In-place Postgres major version upgrades

You can upgrade a PGD node to a newer major version of Postgres using the command-line utility bdr_pg_upgrade.

bdr_pg_upgrade internally uses the standard pg_upgrade

with PGD-specific logic to ensure a smooth upgrade.

Terminology

This terminology is used when describing the upgrade process and components involved:

Oldcluster — The existing Postgres cluster node to upgrade, the one from which to migrate data.

Newcluster — The new Postgres cluster that data is migrated to. This cluster node must be one major version ahead of the old cluster.

Precautions

Standard Postgres major version upgrade precautions apply, including the fact both clusters must meet all the requirements for pg_upgrade .

Additionally, don’t use bdr_pg_upgrade if other tools are using replication slots and replication origins. Only PGD slots and origins are restored after the upgrade.

You must meet several prerequisites for bdr_pg_upgrade:

  • Disconnect applications using the old cluster. You can, for example, redirect them to another node in the cluster.

  • Configure peer authentication for both clusters. bdr_pg_upgrade requires peer authentication.

  • PGD versions on both clusters must be exactly the same and must be version 4.1.0 or later.

  • The new cluster must be in a shutdown state.

  • You must install PGD packages in the new cluster.

  • The new cluster must be already initialized and configured as needed to match the old cluster configuration.

  • Databases, tables, and other objects must not exist in the new cluster.

We also recommend having the old cluster up prior to running bdr_pg_upgrade. The CLI starts the old cluster if it’s shut down.

Usage

To upgrade to a newer major version of Postgres, you must first install the new version.

bdr_pg_upgrade command-line

bdr_pg_upgrade passes all parameters to pg_upgrade. Therefore, you can specify any parameters supported by pg_upgrade .

Synopsis

bdr_pg_upgrade [OPTION] ...

Options

In addition to the options for pg_upgrade, you can pass the following parameters to bdr_pg_upgrade:

Requiredparameters

Specify these parameters either in the command line or, for all but the --database parameter, in their equivalent environment variable. They’re used by bdr_pg_upgrade.

  • -b, --old-bindir — Old cluster bin directory.

  • -B, --new-bindir — New cluster bin directory.

  • -d, --old-datadir — Old cluster data directory.

  • -D, --new-datadir — New cluster data directory.

  • --database — PGD database name.

Optionalparameters

These parameters are optional and are used by bdr_pg_upgrade.

  • -p, --old-port — Old cluster port number.

  • -s, --socketdir — Directory to use for postmaster sockets during upgrade.

  • --check — Specify to only perform checks and not modify clusters.

Otherparameters

Any other parameter that’s not one of the above is passed to pg_upgrade. pg_upgrade accepts the following parameters:

  • -j, --jobs — Number of simultaneous processes or threads to use.

  • -k, --link — Use hard links instead of copying files to the new cluster.

  • -o, --old-options — Option to pass to old postgres command. Multiple invocations are appended.

  • -O, --new-options — Option to pass to new postgres command. Multiple invocations are appended.

  • -N, --no-sync — Don’t wait for all files in the upgraded cluster to be written to disk.

  • -P, --new-port — New cluster port number.

  • -r, --retain — Retain SQL and log files even after successful completion.

  • -U, --username — Cluster’s install user name.

  • --clone — Use efficient file cloning.

Environment variables

You can use these environment variables in place of command line parameters:

  • PGBINOLD — Old cluster bin directory.

  • PGBINNEW — New cluster bin directory.

  • PGDATAOLD — Old cluster data directory.

  • PGDATANEW — New cluster data directory.

  • PGPORTOLD — Old cluster port number.

  • PGSOCKETDIR — Directory to use for postmaster sockets during upgrade.

Example

Given a scenario where:

  • Old cluster bin directory is /usr/lib/postgresql/13/bin .

  • New cluster bin directory is /usr/lib/postgresql/14/bin .

  • Old cluster data directory is /var/lib/postgresql/13/main .

  • New cluster data directory is /var/lib/postgresql/14/main .

  • Database name is bdrdb .

You can use the following command to upgrade the cluster:

bdr_pg_upgrade \
- -old-bindir /usr/lib/postgresql/13/bin \
- -new-bindir /usr/lib/postgresql/14/bin \
- -old-datadir /var/lib/postgresql/13/main \
- -new-datadir /var/lib/postgresql/14/main \
- -database bdrdb

Steps performed

These steps are performed when running bdr_pg_upgrade.

Note

When –check is supplied as an argument to bdr_pg_upgrade, the CLI skips steps that modify the database.

PGD Postgres checks

Steps

–check supplied

Collecting pre-upgrade new cluster control data

run

Checking new cluster state is shutdown

run

Checking PGD versions

run

Starting old cluster (if shutdown)

skip

Connecting to old cluster

skip

Checking if bdr schema exists

skip

Turning DDL replication off

skip

Terminating connections to database.

skip

Disabling connections to database

skip

Waiting for all slots to be flushed

skip

Disconnecting from old cluster

skip

Stopping old cluster

skip

Starting old cluster with PGD disabled

skip

Connecting to old cluster

skip

Collecting replication origins

skip

Collecting replication slots

skip

Disconnecting from old cluster

skip

Stopping old cluster

skip

pg_upgrade steps

Standard pg_upgrade steps are performed.

Note

If supplied, –check is passed to pg_upgrade.

PGD post-upgrade steps

Steps

–check supplied

Collecting old cluster control data

skip

Collecting new cluster control data

skip

Advancing LSN of new cluster

skip

Starting new cluster with PGD disabled

skip

Connecting to new cluster

skip

Creating replication origin, repeated for each origin

skip

Advancing replication origin, repeated for each origin

skip

Creating replication slot, repeated for each slot

skip

Stopping new cluster

skip