants write access, and k grants file locking, which PostgreSQL requires to coordinate concurrent access to data files.

This rule covers the whole subtree under the mount point, so PostgreSQL can write files in base/, global/, and pg_wal/.

Reload AppArmor so the change takes effect:

sudo systemctl reload apparmor

Alternatively, you can reload the profile by parsing it directly:

sudo apparmor_parser -r /etc/apparmor.d/usr.lib.postgresql.14.bin.postgres

The path here points to the main profile file in /etc/apparmor.d/, not the local override file in /etc/apparmor.d/local/. When apparmor_parser reloads the main profile, it automatically includes any rules in the corresponding local override file, so editing the local file and reloading the main profile is the correct sequence.

Start PostgreSQL, then confirm it is online and that psql reports the new data_directory.

Start the service:

sudo systemctl start postgresql

Check service status:

sudo systemctl status postgresql

● postgresql.service - PostgreSQL RDBMS
     Active: active (exited) since ...

Confirm the cluster is online and points to the new directory:

pg_lsclusters

Ver Cluster Port Status Owner    Data directory                             Log file
14  main    5432 online postgres /mnt/volume_nyc1_01/postgresql/14/main    ...

Verify inside psql:

sudo -u postgres psql

SHOW data_directory;

             data_directory
----------------------------------------
 /mnt/volume_nyc1_01/postgresql/14/main
(1 row)

Run additional checks to catch obvious data access issues:

-- List all databases to find your database name
\l

                                  List of databases
   Name    |  Owner   | Encoding |   Collate   |    Ctype    |   Access privileges
-----------+----------+----------+-------------+-------------+-----------------------
 myapp_db  | postgres | UTF8     | en_US.UTF-8 | en_US.UTF-8 |
 postgres  | postgres | UTF8     | en_US.UTF-8 | en_US.UTF-8 |
(2 rows)

-- Connect to your database and list its tables
\c myapp_db
\dt

          List of relations
 Schema |    Name     | Type  |  Owner
--------+-------------+-------+----------
 public | users       | table | postgres
 public | orders      | table | postgres
(2 rows)

-- Spot-check a critical table row count to confirm data integrity
SELECT COUNT(*) FROM users;

 count
-------
  1042
(1 row)

Remove the backup only after the new data directory has been running stably for at least one application cycle.

Before removal, confirm that pg_lsclusters shows the correct Data directory path, the application has completed at least one full read/write cycle without errors (for example, one business day of normal traffic or one successful deployment and smoke test), and no application errors have appeared in logs since the migration.

Then remove the backup directory:

sudo rm -Rf /var/lib/postgresql/14/main.bak

When migration steps do not work, narrow down the cause by checking service status, AppArmor enforcement, and the exact configuration file PostgreSQL is reading.

PostgreSQL Fails to Start After Migration

Symptom: systemctl start postgresql returns a failure, and journalctl shows permission denied messages for the new data directory path.

Cause 1: The AppArmor profile was not updated. Before pointing to Step 6, confirm AppArmor is the actual cause by checking the PostgreSQL journal for the specific error:

sudo journalctl -u postgresql -n 50

Look for lines containing Permission denied referencing the new data directory path. If those lines are present, complete Step 6, reload AppArmor, and try starting PostgreSQL again. If the journal shows a different error, continue to Cause 2 and Cause 3 before revisiting the AppArmor profile.

Cause 2: Incorrect filesystem ownership or permissions. Fix:

sudo chown -R postgres:postgres /mnt/volume_nyc1_01/postgresql
sudo chmod 700 /mnt/volume_nyc1_01/postgresql/14/main

Cause 3: The data_directory path contains a typo or trailing whitespace. Fix: re-check the configured value in the active postgresql.conf file using grep.

PostgreSQL Starts But Loads the Old Data Directory

Symptom: SHOW data_directory; returns the old path after restart.

Cause 1: You updated the wrong postgresql.conf file. This happens when multiple PostgreSQL versions are installed or when you edited a file under a different version directory.

Cause 2: The change was not saved before restarting PostgreSQL.

Fix: confirm which config file PostgreSQL is using:

sudo -u postgres psql -c "SHOW config_file;"

Then verify data_directory inside that exact file.

Connection Refused After Migration

Symptom: psql or applications cannot connect, even though PostgreSQL appears to be running.

Cause 1: pg_hba.conf in the new data directory differs from what your environment expects, or PostgreSQL authentication settings were not copied as expected.

Cause 2: The socket directory configuration changed, or the client is connecting to the wrong socket path.

Fix: check PostgreSQL logs:

sudo journalctl -u postgresql --since "10 minutes ago"

Also check the configured Unix socket directories:

sudo -u postgres psql -c "SHOW unix_socket_directories;"

Use these Q&A entries to handle common variations and failure modes during a data directory migration.

Q: How do I move a PostgreSQL data directory to a new location on Ubuntu?

A: Stop PostgreSQL so the data files do not change during the copy. Copy the stopped data directory to the new mount point with rsync, then update the active postgresql.conf data_directory value to the new path. Add an AppArmor local rule for that path, start PostgreSQL again, and verify with psql that SHOW data_directory; reports the new directory.

Q: Where is PostgreSQL data stored on Ubuntu by default?

A: PostgreSQL stores cluster data under /var/lib/postgresql/<version>/main. The <version> segment matches the PostgreSQL major version installed on your system. You can confirm the exact directory with SHOW data_directory; in psql.

Q: How do I change PostgreSQL’s data directory on Linux?

A: Edit postgresql.conf and change the data_directory directive to the new absolute path. Restart PostgreSQL so it reads the updated configuration and starts using the migrated directory.

Q: How do I set the data directory path for PostgreSQL in Linux?

A: Set data_directory in postgresql.conf to the absolute path of the migrated cluster directory. The directive typically looks like data_directory = '/mnt/your-mount/postgresql/<version>/main'. Substitute your mount point and PostgreSQL major version.

Q: Why does PostgreSQL fail to start after moving the data directory?

A: The most common cause on Ubuntu is an AppArmor profile restriction that does not allow PostgreSQL to access the new path. Check Step 6 first, then verify ownership and permissions on the new directory. Also confirm that data_directory in the active postgresql.conf matches the real path on disk.

Q: Can I use a symbolic link instead of editing postgresql.conf?

A: Yes, but you still need to account for AppArmor. AppArmor applies to the real target path that the filesystem resolves, so a symlink does not bypass path restrictions. If your profile only allows the default directory, PostgreSQL can still fail until you update the AppArmor local rule.

Q: Is rsync the best tool for copying a PostgreSQL data directory?

A: rsync is typically the best choice because it preserves permissions and special metadata while copying recursively. Since you stop PostgreSQL first, rerunning rsync after interruptions converges on the same target state without needing to recopy everything from scratch.

Q: Do I need to reinitialize the database after moving the data directory?

A: No. Moving an existing data directory does not require initdb because PostgreSQL is already initialized inside the migrated directory. initdb is for creating a brand-new cluster, not for relocating an existing one.

After you stop PostgreSQL, copy the data directory to the new mount point, update postgresql.conf, and update AppArmor, PostgreSQL should start cleanly from the new path. Verify the migration with pg_lsclusters and psql, then keep the .bak directory until you confirm stable application behavior.

For related migration patterns, see the tutorial collection at How to Move a PostgreSQL Data Directory to a New Location.