Upgrading PGD clusters manually v5
Because EDB Postgres Distributed consists of multiple software components, the upgrade strategy depends partially on the components that are being upgraded.
In general, you can upgrade the cluster with almost zero downtime by using an approach called rolling upgrade. Using this approach, nodes are upgraded one by one, and the application connections are switched over to already upgraded nodes.
You can also stop all nodes, perform the upgrade on all nodes, and only then restart the entire cluster. This approach is the same as with a standard PostgreSQL setup. This strategy of upgrading all nodes at the same time avoids running with mixed versions of software and therefore is the simplest. However, it incurs downtime and we don't recommend it unless you can't perform the rolling upgrade for some reason.
To upgrade an EDB Postgres Distributed cluster:
- Plan the upgrade.
- Prepare for the upgrade.
- Upgrade the server software.
- Check and validate the upgrade.
There are broadly two ways to upgrade each node:
- Upgrade nodes in place to the newer software version. See Rolling server software upgrades.
- Replace nodes with ones that have the newer version installed. See Rolling upgrade using node join.
You can use both of these approaches in a rolling manner.
While the cluster is going through a rolling upgrade, mixed versions of software are running in the cluster. For example, suppose nodeA has PGD 3.7.16, while nodeB and nodeC has 4.1.0. In this state, the replication and group management uses the protocol and features from the oldest version (3.7.16 in this example), so any new features provided by the newer version that require changes in the protocol are disabled. Once all nodes are upgraded to the same version, the new features are enabled.
Similarly, when a cluster with WAL-decoder-enabled nodes is going through a rolling upgrade, WAL decoder on a higher version of PGD node produces LCRs (logical change records) with a higher pglogical version. WAL decoder on a lower version of PGD node produces LCRs with lower pglogical version. As a result, WAL senders on a higher version of PGD nodes are not expected to use LCRs due to a mismatch in protocol versions. On a lower version of PGD nodes, WAL senders may continue to use LCRs. Once all the PGD nodes are on the same PGD version, WAL senders use LCRs.
A rolling upgrade starts with a cluster with all nodes at a prior release. It then proceeds by upgrading one node at a time to the newer release, until all nodes are at the newer release. There must be no more than two versions of the software running at the same time. An upgrade must be completed, with all nodes fully upgraded, before starting another upgrade.
An upgrade process can take more time when caution is required to reduce business risk. However, we don't recommend running mixed versions of the software indefinitely.
While you can use a rolling upgrade for upgrading a major version of the software, we don't support mixing PostgreSQL, EDB Postgres Extended, and EDB Postgres Advanced Server in one cluster. So you can't use this approach to change the Postgres variant.
Downgrades of the EDB Postgres Distributed aren't supported. They require that you manually rebuild the cluster.
A rolling upgrade is where the server software upgrade is upgraded sequentially on each node in a cluster without stopping the the cluster. Each node is temporarily stopped from partiticpating in the cluster and its server software upgraded. Once updated, it is returned to the cluster and it then catches up with the cluster's activity during its absence.
The actual procedure depends on whether the Postgres component is being upgraded to a new major version.
During the upgrade process, you can switch the application over to a node that's currently not being upgraded to provide continuous availability of the database for applications.
The other method to upgrade the server software is to join a new node to the cluster and later drop one of the existing nodes running the older version of the software.
For this approach, the procedure is always the same. However, because it includes node join, a potentially large data transfer is required.
Take care not to use features that are available only in the newer Postgres version until all nodes are upgraded to the newer and same release of Postgres. This is especially true for any new DDL syntax that was added to a newer release of Postgres.
bdr_init_physical makes a byte-by-byte copy of the source node
so you can't use it while upgrading from one major Postgres version
to another. In fact, currently
bdr_init_physical requires that even the
PGD version of the source and the joining node be exactly the same.
You can't use it for rolling upgrades by way of joining a new node method. Instead, use a logical join.
Upgrading a CAMO-enabled cluster requires upgrading CAMO groups one by one while disabling the CAMO protection for the group being upgraded and reconfiguring it using the new commit scope-based settings.
We recommended the following approach for upgrading two BDR nodes that constitute a CAMO pair to PGD 5.0:
offfor transactions on any of the two nodes, or redirect clients away from the two nodes. Removing the CAMO pairing while attempting to use CAMO leads to errors and prevents further transactions.
- Uncouple the pair by deconfiguring CAMO either by resetting
bdr.camo_parter_of(when upgrading from BDR 3.7.x) or by using
bdr.remove_camo_pair(on BDR 4.x).
- Upgrade the two nodes to PGD 5.0.
- Create a dedicated node group for the two nodes and move them into that node group.
- Create a commit scope for this node group and thus the pair of nodes to use CAMO.
- Reactivate CAMO protection again by either setting a
default_commit_scopeor by changing the clients to explicitly set
bdr.enable_camofor their sessions or transactions.
- If necessary, allow clients to connect to the CAMO protected nodes again.
Each major release of the software contains several changes that might affect compatibility with previous releases. These might affect the Postgres configuration, deployment scripts, as well as applications using PGD. We recommend considering these changes and making any needed adjustments in advance of the upgrade.
See individual changes mentioned in the release notes and any version-specific upgrade notes.
Upgrading EDB Postgres Distributed on individual nodes happens in place. You don't need to back up and restore when upgrading the BDR extension.
The BDR extension upgrade process consists of a few steps.
During the upgrade of binary packages, it's usually best to stop the running Postgres server first. Doing so ensures that mixed versions don't get loaded in case of an unexpected restart during the upgrade.
The first step in the upgrade is to install the new version of the BDR packages. This installation installs both the new binary and the extension SQL script. This step is specific to the operating system.
Once packages are upgraded, you can start the Postgres instance. The BDR extension is upgraded upon start when the new binaries detect the older version of the extension.
The process of in-place upgrade of Postgres depends on whether you're upgrading to new minor version of Postgres or to a new major version of Postgres.
Upgrading to a new minor version of Postgres is similar to upgrading the BDR extension. Stopping Postgres, upgrading packages, and starting Postgres again is typically all that's needed.
However, sometimes more steps, like reindexing, might be recommended for specific minor version upgrades. Refer to the release notes of the version of Postgres you're upgrading to.
Upgrading to a new major version of Postgres is more complicated than upgrading to a minor version.
EDB Postgres Distributed provides a
bdr_pg_upgrade command line utility,
which you can use to do in-place Postgres major version upgrades.
When upgrading to a new major version of any software, including Postgres, the BDR extension, and others, it's always important to ensure the compatibility of your application with the target version of the software you're upgrading.
After you upgrade your PGD node, you can verify the current version of the binary:
Always check your monitoring after upgrading a node to confirm that the upgraded node is working as expected.
HARP can temporarily coexist with the new connection management configuration. This means you can:
- Upgrade a whole pre-5 cluster to a PGD 5 cluster.
- Set up the connection routing.
- Replace the HARP Proxy with PGD Proxy.
- Move application connections to PGD Proxy instances.
- Remove the HARP Manager from all servers.
We strongly recommend doing this as soon as possible after upgrading nodes to PGD 5. HARP isn't certified for long-term use with PGD 5.
TPA provides some useful tools for this and will eventually provide a single-command upgrade path between PGD 4 and PGD 5.